Non-Linear Modeling and Identification of A Permanent Magnet DC Motor
Non-Linear Modeling and Identification of A Permanent Magnet DC Motor
Non-Linear Modeling and Identification of A Permanent Magnet DC Motor
C. A. Pérez-Gómez J. U. Liceaga-Castro
MCIE Departamento de Electrónica
Universidad Autónoma Metropolitana, Azcapotzalco Universidad Autónoma Metropolitana, Azcapotzalco
CDMX, México CDMX, México
al2183802746@azc.uam.mx, cart.gom@gmail.com julc@correo.azc.uam.mx
I. I. Siller-Alcalá
Departamento de Electrónica
Universidad Autónoma Metropolitana, Azcapotzalco
CDMX, México
sai@azc.uam.mx
Abstract—In this work, the modeling and identification of speed. However, for position control purposes, where the
the non-linear model of a permanent magnet DC motor is motor operates around zero speed, these nonlinearities induce
obtained and validated using real-time data. The model includes oscillations in the motor response, avoiding the possibility
the non-linearities of the so call dead zone and Coulomb plus
viscous friction that significantly affect electric motors when of achieving zero steady state error. Moreover, including
used as position actuators. The dead zone is modeled using these non-linearities becomes a must when high precision is
the approach called “hard dead zone” which is considered a required specially under variable loads, [8]–[10]. Obtaining
better approximation to the real phenomenon. The resulting non- good models for electromechanical systems, including DC
linear model, simulated in Matlab®/Simulink™, is compared and motors, involves modeling non-linearities such as dead zone
validated to the Maxon DC motor responses of the Quanser DC
motor control trainer system. and friction. However, not only the modeling but the iden-
Index Terms—Nonlinear Model, DCMCT System, Dead zone, tification of these non-linearities is not trivial, so building
Coulomb Friction, DC motor. an accurate model is a difficult task, [11], [12]. In recent
years, new approaches have been developed to obtain more
I. I NTRODUCTION precise models better representing the behavior of systems
The most common type of actuator in electromechanical such as electric motors. Also, by having better and more
systems is the direct current (DC) motor. The permanent accurate models, the validation and testing of different control
magnet DC motor is the simplest type of these devices. They approaches through digital simulations allows reducing real
have been extensively used due to their small physical size, time implementation problems in the actual process, [13].
relatively low cost, and high efficiency in a wide variety of
applications such as: windshield wipers, personal computer In this paper a validate PMDC motor non-linear model in a
fans or as electromagnetic actuators in biomedical equipment, low speed range is obtained. The model includes the dead zone
robotic manipulators and many other industrial applications, and friction, whose parameters are identified by experimental
[1]–[3]. tests. The selected PMDC motor is the Maxon motor of the
The simplicity of its construction makes the permanent Quanser DC Motor Control Trainer (DCMCT) system. The
magnet DC (PMDC) motor relatively easy to understand and dead zone is modeled by the “hard dead zone” model and
control. These motors do not have field winding; therefore, the friction model used is the Coulomb plus viscous friction
the control is through the armature winding. The general model. Therefore, the main objective of this paper is to provide
control methods for these devices are based on the classic a validated nonlinear model for a PMDC which includes the
PID controllers, modern and intelligent controllers. [4], [5] most important nonlinearities for position control purposes
Usually, this type of motors is modeled as second order with a methodology for its identification.
linear system neglecting or disregarding the non-linearities
present in the system like the dead zone and Coulomb friction, The paper is structure as follows. Section II presents the
[6], [7]. Neglecting or ignoring these non-linearities can affect model for the PMDC motor and the non-linearities. Section
the performance of control systems, especially in those cases III explains the system’s parameters estimation. Section IV
where the control system is designed to operate at low speeds describes the design model in Matlab®/Simulink™ including
or for position control objectives. That is, in the case of the dead zone and Coulomb plus viscous friction. The model
speed control systems this simplification does not affects its is validated in Section V and, finally in Section VI conclusions
performance as these non-linearities act at very low or zero are presented.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 09,2022 at 16:38:13 UTC from IEEE Xplore. Restrictions apply.
II. M ATHEMATICAL M ODELING v(t)
mr
A. PMDC motor model
The PMDC motor can be modeled using two linear equa-
tions for the electrical and mechanical subsystems. bl
br u(t)
di(t)
v(t) = Ri(t) + L + Ea (1)
dt
d2 θ(t) dθ(t) ml
Tm = km i(t) = J +b (2)
dt2 dt
Equation (1) represents the electrical subsystem where v(t) Fig. 1. “Smooth” dead zone
is the applied armature voltage, i(t) is the armature current,
Ea the counter electromotive force and, L and R represent the
B. Dead Zone Model
inductance and resistance of armature winding, respectively.
The expression (2) represents the mechanical subsystem Commonly, the non-linear dead zone is represented by the
where Tm is the magnetic torque, km is the motor constant, “smooth” dead zone shown in the figure 1,
J is the motors rotor equivalent moment of inertia, b is the where u(t) is the input, v(t) is the output, bl and br are
friction coefficient and θ is the angular position of the rotor. the left and right break points while ml and mr are the
In this model the counter electromotive force present in the slopes of the dead zone, respectively. When the dead zone
motor is assumed negligible is symmetric bl = br and ml = mr . [15], [16] Nevertheless,
this approximation does not necessarily accurately represent
The transfer function Ge (s) of the electrical subsystem
the real physical phenomenon in the case of electric motors.
neglecting the counter electromotive force, since it is very
For this reason, the dead zone is modeled as shown in
small at low speed, results in:
figure 2, [17] this representation is called the “hard dead zone”
I(s) 1
Ke which is considered a better approximation of the non-linear
Ge (s) = = L R = (3) phenomenon that occurred in the PMDC motor.
V (s) R s + 1 (τ e s + 1)
190
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 09,2022 at 16:38:13 UTC from IEEE Xplore. Restrictions apply.
is a phenomenon occurring between the electrical subsystem R L
and the mechanical subsystem. +
i(t)
TABLE I
M AXON M OTOR DATA
0
0 0.5 1 1.5 2
t, [s]
Parameter Value
R 10.6 Ω
Fig. 5. Current response in the blocked rotor test
L 0.825 × 10−3 H
km 50.2 × 10−3 Nm/A
J 12.1 × 10−7 kg m2
The steady state gain obtained for these responses is Ke =
0.421762 with a steady state time ts = 0.03 s. Thus, the time
191
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 09,2022 at 16:38:13 UTC from IEEE Xplore. Restrictions apply.
constant is τe = 0.0075. The resulting transfer function for km
b= = 0.0388 N m (13)
the electrical subsystem is shown in (9). Kmot R
Ke 0.421762 J = τmot b = 3.104 42 × 10−3 kg m2 (14)
Ge (s) = = (9)
τe s + 1 0.0075s + 1 It is well known that viscous friction coefficient b depends
Therefore, by equation (3) resistance R and inductance L on speed, however, it will be assumed constant since the
result in: objective is to obtain and validate a motor model for a narrow
1 speed range around zero speed.
R= = 2.3724 Ω (10)
Ke C. Non-linear parameters
L = τe R = 17.7933 × 10−3 H (11) Coulomb friction is estimated by experimental observation.
B. Speed measurement at a step input That is, the Coulomb friction value is manually adjusted based
on the model’s speed and position responses to a triangular
The mechanical parameters are estimated following the input voltage signal and compared to actual motor responses.
same procedure as that applied to the electrical subsystem. Similarly, the viscous friction value is adjusted so that the
That is, is possible to identify equation (6) by the analysis of model responses are as similar as possible to the actual motor
the transient and steady state responses of the rotor velocity responses. The triangular input signal allows better observation
ω(t) to different step input voltages. Rotor velocity ω(t) of this non-linear phenomena when the system operates at
responses are shown in figure 6. different speeds around zero speed.
The values for the Coulomb coefficient and the viscous
1.8 coefficient obtained are bc = 0.005 and bv = 0.0314,
Vin = 0.5
Vin =1 respectively.
Vin = 1.5 The dead zone parameters are estimated by measuring
1.6
Vin =2
Vin = 2.5 the input voltage and the current generated in the armature
at which the motor starts to move, so the magnetic torque
1.4
breaking point of the dead zone can be estimated, equation
(4).
1.2 The motor starts to move at ±0.3 V with an armature current
i(t) = 0.1265 A. Therefore, the torque break points are
Tmδr = ±6.35 × 10−3 .
1
Finally, the estimated parameters of Maxon PMDC motor
ω, [V]
Parameter Value
0.4 R 2.3724 Ω
L 17.7933 × 10−3 H
J 3.104 42 × 10−3 kg m2
0.2 bv 0.0314 N m
bc 0.005 N m
δr 6.35 × 10−3
0
0 10 20 30
t, [s]
IV. M ODEL IN M ATLAB®/S IMULINK™
Fig. 6. Speed response at a step input The Matlab®/Simulink™ model of the PMDC motor is
shown in figure 7, where the dead zone is between the
From figure 6, steady state gain and steady state time result electrical and mechanical subsystems.
in Kmot = 0.545287 and ts = 0.32 s, respectively. Thus, The symmetric dead zone is modeled by
process time constant is τmot = 0.08 and the resulting motor Matlab®/Simulink™ blocks as shown in figure 8. The
transfer function is given by equation (12). km i(t)in represents the magnetic torque generated by the
electric subsystem and the km i(t)out is the magnetic torque
0.545287 supplied to the mechanical subsystem. For values less than
G(s) = (12)
0.08s + 1 Tmδr = ±6.35 × 10−3 no torque is induced to the mechanical
From equation (6) the viscous friction coefficient b and the subsystem.
inertia J can be calculated, resulting in:
192
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 09,2022 at 16:38:13 UTC from IEEE Xplore. Restrictions apply.
E LECTRICAL S UBSYSTEM M ECHANICAL S UBSYSTEM
2 Real motor
+ i(t) + θ̇ Non-linear model
v(t) 1/L km 1/J θ
− −
D EAD Z ONE
1.5
R 0.2
F RICTION 0
1
−0.2
0.5
ω, [V]
0
−0.5
−1
−1.5
90 95 100
V. VALIDATION
−5
The non-linear PMDC motor model and the real system are
compared using the Matlab®/Simulink™real time package.
For the tests a triangular input is supply for both, the model
and the real system, in the voltage range [−2, 2]V. Through −10
193
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 09,2022 at 16:38:13 UTC from IEEE Xplore. Restrictions apply.
The rotor position responses is shown in figure 11. This [10] J. d. J. Rubio, Z. Zamudio, J. Pacheco, and D. Mújica Vargas, “Propor-
tional derivative control with inverse dead-zone for pendulum systems,”
response is obtained by integrating the tachometer signal Mathematical Problems in Engineering, vol. 2013, 2013.
output and comparing it with the response of the non-linear [11] C.-K. Lai, “A hybrid scheme motion controller by sliding mode and two-
model. The dead zone can be clearly seen at the top and bottom degree-of-freedom controls to minimize the chattering,” Mathematical
Problems in Engineering, vol. 2014, 2014.
of the curve. It is possible to notice that the error between both [12] C.-C. Chiang, “Adaptive fuzzy tracking control for uncertain nonlin-
responses is very small for the tested voltage range. ear time-delay systems with unknown dead-zone input,” Mathematical
Problems in Engineering, vol. 2013, 2013.
VI. C ONCLUSIONS [13] J. D. Fortgang, L. E. George, and W. J. Book, “Practical implementation
of a dead zone inverse on a hydraulic wrist,” in ASME 2002 international
In this work, a non-linear PMDC motor model is proposed. mechanical engineering congress and exposition. American Society of
Mechanical Engineers Digital Collection, 2002, pp. 149–155.
This model takes into account two of the most important non- [14] F. MONASTERIO and A. Gutiérrez, “Modelo lineal de un motor de
linear phenomena, dead zone and friction, that significantly corriente continua,” 2012.
affect the performance of designed controllers, especially in [15] D. Recker, P. Kokotovic, D. Rhode, and J. Winkelman, “Adaptive non-
linear control of systems containing a deadzone,” in [1991] Proceedings
those systems that operate at low speeds or for those in whose of the 30th IEEE Conference on Decision and Control. IEEE, 1991,
position control is essential. pp. 2111–2115.
The dead zone is modeled using an approach called the [16] C. Hu, B. Yao, and Q. Wang, “Adaptive robust precision motion
control of systems with unknown input dead-zones: A case study with
“hard dead zone”, where the inertia breakdown is represented comparative experiments,” IEEE Transactions on Industrial Electronics,
as a sudden displacement, better describing the phenomenon vol. 58, no. 6, pp. 2454–2464, 2010.
present in the real system. The friction phenomenon is repre- [17] T. Wescott, “Controlling motors in the presence of friction and back-
lash,” in Wescott Design Services, Seminars, vol. 42, 2013, pp. 2–3.
sented using the Coulomb plus viscous friction model which [18] H. Olsson, K. J. Åström, C. C. De Wit, M. Gäfvert, and P. Lischinsky,
better represent the friction a low and zero speed. “Friction models and friction compensation,” Eur. J. Control, vol. 4,
To obtain the model, the motor parameters are estimated no. 3, pp. 176–195, 1998.
[19] V. Van Geffen, “A study of friction models and friction compensation,”
using the transfer functions corresponding to the electrical and 2009.
mechanical subsystem by applying the Strecj method. [20] B. Armstrong-Hélouvry, P. Dupont, and C. C. De Wit, “A survey of
models, analysis tools and compensation methods for the control of
The model was validated with the Quanser DCMCT system machines with friction,” Automatica, vol. 30, no. 7, pp. 1083–1138,
that has an integrated Maxon PMDC motor. The non-linear 1994.
PMDC motor model was simulated and compared with the real [21] Y. Liu, J. Li, Z. Zhang, X. Hu, and W. Zhang, “Experimental comparison
of five friction models on the same test-bed of the micro stick-slip
system for a certain range of voltages where it was possible motion system,” Mechanical Sciences, vol. 6, no. 1, p. 15, 2015.
to clearly observe the non-linear phenomena present in the [22] S. Skoczowski and A. Osadowski, “A simple identification method for
analyzed motor, observing through the responses of both the the order of the strejc model and its application to autotuning,” in
Intelligent Components and Instruments for Control Applications 1994.
linear model and the motor Maxon DC permanent magnet, Elsevier, 1994, pp. 319–325.
that the model obtained is a very good approximation of the
real system at low speeds. Finally, it is also concluded that a
viscous friction model that depends on speed is necessary if
a model for wider speed ranges is required.
R EFERENCES
[1] N. P. Mahajan and S. Deshpande, “Study of nonlinear behavior of
dc motor using modeling and simulation,” International Journal of
Scientific and Research Publications, vol. 3, no. 3, pp. 576–580, 2013.
[2] S. Syukriyadin, S. Syahrizal, G. Mansur, and H. Ramadhan, “Permanent
magnet dc motor control by using arduino and motor drive module
bts7960,” in IOP Conference Series: Materials Science and Engineering,
vol. 352, no. 1. IOP Publishing, 2018, p. 012023.
[3] J. E. Carryer, R. M. Ohline, and T. W. Kenny, Introduction to mecha-
tronic design, 2011.
[4] A. Hughes and B. Drury, Electric motors and drives: fundamentals,
types and applications. Newnes, 2019.
[5] S. Moussavi, M. Alasvandi, and S. Javadi, “Speed control of permanent
magnet dc motor by using combination of adaptive controller and fuzzy
controller,” International Journal of Computer Applications, vol. 52,
no. 20, 2012.
[6] A. Alkamachi, “Permanent magnet dc motor (pmdc) model identification
and controller design,” Journal of Electrical Engineering, vol. 70, no. 4,
pp. 303–309, 2019.
[7] H. Chu, W. Tao, B. Gao, Q. Liu, and H. Chen, “Speed control of the
permanent-magnet dc motor subjected to uncertainty and disturbance,”
in 2016 35th Chinese Control Conference (CCC). IEEE, 2016, pp.
4664–4669.
[8] S. Brennan and A. Alleyne, “Dead-zone non-linearities,” 1999.
[9] J.-H. Horng, “Neural adaptive tracking control of a dc motor,” Informa-
tion sciences, vol. 118, no. 1-4, pp. 1–13, 1999.
194
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on February 09,2022 at 16:38:13 UTC from IEEE Xplore. Restrictions apply.