Parameter Identification of DC Motors

Author:
Dipl.-Ing. Ingo Vllmecke

Advantages of the Parameter Identification Method

Saving time and money in the testing process: no mechanical coupling necessary
Full information: Entire characteristic curve and values of any motor

www.imc-berlin.com Berlin: +49 (0) 30-46 70 90-0 Hotline: +49 (0) 30 46 70 90 26 Frankfurt: +49 (0) 61 72-59 05 0 1
 DC-motors are used in a wide variety of applications touching our R L KR

daily lives, where they serve to relieve us of much work. However, i(t) (t)
the large number of DC-motors used is attended by a large amount
U(t) Ug() Mi(iA) ML
of time and resources devoted to inspecting them at the end of
their production cycle. The time needed for this process should be
kept as brief as possible so that the inspection procedure is not the
slowest part of the production process. Due to the increasing mass Fig. 1: Electromechanic equivalent circuit
production of these motors, procedures for inspecting them have
been developed which are able to deter-mine the test objects 1.1 Energy model
characteristic curves within seconds. Such procedures are known To better interpret Equations (1) and (2), we now proceed to model
as Parameter Identification procedures. They determine the the energy relationships.
parameters without applying any external load, simply by
measuring the current and voltage. The time and apparatus Thus, Equation (1) is multiplied with the current i and integrated over
required for attaching a load and for aligning the test object with a an indefinite interval.
loading mechanism can thus be totally omitted. di
u idt = R i dt + k idt + L i dt dt
2
(3)
1 Introduction

The first term in Equation (3) describes the energy transferred

The dynamic behaviour of d.c. motor can be described using two
electrically, the second the Ohmic losses, the third the mechanical
equations.
energy contained in the system and the last the energy stored as
inductivity.
The first equation describes the electrical behavior
di
u = R i + k + L (1) Equation (2) is multiplied by the rotational frequency and
dt integrated over an indefinite interval.
d
The second equation describes the mechanical behavior J dt = k i dt k r 2 dt M L wdt (4)
dt
d
J = k i kr M L (2) The first term in Equation (4) describes the rotation energy stored in
dt the mechanical system, the second term the mechanical energy, the
where the algebraic symbols represent the following: third the speed-proprotional energy losses and the last the
mechanical energy delivered including wasted energy due to static
Symbol Unit Definition friction.
u V electric terminal voltage
Einput
i A electric armature current electrical
part
1/s rotational freqency
R Ohmic ferrule resistor Eelectrical loss

k Vs generator constant Emagnetic

Etransfer
L H inductivity mechanical
part
J kgm moment of inertia
Eangular kinetic
kr Nms sliding friction
ML Nm load Efrictional loss

Note: The load also reflects the moment of static friction inherent in the Eoutput
system. Equations (1) and (2) can now be summarized in a single
equivalent electromechanic circuit. Fig. 2 : Diagram of energy distribution

www.imc-berlin.com Berlin: +49 (0) 30-46 70 90-0 Hotline: +49 (0) 30 46 70 90 26 Frankfurt: +49 (0) 61 72-59 05 0 2
 1.2 Transfer function of a DC motor From Equation (8), it follows that:
Since only the terminal quantities voltage and armature current
are used in identifying a DC motor's parameters, the course of the k
I (s) 1 s+ r
spectrum of the armature current-to-terminal voltage ratio is of J (9)
=
particular interest. On the basis of this transfer function it is kML 1 L 2 R kr R kr k 2
U (s) + s + s + + +
J kr L J L J RJ
possible to make statements about at which point in the s + s
J
frequency range excitation is useful, since relevant parameter
changes take effect in that frequency range.
We now introduce the following identities:
A simple example should serve to affirm this: Symbol Unit Definition
ele s electrical time constant
To identify the parameters of an electric low-pass with a cutoff mech s mechanical time constant
frequency of 10 kHz, it isn't a practical approach to excite it with kA 1/s run-down constant
a 10 Hz oscillating quantity, since, considering the measurement V A/Vs gain
precision, the input and output signals are approximately the U#(s) V kM L 1
same (transfer factor approx. 1, phase-shift between input and U ( s) +
J k
s + r s
output signals approx. 0 degrees). Only if the excitation J
approaches the cutoff frequency do the filter parameters become
and inserting these abbreviations results in the following transfer
noticeable and is there a chance of determining them with
equation
reasonably high accuracy.

I ( s) s + kA (10)
To determine the transfer function, Equations (1) and (2) are =V
U # ( s) 1 1 1
transferred into the Laplace domain. s 2 + s + k A + kA +
ele ele mech
voltage current rotation speed

U ( s) = R I ( s) + k ( s) + L s I ( s) (5) V
12
A
50
10^3 RPM
4.8
45 4.6

J s ( s) = k I ( s) k r ( s) M L
11 4.4
(6) 10
40
35
4.2
4.0
30 3.8

From Equation (5) we thus find the rotational frequency 9 25

20
3.6
3.4
8 3.2
15
3.0
10
7 2.8

k I ( s) M L
5 2.6
6 0 2.4

( s) = (7) -5 ele 2.2

kr + s J
5 2.0
-10
1.8
-15
4 1.6
-20 1.4
3 -25 1.2
-30 1.0
2 0.8
-35
By using (7) in (5) and making some rearrangements, we obtain 1
-40
0.6
0.4
-45 mech 0.2

the equation 0 -50

1.27 1.28 1.29 1.30 1.31 1.32 1.33 1.34
0.0

Fig. 3 : Electrical and mechanical time constants of the electrical DC-

LJs2 + s(RJ + Lkr) + RKr + k 2 kM 1 (8)
U(s) = I (s) L motor
kr + sJ J kr
s + s
J
The constant ele is referred to as the electrical time constant of

Equation (8) describes the relationship between the terminal the DC motor. The electrical time constant is a measure of the
quantities U and I and to the load ML. current's reaction time upon change in the terminal voltage.

www.imc-berlin.com Berlin: +49 (0) 30-46 70 90-0 Hotline: +49 (0) 30 46 70 90 26 Frankfurt: +49 (0) 61 72-59 05 0 3
 The constant mech is referred to as the mechanical time constant Rotation Speed Current output power efficiency

of the DC motor. The mechanical time constant is a measure of RPM

3600
A
65
W
190
%
100

the RPM's reaction time upon change in the terminal voltage. 3400
60
180 95

3200 170 90

Using Equation (9), the DC-motor's frequency response can be 3000 55 160 85

150 80
2800
shown for fixed parameters. 2600
50
140 75

45 70
130
2400
65
120
2200 40
60
The figures below show the frequency responses in terms of both 2000
110
55
35 100
magnitude and phase, as well as the characteristic curve for a 1800
30 90
50
45
1600
80
motor with given parameters. 1400 25
70
40
35
1200
20 60
30
1000
50 25
15
Parameter Unit Value 800 40 20
600 10 30 15
R 0.19 400
5
20 10
200 10 5
L H 0.0005 0 0 0 0
0.0 0.5 1.0 1.5 2.0 2.5
k Vs 0.0323 Nm

J kgm 7.5e-5 Fig. 6: Characteristic curve of a DC motor

kr Nms 2e-5
In addition to the curves of the respective frequency responses, the
magnitude poles magnitude tau magnitude
frequencies which are proportional to the mechanical and electric
6.0

5.5
time constants, as well as the frequencies of the transfer function's
5.0 poles are plotted.
4.5

4.0
The magnitude frequency response of the DC motor corresponds
3.5

3.0
to a band-pass filter with a center frequency lying between the
2.5 motor's two time constants.
2.0
The phase frequency response of the DC motor corresponds to the
1.5

1.0
phase frequency response of a band-pass. The zero in the transfer
0.5 function's numerator causes the phase to approach zero at low
0.0 frequencies.
110-2 110-1 1100 1101 1102 1103 1104 1105 1106
Hz

Fig. 4: Magnitude frequency response of a DC motor The DC motor in the example described has two real poles in the
phase poles phase tau phase
transfer function. These poles appear in the frequency response as
Grad
100
90 simple but separate poles. Conversely, there are also motors with
80
70 conjugated complex poles which can be represented in the frequency
60
50 response as double-poles at the band-pass center frequency.
40
30
20
10 Taking a look at the transfer function, it is now possible to determine
0
-10
-20
certain regions in which the DC motor's parameters can be identified.
-30
-40
As indicated above, this results in a dependence on the numerator
-50
-60
polynomial's zero-crossing in the lower range of the transfer function;
-70
-80 the zero-crossing in turn depends on the run-down constant. In
-90
-100 addition, in this range the magnitude frequency is almost zero,
110-2 110-1 1100 1101 1102 1103 1104 1105 1106
Hz making estimation basically impossible and the run-down constant
Fig. 5 : Phase frequency response of a DC motor not to be determined with adequate precision.

www.imc-berlin.com Berlin: +49 (0) 30-46 70 90-0 Hotline: +49 (0) 30 46 70 90 26 Frankfurt: +49 (0) 61 72-59 05 0 4
 In the region of the DC-motor's time constants, sufficiently large
k2 L
amplitudes result for the purpose of estimating the system's = Rk A (17)
J T1T2
poles. Then, the DC-motor's parameters can be calculated from
the poles. If the numerator polynomial's poles have been determined and the
motor's run-down constant is additionally known, these can be
From Equation (10), we see that for the zero-crossings: used to find the motor's parameters. The last actual purpose of the
estimate is to determine the numerator polynomial's poles.
2
1 1
+ kA + kA Upon closer inspection of the transfer function and of the frequency
ele 1 k + 1 (11)
s01,2 = ele response, it is seen that the system can be divided into a high-pass
2 2 ele A mech and a low-pass filter.


kr kr
The system's poles become purely real if the square root I ( s) s+ s + 1
=V J =V J (18)
expression is positive, from which follows: #
U ( s) (1 + sT1 )(1 + sT2 ) 1 + sT2 1 + sT1
ele
mech > 4 (12)
(1 k A ele )
2
The first term in Equation (18) represents a high-pass filter, the
second a low-pass.
All DC motors meeting the conditions in Equation (12) have purely
real poles.

1.3 Estimating the transfer function's poles

For all motors for which Equation (12) applies, the following
approach to determining the parameters is available: Fig. 7: Representation of the DC-motor's individual frequency
responses
I ( s) s + kA
=V
(1 sT1 )(1 sT2 )
(13)
#
U ( s) If the motor is now driven at frequencies f1 and f2 in succession, the
factors in Equation (18) can be separated and calculated individually.
As Figure 3 clearly shows, when driven at 1, the low-pass returns a
this equation can be multiplied out to yield
magnitude factor of approx. 1 and a phase shift of zero degrees, so
that the low-pass can be neglected at first approximation. Thus we
I ( s) V s + kA
= (14) obtain the following transfer function:
#
U ( s) T1T2 T +T 1
s2 + s 1 2 + kr
T1T2 T1T2
I (1 ) j1 +
V1 J (19)
U (1 )
#
1 + j1T2
By comparing coefficients with Equation (9), we obtain the
By multiplying out Equation (19) and splitting it into its real and
following conditional equations for the DC-motor's parameters.
imaginary parts, we find the system of equations (where the indices
TT
L= 1 1 (15) I and R denote the real and imaginary component of the respective
V quantity, e.g. UR= Re{ U#(j) }.

T + T2 I R 1T2 I I = V1U R *
R= 1 kA L (16)
T1T2 * (20)
I I + w1 I R T2 = V1U I

www.imc-berlin.com Berlin: +49 (0) 30-46 70 90-0 Hotline: +49 (0) 30 46 70 90 26 Frankfurt: +49 (0) 61 72-59 05 0 5
 The following identities are used using the following identities
* kr
U R = U R J 1U I kr kr
k (21) U R J U I 2 + 2 T2 (2U R + U I )
J
U I * = U I r + w1U R U R =
J 1 + 2 2 T2 2 (25)
kr kr
U I J + U R2 + 2 T2 (2U I U R )
J
and the equations for determining T2 and V1 result.
U I = 1 + 2 2 T2 2

I RU I * I I U R *
2
T = Using Equations (22) and (24), the motor parameters can now be
w1 ( I RU R * + I I U I * )
(22) determined by iteration. To do this, the results from (22) are applied
I R2 + I I 2
V1 = U * I + U * I to (24) until the iteration algorithm converges. Then the test object's
R R I I parameters can be found using Equations (15) through (17).

2 Identifying a sample motor's parameters

Thus we obtain the first value estimate for determining a pole of
the transfer function T2 and of the gain factor V1. For the example of a real motor described above, the following
transfer function results:
These value estimates can now be used in estimating the second
parameter. For this purpose, Equation (18) is rearranged as 2105
s+ (26)
I(s)
=
1 . 105
75
follows: 5 5
019 210 019 . 210
# 2
U (s) 00005
. . 00323
.
s2 + s + + +
00005
. . 105 00005
75 . . 105 019
75 . 105
. 75
kr
jw2 +
I ( w2 )(1 + j2 T1 ) = V2U ( w2 ) J (23)
1 + jw2 T2 which solves as

I ( s) s + 0.2666
= 2000 2 (27)
We now solve for T1 and V2, which, after separating the real and #
U ( s) s + 380.2666s + 27922.4
imaginary components in Equation (24), are determined by the
conditional equations:

and implies the following values for characteristic quantities:

ele = 2.63ms
I RU I I I U R mech = 13.66ms
1
T =
w2 ( I RU R + I I U I ) kr
IR2 + II 2
(24) = 0.2666
V = J
2 U I +U I
R R I I

www.imc-berlin.com Berlin: +49 (0) 30-46 70 90-0 Hotline: +49 (0) 30 46 70 90 26 Frankfurt: +49 (0) 61 72-59 05 0 6
 Typical plots of current and voltage with the inverse values of the The iteration ultimately returns the following parameters:
electrical and mechanical time constants as the excitation
frequencies are displayed in Figure 8. T1=0.00356063
T2=0.0100582
Voltage Current V1=0.0716225
V A
12 50 V2=0.0716479
45
11
40

10
35 This translates to the following parameters for the test object:
30
9 25

8
20
R= 0.1904
15

7
10 L= 0.000501
5
6 0
k= 0.03233
-5
5
-10
-15
4
-20
3 -25
Thus the accuracy of the parameters determined is within 0.2%.
-30
2
-35
-40
1
-45
0 -50
0.8 0.9 1.0 1.1 1.2 1.3

Fig. 8: Plot of current and voltage for determining the parameters

of a sample motor
V2 V1
10^-3
105.00

96.25

87.50

78.75

70.00
T1
10^-3
3.600

3.425

3.250

3.075

2.900
T2
10^-3
17.00

15.25

13.50

11.75

10.00
0 1 2 3 4 5 6 7 8 9 10

Fig. 9: Plot of identified parameters T1, T2, V1, V2

Figure 9 shows a plot of the parameters T1, T2, V1 and V2. The
convergence produced by this procedure is clear to see.

www.imc-berlin.com Berlin: +49 (0) 30-46 70 90-0 Hotline: +49 (0) 30 46 70 90 26 Frankfurt: +49 (0) 61 72-59 05 0 7