IJE TRANSACTIONS B: Applications Vol. 30, No.

5, (May2017) 668-677

International Journal of Engineering RESEARCH

NOTE
Journal Homepage: www.ije.ir

Optimal Design of a Brushless DC Motor, by Cuckoo Optimization Algorithm

M. Niaz Azari*a, M. Samamib, S. M. Abedi Pahnehkolaeib
a Department of Electrical Engineering, University of Science and Technology of Mazandaran, Behshahr, Iran
b Department of Electrical Engineering, Sari Branch, Islamic Azad University, Sari, Iran

PAPER INFO A B S T R A C T

Paper history: This contribution deals with an optimal design of a brushless DC motor, using optimization algorithms,
Received 04 December 2016 based on collective intelligence. For this purpose, the case study motor is perfectly explained and its
Received in revised form 11 February 2017 significant specifications are obtained as functions of the motor geometric parameters. In fact, the
Accepted 10 March 2017
geometric parameters of the motor are considered as optimization variables. Then, the objective
function has been defined. This function consists of three terms i.e. losses, construction cost and the
Keywords: volume of the motor which should be minimized simultaneously. Three algorithms i.e. cuckoo, genetic
Brushless DC Motor and particle swarm have been studied in this paper. It is noteworthy that, cuckoo optimization
Cuckoo Algorithm algorithm has been used for the first time for brushless DC motor design optimization. A comparative
Objective Function study between the mentioned optimization approaches shows that, cuckoo optimization algorithm has
Optimal Motor Design been converged to optimal response in less than 250 iterations and its standard deviation is 0.03 ,
while the convergence rate of the genetic and particle swarm algorithms are about 400 and 450
iterations with standard deviations of 0.07 and 0.06 , respectively for the case study motor. The
obtained results show the best performance for cuckoo optimization algorithm among all mentioned
algorithms in brushless DC motor design optimization.
doi: 10.5829/idosi.ije.2017.30.05b.06

1. INTRODUCTION1 maximum efficiency with minimal material cost.

Consequently, a comparative experiment among non-
The use of DC motors has become common in industry, dominated sorting genetic algorithm, multi-objective
due to highlighted specifications such as vast speed particle swarm and (OMOCRO) shows the best
control and high efficiency [1-3]. However, the presence performance of (OMOCRO) for BLDC motor design
of commutator and brushes can be considered as a optimization. Reference [6], has proposed a novel
major disadvantage of such motors due to constant optimization method, search region management
erosion of the mentioned components which can finally (SRM), in order to improve the efficiency of the local
lead to an increase in safety hazard and the maintenance search algorithms. The mentioned method has been
cost. But this problem has been solved by the use of tested for optimal design of a BLDC motor with the
Brushless DC (BLDC) motors. In these motors electric help of FEA, in order to minimize the torque ripple. It
circuits have been applied instead of commutator and has been stated in literature [7] a Multi-objective Krill
brushes [2, 4]. Herd Algorithm (MOKH), using the beta distribution in
So far, several investigations have been conducted the inertia weight tuning, has been proposed for
on optimization of BLDC motor design. As stated in electromagnetic optimization of a Brushless DC Motor
literature [5], a BLDC motor has been optimized by with a promising performance.
orthogonal multi-objective chemical reaction Another investigation [8] has proposed the genetic
optimization algorithm (OMOCRO), in order to achieve algorithm for topology optimization of the stator teeth in
a BLDC motor in order to reduce the torque ripple
*Corresponding Author’s Email: [email protected] (Milad
without decreasing the average torque. Son et al. [9]
Niaz Azari) optimized BLDC motor through a population based
algorithm called interstellar search method (ISM) with

Please cite this article as: M. Niaz Azari, M. Samami, S. M. Abedi Pahnehkolaei, Optimal Design of a Brushless DC Motor, by Cuckoo
Optimization Algorithm, International Journal of Engineering (IJE), TRANSACTIONSB: Applications Vol. 30, No. 5, (May 2017) 668-677
669 M. Niaz Azari et al. / IJE TRANSACTIONS B: Applications Vol. 30, No. 5, (May 2017) 668-677

mesh adaptive direct search. Reduction of the torque TABLE 1. Parameters of BLDC motor
ripple has been considered as the main objective of this number of pole pairs P winding thickness lw (m)
paper. Another study [10] deals with the optimal design cross sectional area Ac
of an interior permanent magnet BLDC motor, using mechanical air gap lg (m)
of the winding (mm2)
cost effective ferrite magnets in order to maximize the
pole–arc per pole–
flux density and minimize the torque ripple. The genetic β rotor radius rr (m)
pitch ratio
algorithm has been applied for flared shape rotor
structure optimization. Kim et al. [11] have optimized magnet thickness lm (m) current density Jcu ( Am 2 )
the anisotropic ferrite magnet shape and magnetization wire gauge and
stator/rotor core
direction of an interior Permanent Magnet BLDC Motor ly (m) stator/rotor axial ls (m)
thickness
in order to maximize back-EMF of the mentioned motor length
with the help of (FEM). On the other hand, a 2-D
analytical solution to predict the distribution of
magnetic field and comparing the results with 2-D
(FEM) in ironless BLDC motor, used in flywheel, has
been raised by Liu et al. [12]. Investigation [13],
discusses an outer rotor type motor design, used in the
blower system of a vehicle in accordance to a BLDC
and also BLAC motor with the help of finite element
analysis.
In most of literatures mentioned that the influence of
the required speed has been neglected in optimization
and as a result, the motor power has not been well
defined [5-10]. On the other hand, the applied
optimization approaches are based on simple analysis
with sensitivity to initial conditions which have been
widely used in recent years. Therefore, employing a
more up to date optimization algorithm seems to be
vital. This investigation provides a detailed study in
order to represent the essential equations for BLDC Figure 1. The structure of the studied BLDC motor
motor design, considering: Both speed and torque as
mechanical required parameters and, using cuckoo
optimization algorithm (COA) as a suitable approach rotor core flux density (Br), should be given in the knee
for motor optimal design. To this end, the geometric point of the B-H curve.
parameters of the motor are considered as the Assuming the conductor and the magnetic field, are
optimization variables. Then the objective function is orthogonal to each other, the total torque can be
defined, based on minimization of losses, construction obtained as follows [14, 15]:
cost and the volume of the motor. Finally the obtained T  Aw J cu k f k c lB r (1)
results of the three optimization approaches have been
compared and the COA has been extracted as the best Aw   lw (2rr  2l g  lw ) (2)
method.
where, l and k c represent the length of the conductor
and the correction factor, respectively and Aw is the
2. MATERIALS OF A BLDC MOTOR AND THE
APPLIED METHODS cross section of the coil.
Regardless of the armature reaction and also the
2. 1. BLDC Motor Structure Figure 1 depicts the reluctance of the stator and rotor core, the magnetic flux
structure of the studied motor in addition to its density will be as follows [14, 15]:
geometrical parameters [14]. Furthermore, the shown Fm Brlm
Bg  
parameters, in Figure 1, have been introduced in Table Ag   rr  l g  lw  (3)
(rr  l g )ln  
1.  rr  l m 

where, Fm , is the magneto-motive force and  is the

2. 2. Design Features
2. 2. 1. Electromagnetic Torque For obtaining total reluctance of each winding. A g can be obtained as:
the total torque, the specifications which depend on the 
body material of the BLDC motor such as filling factor Ag  ls (rr  l g ) (4)
p
of the coil (kf), permanent magnet and the stator and
 M. Niaz Azari et al. / IJE TRANSACTIONS B: Applications Vol. 30, No. 5, (May 2017) 668-677 670

The electromagnetic torque of the BLDC motor based R and L represent the resistance and the inductance of
on its geometric parameters can be expressed in each winding, respectively.
accordance to Equation (5).
2. 2. 3. Cost of Materials The volume of the applied
 k f k c k 1k  B r l m l s lw (2rr  2l g  lw )J cu
T em  materials which depend on the motor geometry have
r l l  (5) significant impacts on the motor cost as expressed in
ln  r g w 
 rr  l m  Equation (9).
The leakage component of the magnetic field ( k 1 ) and C  C m  Cw  C y (9)
also the active area of the auxiliary coil and magnet
where, C m , C w and C y represent the costs of
( k  ) are expressed with the help of Equation (6) and
permanent magnet, winding and stator/rotor core,
(7), respectively.
respectively. Each term of Equation (9), can be written
k1  1
1 in detail as follows.
(6)
0.9[rr / (  p (l g  lw ))]2  1
C m  c m 1mV m  c m 2 p (10)

 ( , k c )
k  (7) Cw  cw A g k f wV w (11)
kc

 , indicates the span of the active coils, locating in C y  c y  yV t (12)

the PM magnetic field, as shown in Figure 2. This
parameter can be approximated by the following where, c m 1 , cw and c y are the costs per unit mass of
equation.
permanent magnet, winding and core materials,
  min( , k c )[k s  (1  k s ) tanh(    k c )] (8) respectively.  m , w and  y represent the mass densities
of permanent magnet, winding and rotor/stator core,
It is noteworthy that, k s  1 and  is obtained by respectively. Finally, V m , V w and V t are the volumes
experience and testing. of the permanent magnet, winding and rotor/stator core,
respectively.
2. 2. 2. The Mechanical and Electrical Criteria
For making a relation between the motor geometry and 2. 2. 4. Losses in BLDC Motors Losses in BLDC
the desired velocity, the electrical and also mechanical motors are divided into three categories i.e. electrical,
criteria should be defined in order to restrict the magnetic and mechanical losses. The power loss due to
rotational velocity. From the mechanical aspect, the resistance of windings can be obtained as follows.
bearings are able to withstand high rotational speeds.
Therefore, they hardly impose any limitation on Pcu   k f k c k et Aw l s J cu2 (13)
rotational velocity. But on the other hand, other rotating
parts, specially permanent magnet can impose On the other hand, the eddy currents and hysteresis
limitations on the maximum rotational velocity. As a losses are considered as two major magnetic losses of a
result, a non-magnetic rotating sleeve is applied in order BLDC motor.
to enhance the mechanical robustness of the rotor. Assuming equal magnetic flux of the air-gap and the
From electrical point of view, the electrical time core, the maximum magnetic density of the stator is
constant (   L / R ), can limit the maximum rotational expressed as [14]
velocity.  k 1 B r l m
B sy 
 rr  l g  lw  (14)
2 pl y ln  
 rr  l m 

Thus, the following equations are obtained for eddy

current and hysteresis losses, respectively.
Pe  k e  yV sy B sy2 f 2
(15)

Ph  k h  yV sy B syn f (16)

where,  y is the density of the motor material and V sy

Figure 2. Concept of  represents the stator volume. It should be noted that, the
671 M. Niaz Azari et al. / IJE TRANSACTIONS B: Applications Vol. 30, No. 5, (May 2017) 668-677

frequency in Equations (15) and (16) is calculated in problems, large number of parameters and obtaining
accordance to Equation (17). multiple local optima. On the other hand, PSO, has the
same advantages as GA, but with better computational
f  pw r / 2 (17) efficiency by applying statistical analysis and formal
On the other side, the mechanical losses in a BLDC hypothesis testing. But this study has applied the COA
motor can be divided into two categories including; for optimal design of BLDC motor for the first time.
friction and windage. The friction losses can be written The cuckoo optimization algorithm has superiority to
as follows [14]. many other optimization algorithms i.e. GA and PSO,
typically for multi-objective functions. In COA, the
Nb local search is performed with higher efficiency because
Pb  f Fb d i r (18)
2 there is only a single parameter apart from the
population. In fact the only parameter which should be
where, Fb and d i are the load and the internal radius
adjusted is the fraction of the nests needed to be
of the bearing. On the other hand,  f and N b represent abandoned (Pa). This issue improves the computing
the bearing friction factor and the number of bearings power and speed. GA and PSO are common algorithms
respectively. The windage losses can be obtained as and have been completely described in references [17-
follows. 21]. But since, COA, has been used as the main
algorithm for motor optimal design in this study, an
Pw   k rC f air r3rr4l s (19) initial understanding from the concept of these
algorithms is presented as follows.
where, k r and air represent the roughness factor of the
rotor and the air density respectively and C f is the 3. 1. Cuckoo Optimization Algorithm Cuckoo
friction factor which is obtained by Equation (20) in Optimization Algorithm (COA) has been inspired by the
which Re is the Couette-Reynolds number. life of a bird, called cuckoo [16]. The initial population
of COA which forms various societies, consists of
0.5150  g r 
 l /r
0.3

for 500  Re  104 cuckoos and eggs. Each cuckoo has some eggs and also
 Re0.5 an Egg Laying Radius (ELR). The cuckoos lay eggs
Cf   (20)
 l g / rr 
0.3
 inside their equivalent ELR and in the nests of other
0.0325 for 10  Re4
host birds. Among all the eggs, those ones, which are
 Re0.2
similar to the eggs of the host birds can grow up. The
Re  airw r rr l g / air rate of grown eggs indicates the suitability of the area.
The area with more remained eggs has higher profit.
Cuckoos always search for areas with highest profit for
 k f k c k 1k  B r l m l s lw (2rr  2l g  lw )J cu Pe  Ph
T em   egg laying. Therefore, selecting the best place is an
r l l  wr (21) important term which should be optimized by the
ln  r g w 
 rr  l m  cuckoos. The cuckoos which live in the worst habitats
always are removed. Each cuckoo travels a specific
Tout  Tem  (Pw  Pb ) / r (22) percent of the whole path toward the ideal habitat with a
clarified deviation which are known as  and 
In accordance to the obtained magnetic and mechanical
respectively. These two parameters help the cuckoos to
losses, the modified formula of the electromagnetic and
find the ideal habitat. The maximum number of cuckoos
also the output torque can be modified as Equations (21)
should be confined in the specific environment. In fact,
and (22). Also, the total losses of a BLDC motor can be
cuckoos have been clustered and the best habitat is
expressed as:
detected to achieve the objective point. Consequently,
Ptotal  Pcu  Ph  Pe  Pb  Pw (23) the new cuckoo population can travel to the objective
habitat. Now, the survival of eggs in the nest are
checked and the profit value is obtained. A suitable
profit value can lead to stopping the process. Otherwise,
3. OPTIMIZATION METHODS the whole process should start from the beginning in
accordance to the flowchart, presented by Amiri and
These paper apply three different evolutionary
Mahmoudi [16]. In fact, the survival process of cuckoos
algorithms i.e. Cuckoo (COA) [16], genetic (GA) [17,
should finally converge to a condition with only one
18] and particle swarm (PSO) [19-21], for optimal
cuckoo society, containing the same profit values.
design of the BLDC motor. GA is a popular and applied
algorithm because of several reasons such as, its high
intuitiveness, ease of implementation, its high capability 3. 2. Determination of the Effective Parameters
to solve highly nonlinear mixed integer optimization for the Optimization Methods Accordance to
 M. Niaz Azari et al. / IJE TRANSACTIONS B: Applications Vol. 30, No. 5, (May 2017) 668-677 672

literatures [16-21], the determinave parameters of the TABLE 4. Constant parameters of the BLDC motor
optimization algorithms are introuduced. This Quantity Amount Quantity Amount
parameters are  and  for COA; the cross over rate kf wp
0.7 0.02
and also the percentage of mutation for GA and ( C 1 &
C 2 ) for PSO. ( C 1 & C 2 ) determine the traveled kc 0.666 wv 2000/3
distance of a particle in each iteration [19-21]. The
ks 0.95 wc 0.0125
amount of the aforementioned parameters of the
optimization algorithms, are measured in 20 different kr 1 m (kg m-3 ) 7400
conditions. Each measurement is implemented
individually for about 50 times and finally, The  5 w (kg m-3 ) 8900
effective parameters, mentioned in Table 2, are obtained
from the eights, twentieth and fourth implementation of B r (T ) 1  y (kg m-3 ) 7700
COA, GA and PSO respectively.
B syknee (T ) 1.5 c m 1 (£ kg -1 ) 20

4. OPTIMIZATION PROBLEM DATA  (A 2 m 3 ) 1011 c m 2 (£) 1

4. 1. Design Variables and Constant Values The  (m ) 1.8  108 c y (£ kg -1 ) 3

optimization variables are those parameters of the case
study motor that should be optimized. These parameters k h' (W s kg 1T n ) 0.018a c1 (£ mm2 kg-1 ) 0.045
are presented in the following vector.
k e' (W s 2 kg 1T 2 ) 0.00008a c 2 (£ kg -1 ) 5.42
x  [ p  l m l y lw l g l s rr J cu Ac ]T (24)
n 1.92a T em* (Nm ) 10
Generally, there are 10 different design variables that
should be optimized. Other quantities i.e. power losses,  1 w r* (rad s-1 ) 157
output torque, costs and volume of the motor can be
calculated through them. The aforementioned design V * (v ) 140
variables were defined in Table 1. while, the constant
a
parameters of this motor is presented in Table 4. For M19 lamination with a thickness of 0.35mm
In addition to these parameters, other adjustable
coefficients of the COA are also presented in Table 3.
f 0 (x )  w VV t (x )  w P Ptotal (x )  w CC (x ) (25)
4. 2. Objective Function, with Considering the
Constraints An appropriate definition of the where, C (x ) , Ptotal (x ) and V t (x ) represent, cost
objective function with consideration of the constraints
function, power loss function and volume function of
is always known as the most significant issue in solving
the motor, respectively. On the other hand, w C , w P and
any optimization problem. The main concern in this
investigation includes reduction of costs, volume and w V are the related weight of cost function, power loss
power loss of a BLDC motor. For this reason, the function and volume function of the motor. In fact,
objective function will be as follows: these coefficients clarify the impact of each function.
In addition to electrical and mechanical constraints,
some other limitations such as thermal, cost and
TABLE 2. The value of effective optimization parameters manufacturing constraints are of great importance. The
COA GA PSO only electrical constraint is the voltage which can be
obtained with appropriate selection of the winding
  Cross over rate Mutation (%) C1 C2
diameter. Similarly, the mechanical constraints can be
0.05 30 0.8 0.1 1.6 2 expressed as follows.
T em  T em*
 max (26)
r  r
*
TABLE 3. Other COA coefficients
Number of Cuckoos 40 Max number of Cuckoos 200
Min number of eggs 2  30
where, T em* and  r* are arbitrary torque and speed
Max number Of eggs 5  0.05 respectively. rmax is considered as the maximum speed
Number of clusters 2 Max iteration 500
in the arbitrary torque. It should be noticed that, the
manufacturing constraints contain some parameters i.e.
673 M. Niaz Azari et al. / IJE TRANSACTIONS B: Applications Vol. 30, No. 5, (May 2017) 668-677

minimum air-gap ( l gmin ) and the minimum area of the Equations (3) and (4) are needed to define Equation
section ( A ). min (11). The total losses of a BLDC motor can be
c
expressed as Equation (23). This equation has 5 terms
Other constraints, caused by thermal limitations and including, Equations (13), (15), (16), (18) and (19).
saturation effect, are expressed as follows. Equation (14) is essential for defining Equations (15)
B sy  B syknee and (16) and also Equation (20) is necessary for
 (27) defining Equation (19). For clarifying the impact of
k f lw j cu  k
2
each function the related weight of cost function, power
loss function and volume function of the motor is
where, B syknee represents, the magnetic flux density at the
considered as in Equations (25). Equations (26) and (27)
knee point of the B-H curve and k is the maximum express the constraints of the objective function. By
permissible temperature of the windings. considering the constraints in Equation (25), the final
After considering the impact of the electromagnetic objective function is shown as Equation (28). By means
torque, speed and magnetic flux density constraints, the of optimization algorithms which are implemented on
objective function has been modified as follows. Equation (28), the geometric parameters and also the
volume of the motor will be optimal in addition to
f 0 (x )  w VV t (x )  w P Ptotal (x )  w C C (x )  losses and manufacturing cost simultaneously.
1 T em w rmax B  (28)
f u (1  T * )  f u (1  )  f u ( sy knee  1) 
 em w r* B sy 
5. SIMULATION RESULTS AND DISCUSSION
1
f u (x ) 
1  e  x 5. 1. Technical Analysis of the Optimized
Parameters After implementation of the algorithms
where,  is a tiny amount and  is considered as a
according to the effective parameters, mentioned in
constant large number.
subsection 3.2; design variables and constant values,
mentioned in subsection 4.1 and also the modified
4. 3. Summary and Discussion For implementation
objective function, presented in subsection 4.2, the
of the BLDC motor optimization problem, the
optimal parameters of the case study motor are
significant specifications of the motor are obtained as
obtained. These values along with the minimum and
functions of the motor geometric parameters. The
maximum values of the parameters are given in Table 5.
geometric parameters are mentioned in Table.1. In fact
It should be noted that, GA and PSO results are
the mentioned parameters of the motor are considered as
validated with reference [14]. According to Table 5,
optimization variables and other quantities i.e. power
when COA is applied, most of the geometrical
losses, output torque, costs and volume of the motor can
optimized parameters i.e.  , l m , lw , l g and l s have
be calculated and optimized through them. The
objective function consists of three terms including, the lowest values. Therefore, the motor has the lowest
losses, motor volume and manufacturing cost. The cost possible volume and lowest cost. On the other hand, the
term is calculated by Equation (9), this Equation can be cross sectional area of the winding and the current
written in detail in accordance to Equations (10)-(12). density ( Ac and J cu ) are also more applicable.

TABLE 5. The limitations and optimal value of the motor

No Parameters Min Max COA GA PS0
1 β 0.5 1 0.6912 0.7 0.6950

2 (m) l m 0.001 0.015 0.0120 0.0130 0.0124

3 (m) l y 0.002 0.01 0.0081 0.0060 0.0058

4 (m) lw 0.001 0.0055 0.0033 0.0035 0.0034

5 (m) l g 0.001 0.004 0.001 0.001 0.001

6 (m) rr 0.005 0. 1 0.0592 0.0595 0.059

7 (m) l s 0.006 0.6933 0.0730 0.0756 0.0732

8 (mm2) Ac 0. 1 2 1.9551 2 1.9982

-2
9 (Am ) J cu 3  106 6  106 5819800 5800000 5784573
 M. Niaz Azari et al. / IJE TRANSACTIONS B: Applications Vol. 30, No. 5, (May 2017) 668-677 674

As a result, the objective function has the best value, In accordance to Figure 3a, the large number of
using COA, as presented in Table 6. Another significant poles causes an increment in the motor manufacturing
issue in any optimization approach is the convergence cost and also a decrement in the magnetic losses due to
rate of the algorithm. The COA, converged after 250 low density of magnetic flux in stator and rotor core. It
iteration while the GA and PSO converge after 400 and should be noted that, this issue has no impact on the
450 iterations, respectively. This issue indicates the volume of the motor. By considering the proposed
suitable convergence rate of COA. objective function and Figure 3a, it is concluded that,
applying higher number of poles can lead to a better
5. 2. The Impact of Motor Geometrical design of the motor. However, this issue causes an
Parameters on the Objective Function Figure 3 increment in the leakage magnetic flux and a decrement
shows the variation of the objective function, due to in the output torque. According to Figure 3b, the value
changing each geometrical parameter of the motor, of  has no effect on the volume of the motor. But it
while the rest of parameters remain constant. This figure is noteworthy that, a lower value of  can lead to a
is divided into 9 subfigures and is labeled from (a) to (i).
reduction in the cost, magnetic leakage flux, magnetic
Each subfigure depicts the impact of changing each
losses and also the output torque. On the other hand, a
geometrical parameter of the motor i.e. P ,  , l m , l y ,
large value of  can lead to a decrement in the output
lw , rr , l s , l g and J cu , on the objective function, torque due to its effect on increasing the magnetic
respectively. leakage flux.
In all the subfigures, the red line, blue line and green Figure 3c shows that, reduction of l m leads to
point represent the proposed objective function improvement of manufacturing cost, volume and losses
variation, unconstrained objective function and the of the BLDC motor. But on the other hand, the output
optimal point, respectively. torque and the maximum speed of the motor still keep
decreasing as before. As shown in Figure 3d, an
increment in l y will lead to a reduction in the magnetic
TABLE 6. Specifications of the optimized BLDC motor losses and causes an increment in the cost and volume
No Parameters COA-Value GA-Value PSO-Value of the motor. It should be noted that, considering a very
small value for l y , may lead to saturation. Figure 3e,
1 V t (m 3 ) 0.0011 0.00116 0.0012
shows that, the existence of high space for winding has
2 C (£ ) 65.6417 68.86 64.51 different impacts on the cost and volume of the motor.
But in general, it can lead to improvement of the
3 (W) Ptotal 51.2446 56.71 52.3 efficiency. It is noteworthy that, lw should not be
4 (W) Pcu 42.1851 44.81 40.69 lower than a permissible amount.
Otherwise, the motor will not be able to produce a
5 (W) Ph 4.4115 6.18 6.04 suitable torque. The radius of the rotor ( rr ) is
6 (W) Pe 2.4500 3.52 3.46
considered as one of the most significant parameters in
a BLDC motor design. As shown in Figure 3f, by
7 (W) Pb 2.1195 2.12 2.12 reducing the rr value, all the three items in the objective
function will be reduced simultaneously. But it should
8 (W) Pw 0.0783 0.08 0.0697
be noted that, a very small rr value has negative effect
9 W vV t 0.7891 0.776 0.75 on producing the necessary output torque and on the
10 W cC 0.8205 0.861 0.804
other hand, a very large rr value has inverse effect on
the maximum speed of the motor. Figure 3g depicts
11 W p Pl 1.0248 1.136 1.043
that, a small value of l s is favorable, but unfortunately,
12 fo 2.59 2.78 2.71 this small amount can lead to an adverse impact on the
13 Efficiency 96.61% 96.54% 96.46% output torque of the motor.
This issue is not desirable. As shown in Figure 3h,
14 Standard 0.03 0.07 0.06
l g which shows the air-gap amount, is considered to
15 V (Volt) 1.3034e+02 141.1 136 have its minimum value.
16 I (Ampere) 11.6115 11.6 11 Figure 3i indicates that, the output torque and the
(W) Pout copper losses are proportional to the current density and
17 1462 1460 1461.31
the square of current respectively.
675 M. Niaz Azari et al. / IJE TRANSACTIONS B: Applications Vol. 30, No. 5, (May 2017) 668-677

a b c
15 15 15
Unconstrained
10 Constrained 10 10
Optimal
5 5 5

0 0 0
1 2 3 4 5 6 0.5 0.6 0.7 0.8 0.9 1 0 5 10 15
P Beta Lm
d e f
15 15 20
Objective value

15
10 10
10
5 5
5

0 0 0
2 4 6 8 10 1 2 3 4 5 6 20 40 60 80 100
Ly Lw Rr
g h i
20 8 8

15
6 6
10
4 4
5

0 2 2
0 2 4 6 8 0 2 4 6 8 3 3.5 4 4.5 5 5.5 6
Ls Lg Jcu

Figure 3. Objective function variation, due to changing each geometrical parameter of the BLDC motor

Although, increasing the current density can lead to TABLE 7. The comparison of applied methods
an improvement in the cost and volume of the motor, Algorithm Fitness Standard deviation Convergence rate
but it is noteworthy that, the impact of an increment in COA 2.59 0.03 250
the copper losses is able to overcome the two
aforementioned advantages. GA 2.78 0.07 400

PSO 2.71 0.06 450

5. 3. The Comparison of COA Performance with

GA and PSO Figure 4 and Table 7 show the
By considering, Table 7 and Figure 4, it is concluded
convergence rate, fitness and also the standard deviation
that, COA algorithm has the best performance among
of COA, GA and PSO algorithms.
all the described algorithms for optimal design of a
BLDC motor.
3.05
COA-Convergence
3 GA-Convergence 6. CONCLUSION
2.95 PSO-Convergence

2.9 In this paper, an optimal design of a BLDC motor, using

2.85 three optimization approaches i.e. COA, GA and PSO
Fitness

2.8 has been studied. The priorities of parameters

2.75
optimization in design of a motor are different in
various applications. Therefore, the importance of this
2.7
problem has become more obvious due to simultaneous
2.65
parameter optimization. This investigation firstly
2.6 clarifies significant specifications of the motor as
2.55
0 50 100 150 200 250 300 350 400 450 500
functions of the motor geometric parameters. Then, the
Iteraation objective function has been defined in order to minimize
Figure 4. The convergence waveform comparison of COA, the losses, construction cost and the volume of the
GA and PSO motor simultaneously. Three different optimization
 M. Niaz Azari et al. / IJE TRANSACTIONS B: Applications Vol. 30, No. 5, (May 2017) 668-677 676

approaches i.e. COA, GA and PSO have been applied 10. Yoon, K.-Y. and Kwon, B.-I., "Optimal design of a new interior
permanent magnet motor using a flared-shape arrangement of
for the case study motor optimal design. It is noteworthy
ferrite magnets", IEEE Transactions on Magnetics, Vol. 52,
that, COA has been used for the first time for this No. 7, (2016), 1-4.
purpose. The obtained results of three optimization 11. Kim, H.-s., You, Y.-M. and Kwon, B.-I., "Rotor shape
methods have been compared together and finally it is optimization of interior permanent magnet BLDC motor
concluded that, COA can converge to an optimal according to magnetization direction", IEEE Transactions on
response in less than 250 iterations, while this value is Magnetics, Vol. 49, No. 5, (2013), 2193-2196.
400 and 450 iterations for GA and PSO, respectively. 12. Liu, X., Hu, H., Zhao, J., Belahcen, A., Tang, L. and Yang, L.,
"Analytical solution of the magnetic field and emf calculation in
As a result, the proposed method has an acceptable ironless BLDC motor", IEEE Transactions on Magnetics, Vol.
convergence rate. On the other hand, the obtained 52, No. 2, (2016), 1-10.
fitness value and the standard deviation of COA is more 13. Lee, T.-Y., Seo, M.-K., Kim, Y.-J. and Jung, S.-Y., "Motor
applicable, compared with GA and PSO. design and characteristics comparison of outer-rotor-type BLDC
motor and blac motor based on numerical analysis", IEEE
Transactions on Applied Superconductivity, Vol. 26, No. 4,
(2016), 1-6.
7. REFERENCES
14. Rahideh, A., Korakianitis, T., Ruiz, P., Keeble, T. and Rothman,
M., "Optimal brushless DC motor design using genetic
1. Pyrhonen, J., Jokinen, T. and Hrabovcova, V., "Design of algorithms", Journal of Magnetism and Magnetic Materials,
rotating electrical machines, John Wiley & Sons, (2009). Vol. 322, No. 22, (2010), 3680-3687.
2. Tian, W., "Design of permanent magnet brushless DC motor 15. Lai, S. H., "Design optimisation of a slotless brushless
control system based on dspic30f4012", Procedia Engineering, permanent magnet DC motor with helically-wound laminations
Vol. 29, (2012), 4223-4227. for underwater rim-driven thrusters", University of
3. Afjei, E., Hashemipour, O., Saati, M. and Nezamabadi, M., "A Southampton, (2006),
new hybrid brushless DC motor/generator without permanent 16. Amiri, E. and Mahmoudi, S., "Efficient protocol for data
magnet", International Journal of Engineering Transactions B clustering by fuzzy cuckoo optimization algorithm", Applied
Applications, Vol. 20, No. 1, (2007), 77-86. Soft Computing, Vol. 41, (2016), 15-21.
4. Zolfaghari, M. and Taher, S., "Fuzzy approximation model- 17. Arashloo, R. S., Martinez, J. L. R., Salehifar, M. and Moreno-
based robust controller design for speed control of bldc motor", Eguilaz, M., "Genetic algorithm-based output power
International Journal of Engineering-Transactions C: optimisation of fault tolerant five-phase brushless direct current
Aspects, Vol. 28, No. 3, (2014), 426-432. drives applicable for electrical and hybrid electrical vehicles",
5. Duan, H. and Gan, L., "Orthogonal multiobjective chemical IET Electric Power Applications, Vol. 8, No. 7, (2014), 267-
reaction optimization approach for the brushless DC motor 277.
design", IEEE Transactions on Magnetics, Vol. 51, No. 1, 18. Yousefi, A., Rahmani, A., Farzanegan, A. and Rostami, S.,
(2015), 1-7. "Simulation and genetic algorithms for optimizing comminution
6. Lee, T.-Y., Trung, P. X., Kim, J.-W., Kim, Y.-J. and Jung, S.-Y., circuit at gol-e-gohar iron plant (research note)", International
"Search region management method for local search algorithm Journal of Engineering-Transactions C: Aspects, Vol. 26, No.
employing design optimization of brushless dc motor", IEEE 6, (2013), 663-670.
Transactions on Magnetics, Vol. 52, No. 3, (2016), 1-6. 19. Kazerooni, K., Rahideh, A. and Aghaei, J., "Experimental
7. Ayala, H. V., Segundo, E. H., Mariani, V. C. and Coelho, L. d. optimal design of slotless brushless pm machines based on 2-D
S., "Multiobjective krill herd algorithm for electromagnetic analytical model", IEEE Transactions on Magnetics, Vol. 52,
optimization", IEEE Transactions on Magnetics, Vol. 52, No. No. 5, (2016), 1-16.
3, (2016), 1-4. 20. Nahvi, H. and Mohagheghian, I., "A particle swarm optimization
8. Ishikawa, T., Yonetake, K. and Kurita, N., "An optimal material algorithm for mixed variable nonlinear problems", International
distribution design of brushless DC motor by genetic algorithm Journal of Engineering, Vol. 24, No. 1, (2011), 65-78.
considering a cluster of material", IEEE Transactions on 21. Akbarpour, H., Karimi, G. and Sadeghzadeh, A., "Discrete multi
Magnetics, Vol. 47, No. 5, (2011), 1310-1313. objective particle swarm optimization algorithm for FPGA
9. Son, B., Park, G.-J., Kim, J.-W., Kim, Y.-J. and Jung, S.-Y., placement", International Journal of Engineering
"Interstellar search method with mesh adaptive direct search for Transactions C: Aspects, Vol. 28, No. 3, (2015), 410-418.
optimal design of brushless DC motor", IEEE Transactions on
Magnetics, Vol. 52, No. 3, (2016), 1-4.
‫‪677‬‬ ‫‪M. Niaz Azari et al. / IJE TRANSACTIONS B: Applications Vol. 30, No. 5, (May 2017) 668-677‬‬

‫‪Optimal Design of a Brushless DC Motor, by Cuckoo Optimization‬‬ ‫‪RESEARCH‬‬

‫‪NOTE‬‬
‫‪Algorithm‬‬
‫‪M. Niaz Azaria, M. Samamib, S. M. Abedi Pahnehkolaeib‬‬
‫‪a‬‬ ‫‪Department of Electrical Engineering, University of Science and Technology of Mazandaran, Behshahr, Iran‬‬
‫‪b‬‬ ‫‪Department of Electrical Engineering, Sari Branch, Islamic Azad University, Sari, Iran‬‬

‫‪PAPER INFO‬‬ ‫چكيده‬

‫‪Paper history:‬‬
‫‪Received 04 December 2016‬‬ ‫این مقاله به بررسی طراحی بهینه از یک موتور بدون جاروبک جریان مستقیم با استفاده از الگوریتمهای بهینهسازی که بر‬
‫‪Received in revised form 11 February 2017‬‬
‫‪Accepted 10 March 2017‬‬ ‫پایه هوش جمعی هستند می پردازد‪ .‬برای این منظور‪ ،‬ابتدا موتور مورد مطالعه کامالً توضیح داده شده و مشخصههای مهم‬
‫آن به شکل توابعی از پارامترهای هندسی موتور به دست آورده شده است‪ .‬در حقیقت‪ ،‬پارامترهای هندسی موتور به عنوان‬
‫‪Keywords:‬‬
‫‪Brushless DC Motor‬‬ ‫متغیرهای بهینهسازی در نظر گرفته شدهاند‪ .‬سپس تابع هدف تعریف گشته است‪ .‬این تابع متشکل از سه بخش میباشد که‬
‫‪Cuckoo Algorithm‬‬
‫‪Objective Function‬‬ ‫شامل تلفات‪ ،‬هزینه ساخت و حجم موتور میگردند که میبایست به شکل همزمان مینیمم گردند‪ .‬در این مقاله سه‬
‫‪Optimal Motor Design‬‬
‫الگوریتم بهینهسازی شامل الگوریتمهای فاخته‪ ،‬ژنتیک و اجتماع ذرات مورد مطالعه قرار گرفتهاند‪ .‬شایان ذکر است که‪،‬‬
‫الگوریتم فاخته برای اولین بار برای بهینهسازی طراحی موتور بدون جاروبک مورد استفاده قرار گرفته است‪ .‬یک مطالعه‬
‫مقایسهای ما بین الگوریتمهای یاد شده نشان میدهد که‪ ،‬الگوریتم فاخته در کمتر از ‪ 250‬تکرار و با انحراف استانداردی‬
‫معادل ‪  0.03‬به همگرایی می رسد‪ .‬در حالیکه نرخ همگرایی الگوریتم ژنتیک و اجتماع ذرات به ترتیب ‪ 400‬و ‪ 450‬با‬
‫انحراف استانداردی معادل‪  0.07‬و‪  0.06‬میباشد‪ .‬نتایج به دست آمده نشاندهنده بهترین عملکرد الگوریتم فاخته مابین‬
‫الگوریتمهای یاد شده به منظور بهینهسازی طراحی موتور بدون جاروبک جریان مستقیم میباشد‪.‬‬
‫‪doi: 10.5829/idosi.ije.2017.30.05b.06‬‬