Available online at www.sciencedirect.

com

ScienceDirect
Procedia Manufacturing 4 (2015) 322 – 327

Industrial Engineering and Service Science 2015, IESS 2015

Parameter identification of BLDC motor model via metaheuristic

optimization techniques
Danupon Kumpanyaa; Sattarpoom Thaiparnatb; Deacha Puangdownreongc*
a
Faculty of Engineering and Architecture, Ragamangala University of Technology Suvarnabhumi (RUS), Suphanburi, Thai-land.
b
Faculty of Business Administration and Information Technology, RUS, Suphanburi, Thailand.
c*
Department of Electrical Engineering, Graduate School, South-East Asia University, Bangkok, Thailand.

Abstract

The brushless dc (BLDC) motor has been increasingly used in industrial automation, automotive, aerospace, instrumentation and
appliances. Analysis and design of the BLDC motor efficiently require its accurate model and parameters. In this paper, the
parameter identification of the BLDC motor model via well-known metaheuristic optimization search techniques is proposed.
Two trajectory-based methods, i.e. adaptive tabu search (ATS) and intensified current search (ICS) are employed to estimate the
BLDC motor parameters. As simulation results of model identification and validation, both ATS and ICS can provide optimal
BLDC model parameters. The BLDC models obtained show a very good agreement to actual system dynamics. However, the
ICS can pro-vide optimal model parameters faster than the ATS.

© 2015
© 2015TheTheAuthors.
Authors. Published
Published by Elsevier
by Elsevier B.V.B.V.
This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the organizing committee of the Industrial Engineering and Service Science 2015 (IESS
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer review under responsibility of the organizing committee of the Industrial Engineering and Service Science 2015 (IESS 2015)
2015).

Keywords: Parameter identification, bldc motor model, adaptive tabu search, intensified current search

1. Introduction

Since 1970’s, the brushless dc (BLDC) motor has been used in different applications, for example, industrial
automation, automotive, aerospace, instrumentation and appliances [1]. The BLDC motor retains the characteristics
of a brushed DC motor but eliminates the commutator and the brushes. It can be driven by DC voltage source, while-

* Corresponding author. Tel.: +0-000-000-0000 ; fax: +0-000-000-0000 .

E-mail address: deachap@sau.ac.th

2351-9789 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer review under responsibility of the organizing committee of the Industrial Engineering and Service Science 2015 (IESS 2015)
doi:10.1016/j.promfg.2015.11.047
 Danupon Kumpanya et al. / Procedia Manufacturing 4 (2015) 322 – 327 323

the current commutation is electronically done by solid state switches [2]. These make it needs less maintenance and
in many applications the brushed DC motors are replaced by the BLDC motors. By literatures, the BLDC motor
have many advantages over brushed DC motors. Among those, higher speed range, higher efficiency, better speed
versus torque characteristics, longer operating life, noiseless operation and higher dynamic response can be found in
BLDC usage [1,2,3]. Regarding to control context, analysis and design of the BLDC motor control system require its
accurate model. Generally, the BLDC model can be derived as a kind of synchronous machine [4,5]. In another
approach, the model of the BLDC can be developed in the similar manner of brushed DC motor approach [6]. The
problem is how to provide the accurate model parameters of the BLDC motor. By theory, such the model parameters
can be obtained by the blocked-rotor, no-load and retardation tests [7]. However, obtained parameters by these tests
often inaccurate due to its uncertainty and error in testing process.
Recently, metaheuristic optimization techniques have become potential candidates to various engineering
applications. Among the trajectory-based metaheuristic optimization techniques, the adaptive tabu search (ATS) and
the intensified current search (ICS) are the most efficient methods. The ATS has been launched in 2004 [8,9] as a
modified version of the conventional tabu search for solving various engineering optimization problems especially in
control synthesis and design problems. Some applications of the ATS include model and parameter identification
[10] and PID controller design [11]. For the ICS, it has been proposed in 2014 [12] as a modified version of the
conventional current search to solve engineering optimization problems. The ICS has been successfully applied to
design an optimum PID controller for the AVR system [12]. In this paper, both ATS and ICS metaheuristics are
applied to identify the optimal BLDC model parameters. Results obtained from the identified models are compared
with actual system dynamics of the BLDC motor used in this work.

2. BLDC mathematical model

The mathematical model of the BLDC motor can be developed in the similar manner of brushed DC motor
approach. For the brushed DC motor, the model can be derived as the linear transfer function expressed in (1) or (2)
[6], where Ra is armature resistance, La is armature inductance, J is moment of inertia, B is viscous friction

constant. Th
:(s) Kt . (1)
G( s)
V ( s) ( JLa )s 2  ( Ra J  La B)s  ( Ra B  Kt Ke )
:( s) 1/ Ke 1/ Ke . (2)
G( s)
V (s) § Ra J ·§ La · 2 § Ra J · W mW e s 2  W m s  1
¨¨ ¸¸¨¨ ¸¸s  ¨¨ ¸¸s  1
© Kt K e ¹© Ra ¹ © Kt K e ¹
Ra J La .
Wm , We (3)
Kt K e Ra
Based on the principal of Hall Effect, the BLDC motor can be operated in many modes (or phases). However, the
3-phase has a very good precision in control. To perform the BLDC model, the major addition is the phase effecting
the overall result of the BLDC model. The phase peculiarly effects the resistance and the inductance of the BLDC
arrangement. Generally, a simple arrangement of a symmetrical (balanced) 3-phase wye (Y) connection as depicted
in Fig. 1 could provide a brief illustration of all phase concept. With symmetrical arrangement, the mechanical time
written from (3) as expressed in (4) [6].
Commonly, the BLDC motor will be driven by DC input voltage via the particular invertor (or power amplifier).
Such the amplifier can be approximated by the first order model. Therefore, the mathematical model of the BLDC

constant.
324 Danupon Kumpanya et al. / Procedia Manufacturing 4 (2015) 322 – 327

R J J6Ra J (3Ra ) L La La .
Wm ¦ K aK Kt K e Kt K e
, We ¦ Ra 6Ra 3Ra
(4)
t e a

:(s) § K A ·§ 1/ K e ·
G( s ) ¨¨ ¸¸¨ ¸. (5)
¨ ¸
© W A s  1 ¹© W mW e s  W m s  1 ¹
2
V ( s)
i(t)
ia
a
Ra B
T(t) Z(t)
ib La
v(t) n J
La La
Ra Ra
b c
ic
invertor BLDC motor

Fig. 1. Brushless DC motor schematic diagram with symmetrical 3-phase wye (Y) connection

3. Algorithms

Adaptive Tabu Search (ATS). The ATS [8,9] is one of the trajectory-based metaheuristic optimization techniques
based on the stochastic approach and local search principle. It uses the tabu list (TL) and the aspiration criteria as the
keys. The ATS possesses two distinctive mechanisms denoted as backtracking (BT) to escape from entrapment and
the adaptive radius (AR) to speed up the search process. Algorithms of the ATS can be summarized as the pseudo
code shown in Fig. 2(a).
Intensified Current Search (ICS). The ICS [12] is also one of the trajectory-based metaheuristic optimization
techniques based on the stochastic approach and multiple local search. The ICS possesses the memory list (ML) to
escape from entrapment, the adaptive radius (AR) and the adaptive neighborhood (AN) to speed up the search
process. Algorithms of the ICS can be represented as the pseudo code shown in Fig. 2(b).
 Danupon Kumpanya et al. / Procedia Manufacturing 4 (2015) 322 – 327 325

Initialized:
- Objective function f(x), x=(x1,…,xd)T,
- Initialize the search space :,
- tabu list (TL),
- search radius R, k=1.
- maximum allowance of solution cycling jmax,
- number of neighborhood members n,
- Random initial solution x0 within :.
- Xglobal=Xlocal =x0.
while (k<=kmax or termination criteria: TC);
Random neighborhood xi around x0 within R.
Evaluate f(xi) and set the best one as x*.
if f(x*)<f(x0); Update x0=x*, set j=1
and keep x0 in TL.
else keep x* in TL and update j=j+1. end
if (j== jmax);
- Invoke BT by employ any x in TL.
- Activate AR by R=UR, 0<U<1. end
Update Xlocal=x0.
if f(Xlocal)<f(Xglobal); Update Xglobal=Xlocal. end
Update k=k+1.
end
(a)
Fig. 2. Pseudo codes; (a) ATS, (b) ICS
Initialized:
- Objective function f(x), x=(x1,…,xd)T,
- Initialize the search space :,
- maximum allowance of solution cycling jmax,
- number of initial solutions N,
- number of neighborhood members n,
- search radius R, k=j=1.
- Random initial solution Xi within :.
- Evaluate f(Xi) then rank and Xi into <.
- Let x0=Xk as selected initial solution.
- Xglobal=Xlocal =x0.
while (k<=N or termination criteria: TC);
while (j<=jmax);
Random neighborhood xi around x0 within R.
Evaluate f(xi) and set the best one as x*.
if f(x*)<f(x0); Update x0=x* and set j=1.
else Update j=j+1. end
Activate AR by R=UR, 0<U<1.
Invoke AN by n=Dn, D>1.
end
Update Xlocal=x0.
if f(Xlocal)<f(Xglobal); Update Xglobal=Xlocal. end
Update k=k+1 and set j=1.
Let x0=Xk as selected initial solution.
end
(b)

4. Identification problem formulation

Metaheuristic-based parameter identification of BLDC motor model can be represented by the block diagram in
Fig. 3. Referring to the model in (5), this identification problem is finding the appropriate 5 parameters, i.e. Ke, KA,
Wm, We and WA. Regarding to Fig. 3, let f(˜) be the objective function. In this work, the objective function f(˜) is
formulated as the mean of sum absolute error (MSAE) between the experimental speed data, Z(t), and the simulated
speed data from the model, Z*(t). The objective function f(˜) will be fed back to the metaheuristic block (ATS or
ICS) to be minimized by searching optimal parameters that provide good agreement between the experiment and the
model. The experimental setup for the BLDC motor is shown Fig. 4. The E-Tricycle DC blushless motor ZJTEX
(model ZJ-80) in laboratory is conducted. Its rates are of 48-72 V, 800-1500 W, 3650 rpm, 8-pole and wye-
connection. In this work, two specific levels of the BLDC motor speed, i.e. 1400 rpm and 1600 rpm, are tested and
recorded by a 300 MHz SIGLENT digital storage scope. The experimental speed data at 1400 rpm are used in
identification process, while those at 1600 rpm are used in validation process.

Z(t)
BLDC motor + f( )
v(t)

BLDC math model Z (t)
-

Metaheuristics
(ATS, ICS)
Fig. 3. Metaheuristic-based BLDC identification
326 Danupon Kumpanya et al. / Procedia Manufacturing 4 (2015) 322 – 327

Digital storage scope

BLDC motor

Fig. 4. BLDC experimental setup

5. Results and discussions

The algorithms of ATS and ICS were coded by MATLAB run on Intel(R) Core(TM) i5-2430M, 2.4 GHz, 8 GB-
RAM computer. The search parameters of two candidate metaheuristic techniques are priori set for a fair
comparison. For the ATS, the number of neighborhood member n = 10, the search radius R = 0.5 and maximum
allowance of solution cycling jmax = 10. For AR mechanism; (i) if f(˜) < 0.5, then R = 0.25, (ii) if f(˜) < 0.1, then R =
0.1, (iii) if f(˜) < 0.01, then R = 0.01. For the ICS, the number of initial solutions (search directions) N = 10, the
number of neighborhood member n = 10, the search radius R = 0.5, maximum allowance of solution cycling jmax =
10 and the maximum iteration of each search direction = 100. For AR and AN mechanisms; (i) if f(˜) < 0.5, then R =
0.25 and n = 15, (ii) if f(˜) < 0.1, then R = 0.1 and n = 20, (iii) if f(˜) < 0.01, then R = 0.01 and n = 25. For both ATS
and ICS, the search will be terminated once the f(˜) d 0.0010 set as the TC. The max-min boundaries of the model
parameters are preset to perform their corresponding search spaces as stated in (6).

K A [0, 1.5], K e [0, 1.5], W m [0, 2], W e [0, 0.5], W A [0, 0.5] (6)

For model identification, the experimental speed data at 1400 rpm of the BLDC are conducted. By applying the
ATS and ICS within few trial runs, the optimal parameters of the BLDC are successfully achieved as sum-marized
in Table 1. Plots of the model obtained by the ATS and ICS compared with the experimental speed at 1400 rpm can
be observed in Fig. 5. It was found that the models obtained show the best agreement to the experimental speed.
However, the ATS consumes more search time than the ICS does. For validating model, the experimental speed data
at 1600 rpm are employed. Plots of the models compared with the experiment can be observed in Fig. 6. It was
found that both models by the ATS and ICS show the best agreement to the actual dynamics.

Table 1. Obtained model parameters of BLDC motor

Methods Model parameters MSAE MSAE Search time
KA Ke Wm We WA (Iden) (Valid) (sec.)
ATS 1.2473 0.4207 0.3110 0.0001 0.0014 0.0010 0.0033 243.46
ICS 1.2102 0.4101 0.3112 0.0001 0.0013 0.0009 0.0029 124.78
Note Search times are the average values of 10 trial runs.

6. Conclusions

Metaheuristic-based parameter identification of BLDC motor model has been proposed in this paper. Two most
efficient trajectory-based methods, i.e. adaptive tabu search (ATS) and intensified current search (ICS) have been
applied to estimate the BLDC motor parameters. As simulation results of model identification and valida-tion, both
ATS and ICS can provide optimal model parameters. Both BLDC models obtained by the ATS and ICS show an
 Danupon Kumpanya et al. / Procedia Manufacturing 4 (2015) 322 – 327 327

equally very good quality of agreement to actual system dynamics behavior. However, the ICS can provide the
optimal model parameters of the BLDC motor faster than the ATS.

Motor speed
2

Experimental speed
1
BLDC Model by ACS
BLDC Model by ICS
0
0 1 2 3 4 5 6 7
Time (sec.)

Fig. 5. Plots of BLDC model identification

3
Motor speed

Experimental speed
1
BLDC Model by ACS
BLDC Model by ICS
0
0 1 2 3 4 5 6 7
Time (sec.)

Fig. 6. Plots of BLDC model validation

References

[1] P. Yedamale, “Brushless DC (BLDC) Motor Fundementals, ” AN885 Michrochip Technology Inc., 2003.
[2] A. Tashakori, M. Ektesabi and N. Hosseinzadeh, “Modeling of BLDC motor with ideal back-EMF for automotive applications,” The
World Congress on Engineering 2011 No.2, 2011.
[3] G. Prasad, N.S. Ramya, P.V.N. Prasad and G.T.R. Das, “Modelling and simulation analysis of the brushless DC motor by using
MATLAB”, International Journal of Innovative Technology and Exploring Engineering (IJITEE), pp. 27-31, 2012.
[4] Y.S. Jeon, H.S. Mok, G.H. Choe, D.K. Kim, and J.S. Ryu, “A new simulation model of BLDC motor with real back EMF waveform, ”
Computers in Power Electronics, pp. 217-220, 2000.
[5] L. Hao and H.A. Toliyat, “BLDC motor full speed range operation including the flux – weakening region, ” Industry Applications
Conference, pp. 618-624, 2003.
[6] V.K.R. Singh Patel and A.K. Pandey, “Modeling and performance analysis of PID controlled BLDC motor and different schemes of
PWM controlled BLDC motor,” International Journal of Scientific and Research Publications, pp. 1-14, 2013.
[7] P. Vas, Parameter Estimation, “Condition Monitoring and Diagnosis of Electrical Machines, ” Oxford University Press, 1993.
[8] D. Puangdownreong, S. Sujitjorn and T. Kulworawanichpong, “Convergence analysis of adaptive tabu search, ” ScienceAsia Journal of
the Science Society of Thailand, No.30, pp. 183-190, 2004.
[9] S. Sujitjorn, T. Kulworawanichpong, D. Puangdownreong and K.-N. Areerak, “Adaptive tabu search and applications in engineering
design, ” Frontiers in Artificial Intelligent and Applications, IOS Press, Amsterdam, Netherlands, 2006.
[10] D. Puangdownreong, K-N. Areerak, K-L. Areerak, T. Kulworawanichpong and S. Sujitjorn, “Application of adaptive tabu search to
system identification, ” IASTED International Conference on Modelling, Identification, and Control (MIC2005), pp. 178-183, 2005.
[11] D. Puangdownreong and S. Sujitjorn, “Obtaining an optimum PID controller via adaptive tabu search, ” Lecture Notes in Computer
Science 4432, pp. 747-755, 2007.
[12] A. Nawikavatan, S. Tunyasrirut and D. Puangdownreong, “Application of intensified current search to optimum PID controller design in
AVR system, ” AsiaSim-2014, pp. 255-266, 2014.