Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Mathematics and Computers in Simulation 61 (2002) 1–15

Simulation and fault detection of three-phase induction motors


B. Liang a,∗ , B.S. Payne b,1 , A.D. Ball b , S.D. Iwnicki a
a
Department of Engineering and Technology, Manchester Metropolitan University, Manchester M1 5GD, UK
b
School of Engineering, University of Manchester, Manchester M13 9PL, UK

Abstract
Computer simulation of electric motor operation is particularly useful for gaining an insight into their dynamic
behaviour and electro-mechanical interaction. A suitable model enables motor faults to be simulated and the change
in corresponding parameters to be predicted without physical experimentation. This paper presents both a theoretical
and experimental analysis of asymmetric stator and rotor faults in induction machines. A three-phase induction motor
was simulated and operated under normal healthy operation, with one broken rotor bar and with voltage imbalances
between phases of supply. The results illustrate good agreement between both simulated and experimental results.
© 2002 IMACS. Published by Elsevier Science B.V. All rights reserved.
Keywords: Fault simulation; Mathematical models; Condition monitoring; Induction motors

1. Introduction

Induction motors are critical components in many industrial processes. In spite of their robustness they
do occasionally fail and their resulting unplanned downtime can prove very costly. Therefore, condition
monitoring of electrical machines has received considerable attention in recent years. There are many ways
to detect mechanical and electrical problems in induction motors, either directly or indirectly. Directly,
many parameters can be monitored to provide useful indications of incipient faults. These parameters
include case vibration and noise, stator phase current, air-gap or external magnetic flux density. Hargis
et al. showed that mechanical defects are detectable through variations in the vibration of the stator core
[1–6]. It has been suggested by Steele that per-phase current monitoring can provide similar indications.
The advantages of per-phase current monitoring are firstly, it is easy to measure; secondly, the fault
patterns in the current signal are unique and thirdly, an estimation of the number of broken bars can be
made from the analysis [7–12]. Stephan et al. [13] used air-gap search coils to detect and estimate both
the peripheral and radial component of the leakage flux. An axial flux sensing method to detect rotor
short circuits and other malfunctions has also been described by Penman et al. [14]. Indirectly, modelling

Corresponding author. Tel.: +44-161-2471667; fax: +44-161-2471663.
E-mail addresses: b.liang@mmu.ac.uk (B. Liang), bradley.payne@man.ac.uk (B.S. Payne).
1
Tel.: +44-161-2754315.

0378-4754/02/$ – see front matter © 2002 IMACS. Published by Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 8 - 4 7 5 4 ( 0 2 ) 0 0 0 6 4 - 2
2 B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15

and simulation of electrical machine operation under healthy and faulty conditions will also provide
useful information for fault prediction and identification. Computer simulation of motor operation can be
particularly useful in gaining an insight into their dynamic behaviour and electro-mechanical interaction.
With a suitable model, motor faults may be simulated and the change in corresponding parameters can be
simulated. It has been found that an accurate motor simulation can be achieved with a general model and
accurate parameter selection, in addition to using a sophisticated model [15,16]. This can significantly
reduce the computer simulation time and make model-based condition monitoring more reliable and
easily achievable. Following the literature review, it has been found that this area has received only little
attention. Although a few papers exist on simulation and fault diagnosis of induction motors [17], no
asymmetric stator and rotor fault simulation of induction motors have simultaneously been made to date.
A review of literature also revealed that the variation in the rotor transient speed is thought to be insensitive
to fault conditions.
This paper presents both a theoretical and experimental analysis of asymmetric stator and rotor faults
in electric motors. A three-phase induction motor was simulated and operated under normal healthy

Fig. 1. Simulation of a healthy motor during start-up: (a) speed; (b) electromagnetic torque; (c) stator phase-A current; (d) stator
phase-B current; (e) stator phase-C current.
B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15 3

operation, with one broken rotor bar and with voltage imbalances on one phase of supply. In partic-
ular, the rotor transient speed variation was been selected as one of the main parameters to be moni-
tored.

2. Induction motor modelling and simulation

This research aimed to investigate whether induction motor operational simulations with general equa-
tions could be used for condition monitoring and fault diagnosis of induction motors. Therefore, a theo-
retical motor analysis was made based on generalised rotating field theory and by making the following
assumptions which are commonly regarded as appropriate [18]:
• The magnetic permeability of iron is considered to be infinite and the air-gap is very small and
smooth.
• The state of operation remains far from magnetic saturation.
• The self-inductances and mutual-inductances between stator or rotor phases are constant.
• Mutual-inductances between the stator and rotor windings are functions of the rotor position.
• Space magnetic motive force (MMF) and flux profiles are considered to be sinusoidally distributed and
higher harmonics are negligible.

Fig. 2. Stator current waveforms from simulated and measured results during motor start-up.
4 B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15
    
VsA Rs + pLss pMs pMs pMsr cos θr pMsr cos θr1 pMsr cos θr2 IsA
 VsB   pMs Rs + pLss pMs pMsr cos θr2 pMsr cos θr pMsr cos θr1   IsB 
    
 VsC   pMs pMs Rs + pLss pMsr cos θr1 pMsr cos θr2 pMsr cos θr   IsC 
 =   (1)
 Vra   pMsr cos θr pMsr cos θr2 pMsr cos θr1 Rr + pLrr   Ira 
   pMr pMr  
 Vrb   pMsr cos θr1 pMsr cos θr pMsr cos θr2 pMr Rr + pLrr pMr   Irb 
Vrc pMsr cos θr2 pMsr cos θr1 pMsr cos θr pMr pMr Rr + pLrr Irc

here IsA , IsB , IsC are the stator three-phase currents, Ira , Irb , Irc the rotor three-phase currents, Lss , Lrr
the stator and rotor self-inductances, Ms , Mr the stator or rotor phase mutual inductance, Msr the
stator and rotor mutual inductance, p the differential operator, Rs , Rr the stator and rotor resistances,
VsA , VsB , VsC the stator three-phase voltages, Vra , Vrb , Vrc the rotor three-phase voltages, θr the rotor
angle (θr1 = θr + 2π/3, θr2 = θr + 4π/3).
• The rotor bars are electrically insulated as far as the iron surfaces are concerned.

2.1. Induction motor modelling

The necessary electrical model of the three-phase induction motor was obtained using well-docu-
mented motor models [18]. The matrix form of the stator and rotor voltage equations is shown in
Eq. (1).

Fig. 3. Rotor speeds from simulated and measured results during motor start-up.
B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15 5

The electromagnetic torque Te is given by


Te = 23 Msr (Iqs Idr − Ids Iqr ) (2)

The mechanical model can be created by


dωr
J = Te − T l − T f (3)
dt
where Ids , Iqs are the stator d, q phase currents, Idr , Iqr the rotor d, q phase currents, J the inertia of the
rotor, Tl the applied load, Tf the frictional torque loss, ωr the rotor angular velocity.
Digital simulation was implemented in the MATLAB/SIMULINK software package. Fig. 1 presents
the simulation results in rotor speed, electromagnetic torque and three-phase stator currents for a healthy
motor. It can be observed that the rotor speed, electromagnetic torque and three-phase stator currents
become steady after initial fluctuation when the motor runs up. Figs. 2 and 3 compare stator current and
rotor speed from simulated and measured results during motor start-up. They illustrate good agreement
between simulated and measured results has been achieved. It therefore indicates the model and the
selected parameters are appropriate.

Fig. 4. Simulation of 20 V drop in one phase of voltage supply: (a) speed; (b) electromagnetic torque; (c) stator phase-A current;
(d) stator phase-B current; (e) stator phase-C current.
6 B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15

Fig. 5. Comparison of simulated and measured rotor speed for 20 V drop in one phase of voltage supply.

Fig. 6. Measured three-phase stator currents for 20 V drop in one phase of voltage supply.
B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15 7

In order to validate the simulation results an experimental induction motor test rig was developed for
experimentation. The test rig comprises a 3 kW induction motor, a loading DC motor and a resistor bank
to dissipate the electrical energy generated. The instrumentation consisted of accelerometers, Hall effect
current transducers, a high resolution speed encoder (360 pulses per cycle), amplifiers, filters and a DIFA
210 data acquisition system.

2.2. Simulation of imbalanced stator voltage supply faults

Phase imbalance faults of 20 and 40 V drops in one phase of supply were simulated. In order to simulate
imbalanced supply faults little modification to the model was required. The stator voltage supply equations
were changed as follows:
√ √   √  
VsA = 2(E − E0 ) cos ωt, VsB = 2E cos ωt − 23 π , VsC = 2E cos ωt + 23 π (4)
here E is the rated voltage, E0 = 20 or 40 V.
The 20 V drop in one phase condition was simulated and Fig. 4 illustrates the simulated rotor speed,
torque and three-phase stator current results. It can be observed that there is rotor speed and torque

Fig. 7. Simulation of 40 V drop in one phase of voltage supply: (a) speed; (b) electromagnetic torque; (c) stator phase-A current;
(d) stator phase-B current; (e) stator phase-C current.
8 B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15

Fig. 8. Comparison of simulated and measured rotor speed for 40 V drop in one phase of voltage supply.

ripple and there is a noticeable drop in the amplitude of the phase-A current. The ripple frequency
is at 100 Hz, which is two times the UK supply frequency of 50 Hz. This is the typical symptom
of an asymmetric stator fault because the imbalanced stator phase voltages typically result in torque
pulsing and rotor speed fluctuation [17]. Fig. 5 presents a comparison of the measured and simu-
lated rotor speed. As expected, rotor speed fluctuation is clearly visible in the measured data from
the test motor. Although both simulated and measured results are similar, the measured speed ripple
is larger than the simulated speed ripple by about 30 rpm. This could be caused by approximations
used in the model as it only considers the fundamental frequency and non-saturation conditions. Fig. 6
shows experimental three-phase stator currents and the corresponding current decrease in phase-A can be
observed.
Simulation of a 40 V drop in one phase of voltage supply is illustrated in Fig. 7. It can be seen
that there are larger rotor speed and torque fluctuations (at 100 Hz) and a larger decrease in phase-A
current compared with Fig. 4. Fig. 8 provides a comparison of measured and simulated rotor speed.
As for the 20 V imbalance, a larger rotor speed ripple can also be seen in the measured results than
in the simulated results. Although the speed ripple difference between the measured and simulated
results increases to about 50 rpm the frequency is still at 100 Hz. Fig. 9 represents the mea-
sured three-phase stator currents in which a further reduction in the phase-A current can be
observed.
B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15 9

Fig. 9. Measured three-phase stator currents for 40 V drop in one phase of voltage supply.

2.3. Simulation of a single broken rotor bar fault

In order to simulate a broken rotor bar, the mathematical motor model used for simulation of imbal-
anced stator voltages was further modified with respect to each rotor bar’s resistance and inductance
effects. Fig. 10 presents a simple mesh model of a squirrel-cage rotor [16]. For a rotor having Nr bars and
two end rings, there are Nr + 2 loops and 2Nr nodes. This is because Nr bars will have Nr loops. If the
two end rings are considered as another two loops, there will be Nr + 2 loops. The current distribution

Fig. 10. The rotor bar mesh model.


10 B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15

can therefore be specified in terms of Nr + 2 independent rotor currents which flow in loops compris-
ing of two adjacent rotor bars and two end ring segments that join them. Each rotor bar and end ring
segment is characterised by a resistance Rbn (or Ren ) and an inductance Lrn,rn associated with the rotor
loop. In order to get interaction between the stator and rotor the bar current ibar,n = irn+1 − irn = 0
under normal condition. If the bar is broken, then ibar,n = irn+1 − irn = 0, so that irn+1 = irn .
This means that a broken rotor bar can be simulated by forcing the current flowing in that particular
bar to 0.
The following expression can be used to simplify the mathematical model for the voltage equations of
the induction motor [18]:

dψ dL dI
V = RI + = R+ I +L (5)
dt dt dt

here V = [ VsA VsB VsC 0 0 · · · 0 ], I = [ IsA IsB IsC ir1 ir2 · · · irNr ie ],

Fig. 11. Simulation of a broken rotor bar fault: (a) speed; (b) electromagnetic torque; (c) stator phase-A current; (d) stator phase-B
current; (e) stator phase-C current.
B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15 11

and
 
RsA 0 0 0 0 0 0 0
 0 RsB 
 0 0 0 0 ··· 0 
 
 0 0 RsC 0 0 0 ··· 0 
 
 0 Rl −Rb1 ··· −Rbn Ren 
 0 0 
R= ,
 0 0 0 −Rb1 Rl ··· ··· Ren 
 
 ··· ··· ··· ··· ··· ··· ··· ··· 
 
 
 0 0 0 −Rbn ··· ··· Rl Ren 
0 0 0 Ren Ren Ren Ren Nr × Rbn

Fig. 12. Comparison of simulated and measured results in one stator phase current: (a) measured stator phase-A current; (b)
simulated stator phase-A current.
12 B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15
 
LsA,sA MsA,sB MsA,sC Mr1,sA Mr2,sA ··· MrNr ,sA 0
 
 MsA,sB LsB,sB MsB,sC Mr1,sB Mr2,sB ··· MrNr ,sB 0 
 
 
 MsA,sC MsB,sC LsC,sC Mr1,sC Mr2,sC ··· MrNr ,sC 0 
 
 
 Mr1,sA Mr1,sB Mr1,sC Lr1,r1 Mr1,r2 ··· Mr1,rNr Len 
L=



 Mr2,sA Mr2,sB Mr2,sC Mr1,r2 Lr2,r2 ··· Mr2,rNr Len 
 
 
 ··· ··· ··· ··· ··· ··· ··· Len 
 
 
 MrNr ,sA MrNr ,sB MrNr ,sC Mr1,rNr Mr2,rNr · · · LrNr ,rNr Len 
0 0 0 Len Len ··· Len Nr × Len
here irn is the nth rotor loop current, Len the nth end ring self-inductance, Nr the number of rotor bars,
Rbn , Ren the nth bar and end ring resistance and ψ the flux linkage.
The simulation results for a single broken rotor bar are presented in Fig. 11. It shows that a low frequency
of 2sω (s is the rotor slip and ω the supply frequency) modulation exists in rotor speed, electromagnetic
torque and three-phase stator currents. This symptom is the typical fault indication for a broken rotor bar
fault [7]. Figs. 12 and 13 provide a comparison of simulated and measured results for one phase stator
current and rotor speed. The measured results illustrate slight modulation in the stator per-phase current

Fig. 13. Comparison of simulated and measured rotor speeds for a broken rotor bar fault.
B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15 13

Fig. 14. Rotor speed spectra for healthy and faulty conditions: (a) normal condition; (b) broken one rotor bar; (c) 20 V drop of
one phase voltage; (d) 40 V drop of one phase voltage.

and greater modulation in rotor speed with the introduction of this type of fault. These differences are
assumed to be caused by the assumptions used in the model and by slight errors in setting particular
parameter values. For example, one of assumptions was that the rotor bar current is 0 if a bar is broken.
In practice, however, there may be a current path through the laminations between the adjacent bars. This
effect would lead to a decrease in the stator phase current modulation and therefore could account for the
differences between the simulated and measured stator phase currents. Similarly, it is difficult to estimate
exact values for the model parameters (such as bar resistance, self-inductance and mutual-inductance).
Errors in the setting of these parameters could also contribute to the differences between simulated and
measured results.
Finally, Fig. 14 presents the measured rotor speed spectra for the induction motor under healthy and
faulty conditions. It is clearly seen that symptoms for both imbalanced stator voltage supply (2ω =
100 Hz) and a broken rotor bar (2sω = 4.6 Hz) exist in rotor speed spectra.

3. Conclusions

This paper presented the successful simulation and condition monitoring of induction motors with
traditional electrical machine models. It has been demonstrated that a generalised motor model can be
used to simulate induction motor faults to a high degree of accuracy.
14 B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15

The simulated results showed that a clear typical fault symptom of 100 Hz ripple for asymmetrical
stator voltage supply can be found in both electro-magnetic torque and rotor speed. The ripple amplitude
increases with increasing asymmetric stator supply. Experimental results verified the simulations but
showed higher rotor speed ripple amplitudes when compared with the simulated results (with a 30–50 rpm
error).
Another modified model for asymmetric rotor faults (such as broken rotor bars) was also constructed
and examined. The simulated results illustrated noticeable 2sω modulations in stator phase current,
electromagnetic torque and rotor speed. The experimental results also showed the symptoms but with
much less phase current modulation and higher fluctuations in rotor speed than the simulated results. The
reason for the phase current discrepancy may be related to the assumptions and approximations of the
modified model for broken rotor bar.
The simulated and experimental results indicate that model-based fault detection and diagnosis is useful.
The simulation showed that the dynamic characteristics of induction motors under different conditions
can be obtained purely by computer simulation. With a suitable model, motor faults may be simulated
and predicted without any experimental analysis.

References

[1] C. Hargis, B. Gaydon, K. Kamash, The detection of rotor defects in induction motors, in: Proceedings of the International
Conference, Vol. 313, July 1982, IEE Publication, pp. 216–220.
[2] E. Erdelyi, P.A. Erie, Vibration modes of states of induction motors, Transactions of ASME, Paper A-28, 1956,
pp. 39–45.
[3] I.W. Mayes, A.G. Steer, G.B. Thomas, The application of vibration monitoring for fault diagnosis in large turbo-generators,
in: Proceedings of the Sixth Thermal Generation Specialists Meeting, Madrid, May 1981, pp. 567–575.
[4] R.G. Herbert, Computer techniques applied to the routine analysis of rundown vibration data for condition monitoring of
turbine-alternators, in: Proceedings of the International Conference on Condition Monitoring, Swansea, UK, April 1984,
pp. 229–242.
[5] J.R. Cameron, W.T. Thompson, A.B. Row, Vibration and current monitoring for detecting air gap eccentricity in large
induction motors, in: Proceedings of the International Conference on Electrical Machines, Design and Applications, Vol.
254, September 1982, IEEE, London, pp. 173–179.
[6] J.H. Maxwell, Induction motor magnetic vibration, in: Proceedings of the Vibration Institute, Machinery Vibration
Monitoring and Analysis Meeting, Houston, Texas, 1983, pp. 213–218.
[7] M.E. Steele, R.A. Ashen, L.G. Knight, An electrical method for condition monitoring of motors, in: Proceedings of
the International Conference of Electrical Machines on Design and Application, July 1982, IEE, London, pp. 231–
235.
[8] J.B. Kliman, J. Stein, Induction motor fault detection via passive current monitoring, in: Proceedings of the International
Conference on Electrical Machinery, Cambridge, August 1990, pp. 13–17.
[9] J. Stefin, et al., Methods of motor current signature analysis, Electrical Mach. Power Syst. 20 (5) (1992) 463–474.
[10] R.S. Randy, et al., Motor bearing damage detection using stator current monitoring, IEEE Trans. Ind. Appl. 31 (6) (1995).
[11] David, R. Rankin, The industrial application of phase current analysis to detect rotor winding faults in squirrel cage induction
motors, Power Eng. J. 23 (2) (April 1995).
[12] R.R. Schoen, An unsupervised, on-line system for induction motor fault detection using stator current monitoring, IEEE
Trans. Ind. Appl. 31 (6) (1995) 1281–1286.
[13] J. Stephan, M. Bodson, J. Chiasson, Real-time estimation of the parameters and fluxes of induction motors, IEEE Trans.
Ind. Appl. 30 (3) (1994) 746–758.
[14] Penman, J., Dye, M.A.N., Tit, A.J., Bryan W.E., Condition monitoring of electrical drives, in: Proceedings of the IEE, Vol.
133, Part B, No. 3, May 1986, pp. 142–148.
B. Liang et al. / Mathematics and Computers in Simulation 61 (2002) 1–15 15

[15] T.A. Lipo, A. Consoli, Modelling and simulation of induction motors with saturable leakage reactance, IEEE Trans. Ind.
Appl. IA-20 (1984) 180–189.
[16] E. Ritchie, X. Deng, T. Jokinen, Dynamic model of three-phase squirrel cage induction motors with rotor faults, in:
Proceedings of the ICEM, IEE Publication, 1994, pp. 694–698.
[17] M. Benbouzid, A review of induction motor signature analysis as a medium for fault detection, IEEE Trans. Ind. Electron.
47 (5) (2000) 984–993.
[18] P. Vas, Electrical Machines and Drives: A Space-Vector Theory Approach, Clarendon Press, Oxford, 1988.

You might also like