2017 International Conference on Control, Artificial Intelligence, Robotics & Optimization

Series DC Motor Modeling and Identification

Jesús U. Liceaga-Castro, Irma I. Siller-Alcalá, Jorge Jaimes-Ponce & Roberto Alcántara-Ramírez
Departamento de Electrónica
UAM-Azc
México City, México
julc@correo.azc.uam.mx

Abstract—The modeling and identification of a series DC the case of linear first and second order systems this procedure
motor without saturation is presented. The identification is renders good models by estimating its poles and steady state
performed using the Strejc method. The procedure requires only gain. In some cases, it is also possible to estimate the physical
the measurement of the current consumption and the motor parameters of the process.
rotors speed. The resulting model is validated comparing the
responses of the estimated model against real time responses of Therefore, the main objective of the paper is to present the
an actual motor. modeling, without magnetic saturation, and identification of
series DC motor. A second objective is to present a simple
Keywords—Electric Machine; Identification; Series dc motor; procedure to estimate the parameters of the model using only
Linear systems the measurement of the current consumption and the speed of
the motor rotor. This methodology is based on the transient and
I. INTRODUCTION steady state responses of the mechanical and electrical
subsystems of the motor known as the Strecj algorithm. In
Series DC motors, as well as series universal motors, are a order to assess the estimated model comparisons between the
kind electric motors with one voltage supply and the field time responses of the model and an actual motor are presented
winding connected in series with the rotor winding. This series
connection results in a motor with very high starting torque.
However, torque decreases as the speed builds up due to an II. SERIES DC MOTOR MODEL
increment of the back or counter electromotive force EMF. Series DC motors similar to shunt wound DC motors or
This is why series DC motors have poor speed regulation. compound wound DC motors are self-excited DC motors.
That is, increasing the motors load tends to slow its speed They get their name because the field winding is connected
which in turns reduces the back EMF and increases the torque internally in series to the armature winding as shown in figure
to accommodate the load. A limitation of these motors is that 1. They are also considered self-excited motors because instead
the sense of rotation is fixed for most of their applications. In of two separate voltage sources -one for the armature and one
order to change the direction of torque and rotation, it is
for the field winding- they required only one voltage source.
necessary to change the polarity of the current flow.
The electric diagram of a series DC motor is shown in figure
Despite the fact series DC motors generate high torques 2. Based on the electric diagram of figure 2 the differential
with very low current consumption and small dimensions they equations comprising the mechanical and electrical subsystems
are commonly used open loop for short periods of time. This of the series DC motor are given by:
is mainly, as mention above, because they have poor speed
regulation. Nonetheless, this kind of motors can be fully d d
exploited if good closed loop controllers are designed. V (t ) Ra ia (t )  R f i f (t )  La ia (t )  L f i (t )  Ea 
However, in general many control strategies to this kind of dt dt f
motors are based or depend on dynamic cancellations [1]-[5] d
Te (t ) TL (t )  bZ (t )  J Z (t ) 
requiring good models. dt
Models, in particular mathematical models, are a corner
stone for many control strategies. These are normally obtained
by analyzing the physical properties of the phenomena.
However, a second crucial problem is to estimate the values of
the different physical parameters involved in the model. In the
case of linear system and some nonlinear models the algorithm
of Least Square is one of the most popular methodologies to
estimate or identify these parameters. Nonetheless, a necessary
condition to apply this algorithm is that the model must be
linear at its parameters, [6]. Another identification approach is
the Strecj algorithm, successfully applied in [7] for the
identification of the asynchronous machine which is based on Fig. 1. Series connection of a DC motor
the transient response analysis to a step input of the process. In

978-1-5090-6536-3/17 $31.00 © 2017 IEEE 249

248
DOI 10.1109/ICCAIRO.2017.54
 III. MODEL IDENTIFICATION
The identification of the monophasic universal-motor
Koblenz model HC8825M110 with nominal maximum speed
and power of 24000 RPM and 0.815HP is presented.
In particular, these machines must operate loaded in order
to avoid damage. Hence, a steel disc load was added as shown
in figure 3. This extra load is considered as a part of the rotors
inertia. From equations (8) and (9) it is clear that it is possible
to estimate the motors parameters by analyzing the responses
of the current and angular velocity of the rotor to step input
Fig. 2. Electric diagram of a series DC motor voltages. This is performed using the set up experiment
depicted in figure 4.
As ia (t ) i f (t ) equations (1) and (2) reduce to: The experimental setup consists of a voltage source with a
maximum voltage of 50 volts and a maximum current of 3
V (t ) R a  R f  i (t )   La  L f  dtd i(t )  E
a  Amp, the series DC motor, and a power driver with a USB
communication. The power driver was designed considering
d the electric demand of the motor. Hence, the power driver is
Te (t ) TL (t )  bZ (t )  J Z (t )  based on the MOSFET 12N65, which operates with a
dt maximum current of 12Amp and a maximum input voltage of
650 Volts. An equally important element is the ACS711LC
Where, Z (t ) is the rotors speed, Ea represents the counter circuit used to monitor the power consumed by the motor.
electromotive force EMF, TL (t ) is the load torque,
i (t ) ia (t ) i f (t ) is the current, b is the friction coefficient,
J is the rotor’s inertia and Te (t ) is the electromagnetic torque
produced by the motor.
The EMF Ea and Te (t ) depend both on the air-gap flux ) ,
that is:

E a ( t ) Z ( t ) ) (i ) (5)
Te (t ) i ( t ) ) (i ) (6)

The flux )(i) is a function of the current i (t ) so equations

(1)-(4) are non-linear. Also, it is common practice to
approximate the flux )(i) by a linear relation when the
magnetic saturation is neglected, that is: Fig. 3. Series DC motor

) (i ) k0 i (t ) (7)

where k0 is the mutual inductance between the armature and

field coils.
Finally, the nonlinear model of a series DC motor without
saturation is given by:

d
V (t ) Ri (t )  L i (t )  Z (t )i(t ) k0 (8)
dt
d
i 2 (t )k0 TL (t )  bZ (t )  J Z (t ) (9)
dt

Where R Ra  R f and L La  L f Fig. 4. Experimental Set Up

249
250
 Inductive loads together with pulsed excitation signals The mutual inductance k0 can be estimated measuring the
generate reverse currents that may damage switching elements steady state responses of the current i (t ) and the speed of the
such as MOSFETs. Although the MOSFETs used in the
implementation have an internal protection diode, two rotor Z (t ) to a step input voltage V (t ) .
transistors FR307 of rapid recovery were added in order for
additional protection, as shown in figure 5.

Fig. 6. Current step response (rotor shaft locked)

d
It should be noted that under this condition i (t ) | 0 .
Fig. 5. Power driver dt
Rearranging equation (8):
A. Electric Subsytem Identification
From equation (8), it is clear that if the rotor shaft is fully
locked, then Z (t ) 0 t , reducing the electric subsystem to a V (t )  Ri (t )
simple RL circuit with a differential equation given by: k0 (12)
Z (t )i (t )

d
V (t ) Ri (t )  L i (t ) (10) In figures 7 and 8, the current i (t ) and rotor speed, Z (t ) ,
dt
responses to a step input voltage from 0 to 25 volts are shown.
From the steady state responses, the previously estimated
with a transfer function G E ( s ) given by: resistance, R , and equation (12), the mutual inductance value
is: k0 0.17554 N-m/Wb-A .
I (s) 1
GE ( s) (11)
V ( s) Ls  R

The steady state gain and the steady state time of GE ( s )

are K G 1 / R and, t ss 4 L / R respectively. This is in
E E

fact the Strecj method for the identification of first order

systems without delay.
With the rotor shaft locked, the current response i (t ) ,
obtained by feeding a step input voltage V (t ) from 0 to 25
volts is shown in figure 6. The current variation is
approximately 1.2 Amp so the steady state gain
is K G 0.048 , hence R 20.833: . Also, from Figure 6, the Fig. 7. Current step response
E

steady state time of the current response is t ss 0.03 sec .

E
Thus, the inductance is L 156.24 mH .

250
251
 TABLE I. ESTIMATED PARAMETERS

k0 0.17554 N-m/Wb-A

R 20.833:

L 156.24 mH

b 0.000026 N-m/Wb-A

2
J 0.0006206 Kg-m

The Simulink program of the Series DC motor non-linear

model without magnetic saturation is shown in figure 9.
Fig. 8. Rotors speed step response

B. Mechanical Subsystem Identification

From equation (4), the mechanical subsystem without
torque load, TL (t ) 0 , is described by:
d
Te (t ) bZ (t )  J Z (t ) (13)
dt
With a transfer function

Z ( s) 1
GM ( s ) (14)
Te ( s ) Js  b

Similar to the electrical subsystem identification, the

steady state gain and the steady state time for the mechanical
subsystem are K GM 1 / b and t ss 4 J / b , respectively. It
M
should be noted that the electromagnetic input torque
2
Te (t ) k 0 i (t ) can be obtained by measuring the current i (t )
Fig. 9. Simulink program Series DC motor
together with the estimated mutual inductance k0 0.17554 .
C. Model Linearization
Figure 7 shows that the steady state variation of the current
i (t ) is 0.255. Thus, from equations (6) and (7) the variation of A linear approximation of equations (8) and (9) around
any equilibrium point may represent a better model for certain
the electromagnetic torque TE (t ) is 0.01141. On the other control objectives. In this sense, equations (8) and (9) are
hand, from Figure 8, the steady state variation of the speed of rearranged as follows
the rotor is 439.82RPM. Hence, the gain of the mechanical
subsystem is K G 1 / b 38531.73 , resulting in a friction d R k 1
M
i (t )  i (t )  0 Z (t )i (t )  V (t ) (15)
dt L L L
coefficient b 0.000026 N-m/Wb-A .
d b 1 k
Z (t )  Z (t )  TL (t )  0 i 2 (t ) (16)
Estimating the rotor inertia J requires measuring the dt J J J
steady state time t ssM of the velocity Z (t ) to a step input.
However, Figure 6 clearly indicates that the magnetic torque defining
input cannot be assumed as a step input. Nevertheless, the
inertia J was initially estimated using the k0 R
relation J t ssM b / 4 . Further adjustments via trial and error a1 : ; b1 :
J L
were carried out finding that J 0.0006206 Kg-m 2 was b k0 x1 : Z
a2 : ; b2 : and (17)
closer to the actual response of the system. J L x2 : i
The résumé of the estimated parameters of the series DC 1 1
motors is shown in Table I. a3 : ; b3 :
J L

251
252
 The nonlinear state space representation of the series DC The poles of the transfer function (25) are ^3256.2,  0.1`
motor is given by:
so the series DC motor is stable and over damped. It is also
x1 a1 x22  a2 x1  a3TL possible to distinguish the two typical modes of a DC motor
x2 b1 x2  b2 x1 x2  b3V
(18) from the poles:  s  0.1 representing the slow dynamic of the
mechanical subsystem and  s  3256.2  the fast dynamic of
ª x1 º ª a x  a2 x1  a3TL º
2
1 2 the electrical subsystem.
x « x » « » f ( x, u ) (19)
¬ 2¼ ¬ b1 x2  b2 x1 x2  b3V ¼ In [8], was shown that linear models may be suitable for
regulation and tracking control objectives.

 
The equilibrium point x10 , x20 of equation (18) is given by:

x20  b1  b2 x10 
IV. REAL TIME MODEL VALIDATION
a2 x10  a3TL
x20 ; V (20) To validate the nonlinear model the responses of the actual
a1 b3 series DC motor and the identified nonlinear model to different
input voltages were compared.

The linear approximation of (18) around the equilibrium In figures (6) and (7) the step responses of the current and
point (19) is given by: rotors speed to a step input voltage from 0 to 25 volts are
shown. From these figures it is clear that the model matched
x Ax  Bu; y Cx (21) the actual motors response.
where To verify the validity of the model in a range of input
variation and frequency a sinusoidal input voltage was applied.
G f ( x, u ) ª  a2 2a1 x20 º
A « »;
Gx x «¬ b2 x2   b1  b2 x1  »¼
0 0
0 0
1 , x2

G f ( x, u ) ª  a3 0 º
B « 0 b »; (22)
Gu x 0 0
1 , x2 ¬ 3¼

C >1, 0@
with

>TL V@ , y
T
u Z (t ) (23)

If the load torque is assumed zero TL 0

Fig. 10. Rotors speed response to a sinusoidal input

a2 x10 x  b1  b x
0
2
0
2 1 ; ª0º
x20 ; V B «b » (24)
a1 b3 ¬ 3¼

These calculations lead to a model, which can be

considered as a system with one input, voltage V (t ) , subjected
to a torque perturbation.
The transfer function G ( s ) associated to the state space
representation (21) around the equilibrium point (20) with
x10 439.82 RPM is:

Z (s) 1 923.3
G(s) C  sI  A  B (25) Fig. 11. Rotors speed response to a sinusoidal input
V (s) 2
s +626.6s+67.58

252
253
 The input voltage is given by: REFERENCES

[1] Leonardo Amet, Malek Ghanes, Jean-Pierre Barbot, “Sensorless control

V (t ) 25  5sin(0.05t ) (26) of a DC series motor”. IFAC ALCOSP, Caen, France. 2013. <hal-
00923720>
[2] Dongbo Zhao and Ning Zhang, “An Improved Nonlinear Speed
Controller for Series DC Motors”. Proceedings of the 17th World
The frequency of the signal takes into account the slow Congress The International Federation of Automatic Control Seoul,
mode of the mechanical subsystem. The responses of the Korea, July 6-11, 2008
current and rotors speed to the sinusoidal input voltage of [3] Jorge L. Barahona-Avalos, Cornelio H. Silva-López and Jesús Linares-
equation (26) are shown in figures (10) and (11). From these Flores, “Control de velocidad de un motor de CD con conexión en serie
mediante Rechazo Activo de Perturbaciones”. Congreso Nacional de
plots it is clear that the estimated model responses are Control Automático, AMCA, México,2005.
comparable to the actual motor responses.
[4] Samir Metha and John Chiasson, “Nonlinear Control of a Series DC
Motor: Theory and Experiment”. IEEE Transactions on Industrial
V. CONCLUSIONS Electronics, Vol.45, No. 1, 1998.
[5] I. I. Siller-Alcalá, J. U. Liceaga-Castro, A. Ferreyra-Ramírez, R.
Alcántara Ramírez and J. Jaimes-Ponce, “Speed Nonlinear Predictive
The modeling, without magnetic saturation, and estimation Control of a Series DC Motor for bidirectional operation”. Recent
of a series DC motor was obtained. The estimation is based Researches in Mathematical Methods in Electrical Engineering and
only on the transient response of the current and rotor speed Computer Science, p.p 182-187. ISBN: 978-1-61804-051-0, 2011
resulting in a simple and easy to implement procedure. The [6] Ioan Doré Landau and Gianluca Zito. Digital Control Systems, Springer,
2006
resulting model matched the actual response of a real series DC
motor. Also, a linearization of the nonlinear model in an [7] M. Khemliche, S. Latreche, A. Khellaf, “Modelling and Identification of
a Sinchronnous Machine”. First International Symposium on Control
equilibrium point is presented. This linear model which clearly Communications and Signal Processing. March 21-24, Hammamet,
identifies the slow mode of the mechanical subsystem and the Tunisia, 2004
fast mode of the Electrical subsystem may be adequate to [8] I. I. Siller-Alcalá, J. U. Liceaga-Castro, A. Ferreyra-Ramírez, R.
regulation purposes where classical linear controllers can be Alcántara Ramírez and J. Jaimes-Ponce. “Speed Nonlinear Predictive
implemented. Control of a Series DC Motor for bidirectional operation”. Recent
Researches in Mathematical Methods in Electrical Engineering and
Computer Science, p.p 182-187. ISBN: 978-1-61804-051-0.

253
254