See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/262913514

Realtime parameter estimation, calibration and simulation of a DC motor

Article  in  Technics Technologies Education Management · January 2011

CITATIONS READS

0 710

3 authors:

Barış Doğan Ali Buldu

Marmara University Marmara University
10 PUBLICATIONS   10 CITATIONS    27 PUBLICATIONS   76 CITATIONS   

SEE PROFILE SEE PROFILE

Hasan Erdal
Marmara University
32 PUBLICATIONS   82 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Development and Implementation of an Algorithm of Trajectory Planning and Computer Vision Based Obstacle Avoidance for Unmanned Aerial Vehicles View project

intelligence transportation systems View project

All content following this page was uploaded by Hasan Erdal on 21 May 2015.

The user has requested enhancement of the downloaded file.

 technics technologies education management

Realtime parameter estimation,

calibration and simulation of a
DC motor
Hasan Erdal1, Barış Doğan2, Ali Buldu3
1
Electrical and Electronics Engineering Dept., Marmara University, Turkiye,
2
Mechatronics Education Dept., Marmara University, Turkiye,
3
Computer and Control Education Dept., Marmara University, Turkiye.

Abstract establishment, prediction of unknown system

parameters from the experiment data and test of
This paper is about real time estimation and cali- the validity of the established model. In the de-
bration of a brushed wound-pole dc motor param- sign and control optimization of an electric motor,
eters, experimantally. An Electrical Machines Test identiication of the motor model and parameters
Bed (EMTB), a dc motor with a driver and a data accurately provides to obtain a high dynamic per-
acquisition card (DAQ) installed computer is used formance and energy eficiency from the motor.
in the experiment setup. The software application These parameters are generally acquired either
is developed on Matlab-Simulink platform using analytically during the motor design or gener-
Real-Time Windows Target library. Firstly, motor ally from the experimental measurements. Thus,
parameters are estimated by using ODEs. Then es- knowing the system model provides the formuli-
timated parameters are simulated within the model zation of the system characteristics and this for-
and compared with real motor behaviour. Finally, mulization realizes the design and establishment
parameters are calibrated with Matlab-Parameter of high performance control systems. [1-4]
Estimation tool and linear model of the dc motor is There are some studies about the modeling and
acquired according to velocity output. estimation of the parameters of an electric machine
Key words: parameter estimation, parameter in the literature. In [5], an experimental approach
calibration, simulation, modeling, model validation. in estimation of the parameters of a dc motor is
one of these methods. Motor armature winding re-
sistance and inductance, Back-EMF constant, mo-
1. Introduction tor torque constant, moment of inertia and friction
coeficient are calculated throughout the experi-
Diagnostics and model development for elec- ments. The measurements of the armature current
tro-mechanical systems are quite important in the are lattened by using the discrete time Kalman
applications of practical control system design. Filter and angular velocity is calculated.
System diagnostics includes development of a In [6], another study on the experimentally
dynamic system model with the the input-output modeling of PM-DC motor and parameter esti-
measurements obtained from a real system. The mation can be seen. First of all, by using standard
purpose is to establish a reliable mathematical linear differential equations, a three-mass system
model usually an ordinary differential equations model has been developed. Then, it is modeled
(ODE) or a transfer function representing the in- with numerical input-output data and system lin-
put-output relationship of the system. The aim of ear differential equations. The acquired model has
this model is to predict response, controls and the been tested with the discrete-time identiication al-
redirection of the system in the desired course. gorithm. The values experimentally acquired from
The system diagnosis process is constituted of over the model outputs and motor set experiments
experimental planning, data acquisition, model are observed to be close to each other. The mod-

606 Volume 6 / Number 3 / 2011

 technics technologies education management

el has been veriied with the Root-mean-square driver are established in Matlab-Simulink environ-
(RMS) error method. Open loop and closed loop ment. Measurements and experiments have been
with PI controller tests have been made. performed to acquire data necessary for estimat-
Linear and non-linear models can be estab- ing the motor parameters deined in the simulation
lished in order to ind the parameters of a system. model. Estimated parameter values are implement-
Basic non-linear parameters like Coulomb fric- ed into the simulation model and evaluated by ob-
tion and dead zone are included in the non-linear serving real-time response of the dc motor. Then,
model. Hammerstein non-linear system approach Matlab-Parameter Estimation tool has been used
is used to estimate the non-linear system model. for parameter calibration while acquiring approxi-
Thus, a concurrent model of a system like a dc mate values with regard to the real-time output re-
motor can be established by using the linear and sponse of the motor.
non-linear models.[7]
The Hartley modulating functions (HMF) meth-
od can be used for physical parameter estimation 2. Experiment apparatus setup
of the nonlinear continuous-time systems, where
a frequency-weighted least squares formulation is Experiment apparatus setup in this study is con-
applied base on input-output records over a inite stituted of an Electric Machines Test Bed (EMTB),
time interval. In [8], thyristor driven dc-motor pa- an externally-excited wound-pole brushed dc mo-
rameters are experimentally estimated by using the tor with driver, a DAQ card installed computer
HMF-method. Another approach in the modeling and a digital multimeter.
of non-linear systems is using NARMAX (Nonlin- Our testing unit, TQ Equipment Company FH2
ear AutoRegressive Moving Average with eXog- model (EMTB) is designed to test common elec-
enous input) model which is a method giving quite tric machines in the range with various fractional
accurate and eficient results. This method is based horse-powers like Three Phase Wound Pole Motor
on a general parametric model that is constituted (FH100), Step Motor (FH150) and Wound-Pole
of the polynomials that contain various linear and Brushed DC motor (FH50). It has a taco-genera-
non-linear concepts which unite the inputs, out- tor for measuring the rotor speed (rpm), a strain-
puts and modeled system errors. In [9], a dc mo- gauge for obtaining the rotor shaft torque (Nm)
tor is modeled with the NARMAX approach. MLP and a Foucault-brake for altering the load of the
(Multilayer Perceptron) network is used in order to motor. By the help of analog IO ports that EMTB
establish the structure of the NARMAX model. has; measured speed, applied reference voltage to
In [10], DC Decay test has been performed on the motor and motor load values can be transferred
a 4,5 KW ASEA synchronous machine which is to the DAQ installed computer. A standard PC and
one of the necessary methods of estimating the pa- a 100 Ks/S, 12 bit DAQ card is used in this setup.
rameters of synchronous machine. DC Decay test For measuring the motor current, a multimeter that
includes holding a machine’s rotor at a certain posi- has a computer connection via serial port (rs232)
tion and applying a small amount of dc voltage and and software support, is preferred. Experiment ap-
acquiring of the parameters by experimental tech- paratus setup can be seen in Figure 1.
niques. This method is also called standstill time-
domain test or DC Decay test. This method is quite
advantageous and worth noticing since it does not
need any special test equipments.
This study is focused on estimating the model
parameters of a wound-pole brushed dc motor ex-
perimentally to develop a linear model of the sys-
tem by using ODEs. Open-loop experiments have
been performed in order to deine and calibrate the Figure 1. The view of FH2 EMTB and experiment
response of the sensors and the motor driver. Then, apparatus setup.
simulation models of the dc motor and the motor

Volume 6 / Number 3 / 2011 607

 technics technologies education management

Figure 2. Overall structure of connection scheme in the experiment apparatus setup

The overall electrical signal and connection

scheme between the elements of this experiment d ia (t )
apparatus setup as in Figure 2. ea ( t ) = ia ( t ) Ra + La + e b ( t ) ..... (1)
dt

If La is neglected in correlation (1) [13];

2.1 The Structure of the DC Motor and
Mathematical Correlations
ea (t ) = ia (t ). Ra + eb (t ) .................... (2)
The motor used in the experiment is TQ Equip-
ment Companies’ FH50 model which has both se- the result will be as shown in correlation (2).
rial and shunt winding; in this study, it is used as Back-EMF is as follows;
an externally-excited dc motor. When the machine
is operated as a shunt motor and its armature and eb (t ) = K b .ω (t ) ......................... (3)
ield are fed with approximately 110 V, the rotor
speed increases approximately to 1500 rpm with Shaft torque related to the electrical part;
no-load condition [12]. The electro-mechanical
structure of the motor is as shown in Figure 3.
Tm (t ) = K m .ia (t ) ........................ (4)

Shaft torque related to the mechanic part;

d ω (t )
Tm ( t ) = J + B ω ( t ) + T L ( t ) ....... (5)
dt
is as shown above. The terms and explanations
of the terms used in these correlations are deined
as follows:
ea (t ) = Armature Voltage (110 V)
Figure 3. The electro-mechanical structure of the ia (t ) = Armature current (1 A)
used dc motor.
Ra = Armature resistance (Ohm)
According to the above scheme, the correla- eb (t ) = Back-EMF (V)
tions that will be used to calculate the parameters ω (t ) = Angular velocity (rad/s)
of the dc motor are as follows [13]:

608 Volume 6 / Number 3 / 2011

 technics technologies education management

K b = Back-EMF constant (V/rad/s) In the other method for estimating K b ; the mo-
tor is operated as a dc generator while measur-
Tm (t ) = Shaft torque (Nm)
ing the different values it generates at different
J = Inertia (Nm/rad/s2)
B = Friction (Nm/rad/s) rotation-speeds, as a result, the value of K b has
been acquired approximately as 0.7182. Now that
TL (t ) = Load torque (Nm)
this value is close to the value acquired in the irst
K m = Torque constant (Nm/A)
method, found K b value in the irst method will
rad s .
be used as 0.6739
3. Estimation of the motor parameters V

3.1 Estimation of Armature resistance ( Ra ) 3.2 The Estimation of Friction (B) :

and Back EMF Constant ( K b )
According to the correlation (5) above, deriva-
If every term in correlation (2) is divided by ia; tive term is zero since the speed will be constant
in steady-state and therefore the equation in the
ea (t ) ω (t ) correlation (5) will be:
= Kb . + Ra .................... (6)
ia (t ) ia
d ω (t )
= 0 ⇒Tm ( t ) = B ω ( t ) + TL ( t )
(6) is derived. For the estimation of the param- dt
eters in (6), the voltage applied to the motor has A graphic is drawn by using measured shaft
been started from 0 and increased. The rotation-
speed of the shaft and motor current are measured torque Tm (t ) and angular velocity ω (t ) values
then the graphic in Figure 4 is acquired. while the motor is operated under a certain load
torque TL (t ) . B and TL (t ) values are again calcu-
lated according to the equation found in the graph-
ic by curve itting. In this experiment, while the
motor is operating at the nominal voltage value
(110 V), it is loaded until 0.1 Nm is read in the
torque gauge. After this process, the motor rota-
tion-speed is decreased gradually and according to
Tm (t ) and ω (t ) values at each rotation-speed, the
graphic in Figure 5 are acquired.

Figure 4. Estimation of armature resistance and

Back-EMF constant

After applying curve itting to this graphic,

y = 0.6 7 39 x + 31.8 2 5 equation has been ac-
quired. When this equation is matched with the
equation (6), Back-EMF constant and armature
resistance is acquired as follows:
Figure 5. The Estimation of Friction (B).
rad s
K b = 0.6739 , Ra = 31 .825Ω
V After the curve itting process on the graphic,
y = 0.0001x + 0.0865 equation has been ac-

Volume 6 / Number 3 / 2011 609

 technics technologies education management

quired. When this equation is matched with the current at the instance of irst rotation and torque
equation Tm (t ) = B.ω (t ) + TL (t ) , B = 0.0001 and constant K m .
TL = 0.0865 are acquired.
Ta = K m .iamax ........................... (8)
As for acceleration, it can be derived from the
3.3 The Estimation of Torque Constant ( K m ) : slope of the time-speed graphic that will be ac-
quired when the nominal operating voltage is ap-
The correlation (4) can be used to estimate the plied to the dc motor.
torque related to the electrical part. Torque con- ωn
stant can be found by using the current absorbed aw = .............................. (9)
tx
by the motor and measured shaft torque. In the ex-
periment done for this purpose, the motor has been
operated at nominal values in the no-load condi-
tion, then the applied torque has been increased
starting from 0 and Tm (t ) values corresponding to
ia (t ) values are recorded, the graphic in Figure 6
has been acquired as shown below.

Figure 7. Acceleration graphic and the estima-

tion of inertia

In the current measurement process, the irst

rotation current of the motor is measured as
i a m a x = 1.7 3 A . From the acceleration graphic in
Figure 7, it can be observed that the slope value is
Figure 6. Torque - current graphic. approximately 268.29. Accordingly it is acquired
as follows:
After the curve itting process on the graphic,
y = 0.5024 x - 0.0547 equation is found, and T a = K m i a m a x ⇒ 0.5 3 4 9*1.7 3 = 0.9 2 5 38 N m
from this equation K m ≅ 0.5024 is acquired.
Ta 0.9 2 5 38
J= ⇒ = 0.0 0 3 4 4 917 812 8 N m
aw 2 6 8. 2 9 rad s2
3.4 The Estimation of Inertia (J ) :
After covering all the calculations, the param-
Inertia is basically related to acceleration eters that belong to the motor have been approxi-
mately acquired as follows:
torque Ta (t ) and angular velocity a w (t ) as can also
be seen in the correlation (7) below [11]. R a = 31.8 2 5 Ω , K b = 0.6 7 39
V
rad s
Ta (t )
J= ............................. (7) K m = 0.5 0 2 4 N m , B = 0.0 0 01 N m
a w (t ) A rad s

Acceleration torque can be derived from the

J = 0.0 0 3 4 4 9 N m
multiplying of the maximum value of the absorbed rad s2

610 Volume 6 / Number 3 / 2011

 technics technologies education management

4. Analysis and calibration of the estimated have been acquired, the model of the experiment
parameters system has been established in Simulink environ-
ment as seen in Figure 9.
In order to test the accuracy of these estimated
parameters; the model of the system is needed
to be established and the model should be simu-
lated by using estimated parameters. The results
acquired from simulation and the real system re-
sponses are needed to be compared. With the ex-
periments done so far, the parameters of the dc
motor which is the main component in the system
model have been acquired.
As it can be seen in Figure 2, the motor is Figure 8. The three different step response of the
driven by an electronic-driver which generates dc dc motor driver
output voltage (motor operating voltage) between
0-110 VDC with a reference input signal between After the Simulink Simscape model have been
0-6 VDC. For this reason, the response of the established, the parameters acquired by the exper-
driver is determined before observing the motor iments, performed in section 3, have been imple-
model. Three different levels of step inputs have mented into the simulation. After the comparison
been applied to the driver and input – output real- of the real system response and the simulation
time responses are observed. As a result, the fol- results, it has been observed that the simulation
lowing graphic is obtained representing the driver results are almost coincident with the real-time re-
characteristic (Figure 8). sponse. In order to have better simulation results,
In Figure 8, the response of the driver with re- parameter-calibration task has been applied by us-
gard to the three different step inputs demonstrates ing Simulink-Parameter Estimation tool, in which
certain incline and decline slopes, as 22.5 and the values of unknown parameters in both contin-
-20.4(V/s), respectively. After the driver responses uous-time and discrete-time models are estimated

Figure 9. The Simulink Simscape model of the experiment system

Volume 6 / Number 3 / 2011 611

 technics technologies education management

and/or known parameters are calibrated by using the use of “non-linear least squares” method with
experimental data. This is especially useful when “Trust-Region-Relective” algorithm in the tool,
some parameters cannot be measured experimen- which only focuses on inding local minimums.
tally or estimated parameters are needed to be cali- Thus, deining the range for the parameters to be
brated within acceptable values. When the tool is predicted improves the accuracy for the results of
utilized, it compares the real-time measured data the calibration process and reduces the computa-
with the data generated by the Simulink model. tional time elapsed by optimization algorithms.
Using optimization techniques, the software esti- Calibrated and validated motor parameters are
mates the parameters and (optionally) initial con- seen in Table 1 with respect to the estimated pa-
ditions of the states, to minimize a user-selected rameters in section 3.
cost function. The cost function typically calcu-
lates a least-square error between the empirical
and model data signals [14].
Parameter-Estimation tool is started within the
system model created in Simulink environment.
Then input data, which will be applied both to the
model and the real system, is prepared and applied
to real system. Real system response is recorded
since it will be a reference for the tool while cali- Figure 10. Deining estimation parameter value
brating the parameter values. After input and out- ranges within Parameter-Estimation tool.
put data are imported into the tool, the parameters
to be calibrated and their value ranges are deined.
When assumed that more than one parameter will 5. Discussion
be calibrated within the tool, the change in the
each parameter values is expected to affect the When the estimated and calibrated parameters
others. So, it’s important to deine value ranges in in Table 1 are examined, there exists more ratio-
consideration with the calculated values. For ex- nal change in Friction (B) -800.00%- and Inertia
ample, Ra is estimated as 31 .825Ω in section 3.1 (J) -44.97%- values than in the other parameters.
and by measuring with the ohmmeter, the value This is considerable because in physical systems
of Ra is observed approximately 30 ohm. There- like dc motor, estimating B and J values are hard-
fore, the acceptable initial-guess value range for er than the other parameter’s estimation. In Fig-
ure 11, graphic shows the motor’s simulation re-
Ra can be deined as 27 ohm minimum and 35 sponse with the estimated parameters in section 3
ohm maximum in the tool. versus the simulation response with the calibrated
The results of our experiments show that, ig- parameters in section 4 and the motor real output
noring the initial guess min/max values may cause response.
the tool to calculate insensible parameter values,
like inding Ra or J negative. This may be due to

Table 1. Estimated and calibrated parameter values

 
K b  V  B  N m  J Nm 
Parameters Ra (Ω )
 rad s 
Km ( Nm
A ) 
rad 


 r a d s 2 

 s 
Estimated 31.825 0.6739 0.5024 0.0001 0.003449
Calibrated 31.606 0.6798 0.5065 0.0009 0.0050
Rate of Change -0.69% 0.88% 0.82% 800.00% 44.97%

612 Volume 6 / Number 3 / 2011

 technics technologies education management

Figure 11. DC motor non-calibrated vs. calibrated parameters simulation and real output response

6. Conclusion the estimated parameters are calibrated by using

Simulink-Paratmeter Estimation tool. Due to the
Many physical systems can be represented by new parameter values acquired as the result of the
using ordinary differential or difference equations estimation process, simulation results and the real
with variables, referred to a mathematical model. system responses are observed to coincide with
Primarily, important components and correla- great proximity.
tions that interact within the system are identiied. As a result of this study, mathematical model of
Variables (parameters) present in the correlations a wound-pole brushed dc motor has been identi-
are calculated by analyzing the internal correla- ied based on the velocity output data. The identi-
tions and external responses of the system. Thus, ied model and parameters will be the base of fur-
a complete mathematical model of the system is ther studies related to the control of the same dc
acquired to estimate and control the system re- motor with different control algorithms.
sponses in different situations. When considered
that the systems are getting bigger and complicat-
ed nowadays, the importance of modeling the sys- 7. References
tem components and accurately estimating their
parameters will be understood better. 1. Söderstöm, T., Stocia, P., “System Identiication”,
A dc motor is one of the basic actuator that Prentice Hall International Series in Systems &
can be found in many complicated industrial or Control Engineering, 1994.
home-application systems. It converts electricity
2. Melsa, L., J., Sage, p., A., “System Identiication”,
into mechanical energy and it includes both elec- Mathematics in Science and Engineering, Vol. 80,
trical and mechanical components. In this study, a New York, Academic Press, 1971.
dc motor is modeled based on its velocity output
data and the parameters in the model have been 3. Ljung, L., “System Identiication: Theory for the
calculated with the experiments by using math- User”, Prentice Hall Information and System Sci-
ematical correlations. The identiied model and ences Series, 1998.
parameter values have been tested on the model
4. Chapman, J., S., “Electric Machinery and Power
established in the Simulink environment. When
System Fundamentals”, McGraw Hill, 2001.
the real response of the system and the simulation
results with estimated parameters are compared, 5. Samer, S., S., Raed, A., K., “Parameter Identiica-
it is observed that simulation acts almost like the tion of a DC Motor: An Experimental Approach”,
real system with little differences. At this point, Proceedings of the 8th IEEE International Confer-

Volume 6 / Number 3 / 2011 613

 technics technologies education management

ence on Electronics, Circuits & Systems, pp. 981-

984, September 2-5, 2001.

6. Eker, İ., “Open-loop and closed-loop experimental

on-line identiication of a three-mass electrome-
chanical system”, Esevier Mechatronics, Volume
14, Issue 5, pp. 549-565, June, 2004.

7. Kara, T., Eker, I., “Nonlinear modeling and identi-

ication of a DC motor for bidirectional operation
with real time experiments”, Energy Conversion
and Management 45, pp. 1087–1106, 2004.

8. S. Daniel-Berhe, S., Unbehauen, H., “Experimen-

tal physical parameter estimation of a thyristor
driven DC-motor using the HMF-method”, Con-
trol Engineering Practice 6, pp. 615–626, 1998.

9. Rahim, A., N., Taib, N., M. and Yusof, I., M., “Non-
linear System Identiication for a DC Motor using
NARMAX Approach”, AsiaSENSE, pp. 305–311,
2003.

10. Biró, K., A., Szabó, L., Iancu, V., Hedesiu, H.,
C. and Barz, V., “On The Synchronous Machine
Parameter Identiication”, Workshop On Electri-
cal Machines Parameters, May, 2001.

11. Ogata, K., “Matlab for Control Engineers”,

Prentice Hall, Oct 10, 2007.

12. TecQuipment Ltd. FH2/3 MkIV Electrical Ma-

chines Teaching System User’s Manual Book.

13. Ogata, K., “Modern Control Engineering”,

Prentice Hall, 1996.

14. Matlab Product Help & System Identiication

Toolbox™ 7, Lennart Ljung.

Corresponding author
Barış DOĞAN,
Research Assistant in Mechatronics Education Dept.,
Marmara University,
Turkiye,
E-mail: [email protected]

614 Volume 6 / Number 3 / 2011

View publication stats