Sensorless Speed Measurement of Induction Motors
Roberto Micheletti
Department of Electrical Systems and Automation, University of Pisa
Via Diotisalvi. 2 1-56126 P i s , Italy
Abstract -The paper deals with the accurate slip
measurement of induction motors. The proposed procedirre
uses a non-invasive slip nieaszirement scheme based on
digitalfilferingand dvnamic parameter estimation. The slip
measurement is carried ouf M'ithout speed sensor and is
deduced analvzing the magnetic field harnionics spectruni
in pro1imit.v of the induction motor. First the enif induced
waveform, taken from U searching coil, is filtered rising
algorithms based on the discrete Fourier transform. Then
the stator frequencb and rotorfiequency are obtained b?
comparing thefiltered voltage with a mathematical model
using an optimization procedure. The model's parameters
are varied until an adequate match is obtained with the
filtered voltage. Experiniental resulrs are presented to
validate this method.
K@xQ& - Digital filtering, JndiJrtion mofors, Slip
measurement.
1. INTRODUCTION
Induction motor drives are now being more and more in
process industry because of the application of the field
oriented control strategy. However, the performance of such
control method depends strongly on the accuracy of the
motor parameters uscd in the vector controller.
It is well known that the variation of the rotor resistance
and rotor constant time has a most dominant effect on the
control performance. Unfortunately the rotor resistance
depends widely on the rotor temperature and on the slip
frequency, resulting in the deStruction of the decoupled
condition of the flux and torque.
In recent years many studies are then carried out to
overcome this situation. Estimation methods to get rotor
parameters use extended Kalman filter approach [I],
observer technical [2] and adaptive system [3].
It is actively proceeding to research for speedsensorless
vector control which estimates rotor speed and slip
0-7803-7864-4/03/$17.00 02003 IEEE.
Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on October 6, 2008 at 7:39 from IEEE Xplore. Restrictions apply.
frequency without speed sensor. The slip frequency
detection is carried out utilizing rotor slot harmonics [4], [ 5 ]
or sensing and exploiting the stator current [ 6 ] .
This paper presents an algorithm for the accurate
measurement of the slip frequency based on digital filtering
and dynamic parameter estimation method [7]-[9].
The slip frequency measurement is carried out without
speed sensor and is deduced analyzing the magnetic field
harmonics spectrum in proximity of the induction motor.
First the emf induced waveform, taken from a searching
coil, is filtered using algorithms based on the discrete
Fourier transform.
Then the stator frequency, rotor frequency and
consequentially the slip frequency are obtained by
comparing the filtered voltage with a mathematical model
using an optimization procedure. The model's parameters
are vaned until an adequate match is obtained with the
filtered voltage.
The parameters that affect the performance of the
algorithm are essentially the data window size, the sampling
rate and the characteristics of the filter.
Mathematical development of the algorithm is presented
and the effects of key parameters that affect the
performance of the algorithm are discussed.
A representative set of experimental results are
presented.
The proposed method can be reliably applied in rotor
parameter identification in steady state conditions and
running test.
11. ESTIMATION ALGORITHM OF STATOR
AND ROTOR FREQUENCY
The stator frequency f, and rotor frequency f, are
obtained with the same procedure by comparing the
respective filtered voltages with a mathematical model
using an optimization method.
67
 W I S P 2003, Budapest, Hunga y 4-6 Septeritber, 2003

After filtering operation, we obtain a set of "n" samples

-a=Eo . -a=Eo . -a=Eo (7)
of the stator (rotor) voltage. The filtered voltage can be AA ' aR ' aw
approximated by a sinusoid

There results after rearrangement

v(t)=Vsin(at+O) (1)

where V a n d 0 are amplitude and phase,respectively.

gy,"(f.\)-
4Ja1=
By rcwriting the voltage in terms of quadrature
components

"(1) = A sin wt + B cos wt ="(A, B, w, t) (2)

with

0 = tan -'(B/A)

Expanding v(t) in a Taylor series in the neighborhood

of given values A a , Bo, (rh of parameters A, B, w gives

where the higher order terms of the expansion are

ignored and n = (A0 , BO,%).
The principle of operation of the estimation technique is
based on the comparison between the real values of the
filtered voltage and the estimation values of the model.
Thus the problem consists in determining the parameters A,
B, w able to minimize the error between sampled values and
estimation values. The total square error, at instant 4. is
expressed as
Solution of (8) gives the corrections AA, AB and Aw
necessary for updating parameters A , B and w for each
iteration step. This recursive technique permits t o obtain the
unknown stator and rotor frequency with good accuracy.
Finally the slip frequency is given by
s = fJf (9)

Iterative methods are well known for their sensitivity to

the initially guessed values of the unknowns. The initial
values used for the model reference are determined as
follows. The initial value of o, is obtained using the first
where vr(ts) and v(ts) represent the sampled outputs five samp les of the input voltage
of the real system after filtering and the model reference at
time t,,respectively.
Substitution of (4) into ( 5 ) yields 3 [2v(tI+2T&v(t J-v(t, +4TJ]
(10)
? e- 2[v(t l+T J+43t ,+2TJ +v(t ,+3Tsd

The initial values of and & are determined solving the

system:
The total square error is minimized by solving the
partial derivatives of (6) relative to A, B, w evaluated
v ( t , ) = A a s i n w , t , + Bocosw,t,
at Ao.Bo,oh
(11)
v(tl +Ts)=Aosin[(rh(t, + T s ) l + B o c o s [ ~ ( t +Ts)l
,

68

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on October 6, 2008 at 7:39 from IEEE Xplore. Restrictions apply.
 WISP 2003, Budapest, Hiiiignry M September, 2003

111. RESULTS
The measurement algorithm has been verified to
investigate the validity of this technique.
The rated values of the induction motor used in the
experiments are given in Table I .
Table 1 Rated values of a tested induction motol
the emf component at the stator frequency (emf at stator
frequency component).
The second one is a lowpass filter with cutoff frequency
set to 2 Hz. The output of the lowpass filter is the emf
component at the rotor frequency (emf at rotor frequency
component).
The emf at stator and rotor frequency components with
motor loaded at 1480 rpm are shown in Fig. 2.

Power
Volta e Y
Current (U) 21.5 A
Fre uenc
Revolution er minute (rimin)

The slip frequency measurement is carried out analyzing

the waveform of the induced emf, taken from a searching
coil, due to the magnetic field in proximity of the induction
motor.
Fig. 1 shows the waveform of the induced emf with
motor loaded at 1480 rpm; the slip frequency is s=1.33%,
corresponding to a rotor frequency of 0.667 H z .
Useful estimates o f the slip frequency are obtained using ,,me ( 5 )
about 114 cycle of the emf at stator and rotor frequency,
therefore the proposed procedure could be compatible with
control purposes. Fig. 2.Emf at stator and rotor frequency with motor loaded at 1480
The searching coil is put on the frame of the induction rpm
motor.

Fig. 3 shows the waveform of the induced emf with

04
emf (V) motor loadedat 1470 rpm; the slip frequency is s=2.00%,
0.3

0.2 emf (v)

0.1

00

4.1

0.2

0.3

0.4
0 10 20 30
11818 d o 8 ab,. x.wr x (8Yf0,) em,
time (E)

Fig. 1. Waveform of the induced emf with motor loaded at 1480 0 10 20 30

(,,le a07 ad,. x - v a x (aYtO)) em,
m. xime I S )

Fig. 3. Waveform of the induced emf with motor loaded at 1470

Then this waveform is filtered using algorithms based
on the discrete Fourier transform.
The filtering operation is obtained with two eighth order
IIR filters. The first one is a bandpass filter with a center
frequency of 50 H z and bandwidth 4 Hz, whose output is
m.
The emf at stator and rotor frequency components with
motor loaded at 1470 rpm are shown in Fig.4.

69

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on October 6, 2008 at 7:39 from IEEE Xplore. Restrictions apply.
 WISP 2003, Budapest, Hungary 4-6 Septeinbev, 2003

s o l y ~ , . . , . . , . , Next pictures show the waveforms of the induced emf

and the emf at stator and rotor frequency components with,
motor loaded at 1450 rpm and slip frequency ~ 3 . 3 3 % .

emf ( V I
0.4, I I I I I I I I i

,,me (I1

Fig. 4. Emf at stator arc rotor frequency with motor loaded at 1470 4.44 I I I I I I I I I \
m. 0 in 20 30 40 :qo-3 50
(mem06adl, x."srX(a"to,l em,

Figure 5 and 6 show the waveforms of the induced emf ,"(SI

and the emf at stator and rotor frequency components with
motor loaded at 1460 rpm and slip frequency s=2.67%. Fig. 7. Waveform of the induced emf with motor loaded at 1450
run

Fig. 5. Waveform of the induced emf with motor loaded at 1460 rpm
-> 0 . j 1 . . . ~ . ~ . . ~ , 0 02 0 4 06 0 8
,\me
1
(S)
I2 14 1 5 18 2

Fig. 8. Emf at stator and rotor frequency with motor loaded at 1450
rpm

The delay introduced by the filtering operation is

approximately 0.5 s.
Tests have been carried out for different values of the
slip frequency in the range from s = 1.33% to s = 3.33%
with sampling frequency set to 800 Hz.
The elaboration for generating new values of A, B, and
-
: 4 0 , L " " ' ~ ' ' . 1
o at each iteration step is carried out on a data window of
about 1/4 cycle o f the respective filtered voltages; these
0 02 01 06 0 8 I 12 3 4 1 5 1 8 z
,,me (3) values o f sampling frequency and data window size have
Fig. 6. Emf at stator and rotor frequency with motor loaded at 1460 been selected in order to increase the speed of convergence
rpm and to improve the accuracy.

70

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on October 6, 2008 at 7:39 from IEEE Xplore. Restrictions apply.
 WISP 2003, Budapest, H u n g n y 4 4 September, 2003

The initial values ofparameters%, Ao,Bo are estimated

as mentioned in previous section; only six iteration steps are 0.040
required tu get to the convergence. 0.030
Results for the stator and rotor frequency and for the slip
measurement, in the range from s = 1.33% to s = 3.33%, are
E 0.020

shown in Fig.9 and Fig. IO, respectively 2 0.010

0.000
.. -0.010~
50.003
-0.020
5 50.002
6 50.001 Revolution per minute
50.000
E
-;
b
49.999
49.998
49.997
b) Percent error

1480 1470 1460 1450 Fig. 10. Result for the slip measurement in the range from
from 5 = 1.33% to s = 3.33%
Revolution per minute

The proposed system operated satisfactorily for any

a) Stator frequency value of the slip frequency in the previously mentioned
range; the error was in any case within + 0.03 Yo.
-
E.2 000 1 IV. CONCLUSION
5
c
1.500
0 A new digital approach for the accurate measurement o f
g 1.000 the slip frequency o f induction motors has been presented.
0
0.500 The non invasive procedure based on digital filtering
-b and dynamic parameter estimation has been shown to work
p 0.000 J effectively over a nearly wide range of speed and loading
1480 1470 1460 1450 conditions.
Revolution per minute The slip frequency measurcment is carried out without
speed sensor and is deduced utilizing the magnetic field in
proximity of the induction motor. The system apparatus
b) Rotor frequency consists of a searching coil, an ADC hoard and a PC; the
measurement system needs only a signal from a searching
Fig. 9. Result for the stator and rotor frequency in the range
from s = 1.33% to s = 3.33% coil (induced emf) to be digitally filtered using algorithms
based on the discrete Fourier transform. Useful estimates o f
the slip frequency are obtained using about 114 cycle of the
emf at stator and rotor frequency. Sampling rate, data
-e- theoretical -E- experimental window size and the characteristics of the filter are critical
3.500 parameters that affect the performance of the algorithm.
- 3.000
2.500
2.000
The accuracy of the slip measurement, in the previously
mentioned range o f the slip frequency, is not affected by
the load applied to the motor nor b y any consequent
-
0 1.500
changes to the motor parameters.
n1.000
l
The proposed technique is accurate, simple and l o w
0.500
0.000 1
1480 1470 1460 1450
cost; moreover it allows in-field measurement of the slip
frequency of induction motor even in hazardous
environments and it can be reliably applied in rotor
Revolution Per minute parameter identification in steady state conditions and
running test.
Experimental tests confirmed the validity of the
proposed procedure.
a) Slip measurement

71

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on October 6, 2008 at 7:39 from IEEE Xplore. Restrictions apply.
 WISP 2003, Budnprst, Hungti y 4-6 September, 2003

REFERENCES

[ I I Y.R. Kim. S.K.Sul, M.H. Park: Speed SCIISO~IUSS vcctor ~ o n t r u of

l
induction motor using extended Kalman. IEEE 7rons. On Indsrin.
Applicolions. Sep.-Oct. 1994. pp. 1225-1233.
[2] H . Kuhota K. Matrurr. T. Nakano: DSP -based speed adaptive tlux
ahserver of induction motor. IEEE T".OPII n d t r r t ~Applicolioni.
~ Mar:
Apr. 1993. pp. 344-348.
[?I G. Yang. T.H. Chin: Adaptive-speed identification scheme fora vector-
controlled speed sensorless invener-induction motor. IEEE Trans. On
Indnstrv Applicolions. Jul.-Aue. 1993. pp. 820-825.
[4] M. Ishida. K. Iwata: A new slip frequqncy detectorofan induction motor
utilizing rotor slots h a m m i c s . IEEE Trans. On I n d m y Appiicotiom, Ma).
June 1984. pp. 575-582.
[SI K.D. Hum. T.G. Hahetler: A comparison of spectrum estimatioii
techniques far sensorless speed detection in induction machines. IEEE FOLT
On l n d n r i ~Applicarionr.
~: vol. 33. n. 4. 1997.
[6] R Beguenane. M.E.H. Benbouzid. F.A. Capolina: On-lineidmdtication
of induction motor rotor parameters from terminal signals. ElCro-mlMan.
v01.3. no.2. April-June 1996. pp. 51-57.
[7] R . Pintelon. J . Schoukens: An improved sine-wavefimngpmdwehr
characterizing data acquisuiitian channels. IEEE Tram. @EInsmmw"o,t
and Meorsrmrent. April 1996. pp. 588-593.
[8] R. Micheletti. R. Pieri: Non-invasive slip measurement ofinduction
motors. IEEE It?strtrmentotionand Meosweatmi 7 e P d n u l u ~Conference
W T C 2 0 h l . Budapest.May21-23. 2001. vol. I , pp. 1988-61
[9] R. Micheletti. R. Pieri: Accurate measuring method for the slip of
asynchronous motors, 12Ih IMEKO 7C4 lniernnriosol S w q " n n i "

Electrical Meosirrenienr ond lnsrrirnienroriorr ,Zagreb. Croatia. September

25-27. 2002.

72

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on October 6, 2008 at 7:39 from IEEE Xplore. Restrictions apply.