©Civil-Comp Press, 2012

Proceedings of the Eleventh International Conference

Paper 72 on Computational Structures Technology,
B.H.V. Topping, (Editor),
Civil-Comp Press, Stirlingshire, Scotland

Optimum Detailing Design of Reinforced Concrete

Plane Frames to ACI 318-05 using the
Harmony Search Algorithm
A. Akin1 and M.P. Saka2
1
Department of Civil Engineering, Yuzuncu Yil University, Van, Turkey
2
Department of Civil Engineering, University of Bahrain, Isa Town, Bahrain

Abstract
This paper presents the application of the harmony search based algorithm to the
optimum detailed design of special seismic moment reinforced concrete (RC) frames
under earthquake loads based on American Standard specifications. The objective
function is selected as the total cost of the frame which includes the cost of concrete,
formwork and reinforcing steel for individual members of the frame. The modular
sizes of members, standard reinforcement bar diameters, spacing requirements of
reinforcing bars, architectural requirements and other practical requirements in
addition to relevant provisions are considered to obtain directly constructible designs
without any further modifications. For the RC columns, predetermined section
database is constructed and arranged in the order of resisting capacity. The produced
optimum design satisfies the strength, ductility, serviceability and other constraints
related to good design and stated in the relevant specifications. The lateral seismic
forces are calculated according to ASCE 7-05 and it is updated in each iteration.
Number of design examples is considered to demonstrate the efficiency of the
optimum design algorithm proposed. It is concluded that the developed optimum
design model can be used in design offices for practical designs and this study is the
first application of the harmony search method to the optimization of RC frames and
also the optimization of special seismic moment RC frames to date.

Keywords: cost optimization, structural optimum design, harmony search

algorithm, reinforced concrete structures, heuristics.

1 Introduction
In recent years structural optimization has witnessed an emergence of robust and
innovative search techniques that strictly avoid gradient-based search to counteract
with challenges that traditional optimization algorithms have faced for years. The
basic concept behind each of these techniques rests on simulating the paradigm of a

1
biological, chemical, or social systems (such as evolution, immune system, swarm
intelligence or annealing process, etc.) that is automated by nature to achieve the
task of optimization of its own [1,2]. The design algorithms developed using these
meta-heuristic search techniques are particularly suitable for obtaining rapid and
accurate solutions to problems in structural engineering discipline [3,4]. This is
particularly true in the optimum design of steel structures where the design problem
turns out to be a discrete optimization problem when it is formulated according to
design codes used in practice.
In many practical engineering design problems, design variables may consist of
continuous or discrete variables. In structural optimization problems, the design
variables are functions of the cross sections of the members and they are often
chosen from a limited set of available values. For example, steel structural members
are chosen from standard steel profiles in the market, structural timber members are
provided in certain sizes, and reinforced concrete members are usually designed and
constructed with discrete dimensional increments. Design optimization of reinforced
concrete (RC) structures is more challenging than that of steel structures because of
the complexity associated with reinforcement design. Also, only one material is
considered for the optimization problems of steel structures and the structural cost is
directly proportional to its weight in general. But in the case of the optimum design
of concrete structures, three different cost components due to concrete, steel and
formwork are to be considered and the overall cost of the structure is affected from
any slight variation in the quantity of these components. Therefore, the optimization
problem of concrete structures becomes the selection of a combination of
appropriate values of design variables to make the total cost component the
minimum.
In the literature, there are a number of studies on optimization of RC members
and frames. Practical applications of traditional optimization methods are not
suitable for optimum design of RC frames and these algorithms require additional
modifications to fit the discrete nature of structural design variables. Choi and Kwak
[5] haves suggested more practical discrete optimization techniques. They used
direct search method to select appropriate design sections from some pre-determined
discrete member sections for the cost optimization of rectangular beams and
columns of RC frames based on the ACI and Korean codes. Balling and Yao used
the simulated annealing method which is one of the metaheuristic algorithm to
optimize three-dimensional reinforced concrete frames [6]. Discrete variables as
well as limits on the number of reinforcing bars and their topological arrangements
are considered in their study. Rajeev and Krishnamoorthy [7] applied a simple
genetic algorithm to the cost optimization of two-dimensional RC frames. The
detailing of reinforcement is considered as a design variable in addition to cross-
section dimensions. The allowable combinations of reinforcement bars for columns
and beams were tabulated. Camp et al. [8] used genetic algorithms (GAs) by
constructing a database for beams and columns which contains the sectional
dimensions and the reinforcement data in the practical range to optimize for
optimum design of plane frames. Lee and Ahn [9] used to the genetic algorithms to
optimize reinforced concrete plane frames subject to gravity loads and lateral loads.
In their study, the constructing data sets, which contain a finite number of sectional

2
properties of beams and columns in a practical range removed the difficulties in
finding optimum sections from a semi-infinite set of member sizes and
reinforcement arrangements. Kwak and Kim [10] studied on optimum design of RC
plane frames based on pre-determined section database. In their study, pre-
determined section databases of RC columns and beams are constructed and
arranged in order of resisting capacity and optimum solutions are obtained using
direct search method. They also used genetic algorithms for similar optimization
problems [11]. Govindaraj and Ramasamy [12] used genetic algorithms for optimum
detailed design for RC frames based on Indian Standard specifications. The
dimensions and reinforcement arrangement of column, and the dimensions of beam
members alone are considered as a design variables and the detailing of
reinforcements in the beam members is carried out as a sub-level optimization
problem. The modular sizes of members, available standard reinforcement bar
diameters, spacing requirements of reinforcing bars, architectural requirements on
member sizes and other practical requirements are arranged in order to obtain for the
optimum designs to be directly constructible without any further modifications.
This study deals with the problem of optimizing Special Seismic Moment
reinforced concrete (RC) frames subject to ACI 318-05 [13] and ASCE 7-05 [14]. In
the literature, there are a number of studies on optimization of RC members and
frames. In the studies available in the literature the shear design calculations of
concrete members are not considered and the cost of shear reinforcement (ties) is not
taken into consideration in the total cost of the frame. Only the simple constraints
such as capacity and regulations for flexural reinforcement are included. The
detailing of the reinforcement bars is oversimplified and the development length of
bars is not considered in the cost calculations. In some of these studies, the lateral
loading on the frame is considered; however, the values of the lateral loads are taken
as a constant even though the value of lateral loads change with the weight of the
structure subject to seismic specifications. The design algorithm presented in this
study considers covers all required code provisions and obtains the optimum
solution by using harmony search algorithm.

2 Modelling of Optimum Detailing Design Problem of

Reinforced Concrete Plane Frames
Reinforced concrete special moment frames are used as part of seismic force-
resisting systems in buildings that are designed to resist earthquakes. Beams,
columns, and joints in moment frames are proportioned and detailed to resist
flexural, axial, and shearing actions during an earthquake. These moment-resisting
frames are called “Special Moment Frames” because of the additional requirements.
The optimum design algorithm developed for the RC special seismic moment
frames obtains buildable optimum designs where the cross sections are selected
from the design pool. In the design formulation, the objective function is selected as
the cost of the RC structure which includes the cost of concrete, reinforcement and
formwork. The design variables are categorized in two general groups i.e. beam and

3
column members. For the columns, the section database which includes the
dimensions and the reinforcement detailing of column sections is constructed with
the most commonly used sections. The design constraints are implemented
according to ACI 318-05 which considers shear and the seismic design constraints
as well as flexural design constraints. The development lengths of the reinforcement
steel bars are calculated according to the given regulations in the design
specifications. The cost of the shear reinforcement and the impact of the
development length on the cost are considered. In the design of frames, the matrix
displacement method is used for the structural analysis and the load combinations
are taken from the ACI code. The self-weight of RC beams is included in the
structural analysis and it is updated in each iteration of the optimization process. The
lateral seismic forces are calculated according to ASCE 7-05 and it is updated in
each iteration according to the selected design weight.

2.1 Objective Function

The objective function is selected as the total cost of RC frame consisting of
individual cost components of concrete, steel and formwork which is inclusive of
material, fabrication, and labor. The objective function is expressed as,

Minimize
f cost = Cc + Cs + C f (1)
where, Cc = the cost of concrete, Cs = the cost of reinforcing bars, Cf = the cost of
formwork (includes labor and placement).
The cost of concrete, Cc is computed from Eqn. (2) as;

Cc = Vc ⋅ (UCc ) (2)

where, Vc = the total volume of concrete, (UCc ) = the unit cost of concrete ($/m3).

N col Nbeam
Vc = ∑ bi di Lcolumn, i + ∑b w, j h j Lclear beam, j (3)
i =1 j =1

where, Ncol = the number of column members, bi and di = the width and depth of ith
column member with rectangular cross section, Lcolumn, i = the length of ith column
member, Nbeam = the number of beam members, bw,j = the width of jth beam, hj = the
height of jth beam, Lclear beam, j = the clear span length of jth beam.
The cost of reinforcing steel, Cs is computed from Eqn. (3) as;

Cs = Ws ⋅ (UCs ) (4)

where, Ws = the total weight of steel used as reinforcement bar in the concrete frame,
(UCs) = the unit cost of steel ($/kg) .

4
 Ncol Nbar ,i Ncol Ntie ,i Nbeam Nbar ,m Nbeam Ntie ,m
Ws = ∑ ∑ As , j lbar , j + ∑ ∑ Ash,k ltie, k + ∑ ∑ As ,l lbar , l + ∑∑A l
st , n tie , n (5)
i =1 j =1 i =1 k =1 m =1 l =1 m =1 n =1

where, Nbar,.. = the number of longitudinal reinforcement bars placed in the member,
Ntie,.. = the number of ties used in the member, As,.. = the area of reinforcement bars,
lbar,.. = the length of reinforcement bars, Ast,.. , Ash,.. = the area of shear reinforcement
bars (ties) ltie,.. = the length of ties.
The cost of formwork, Cf is computed from Eqn. (4) as;

C f = Af ⋅ (UC f ) (6)

Where, Af = the total formwork area, (UCf ) = the unit cost of formwork ($/m2).

N col Nbeam
Af = ∑ {2(bi + di )Lcolumn , i − ( Areabeams @ joint,i )} + ∑ {(bw, j + 2h j )Lclear beam , j } (7)
i =1 j =1

where, Areabeams@joint,i = the cross-section area of beams connected to the ith column
at joint.
The unit costs are based on market prices and their values changes from time to
time and also from country to country. For these reasons, the unit price data cannot
be fixed and needs to be updated. In the previous studies in the literature, researchers
used different country design specifications for their design and different market
prices for unit costs depending to their countries.

2.2 Design Variables

Design variables are divided two groups as column design variables and beam
design variables and to obtain practical designs beams and columns are separated to
groups. Total number of design variables is determined according to number of
column and beam groups.

2.2.1 Beam Design variables

For beam design groups, the cross-section of beams, the area of reinforcement bars
along all beams, and the area of reinforcement bars placed on the top and bottom of
beam spans and supports are considered as design variables. The cross-sectional
dimensions of the beam are considered as the design variable. In addition to cross-
section dimension of the beam, the areas of longitudinal bars that are placed
continuously at the bottom of all the beams and the tensile reinforcements at the
spans of beams, and supports for each beam are also considered as the design
variable. These design variables and their numbers in the problem are defined in
Table 1. The design variables relevant with reinforcement bars are not defined as the
surface area of reinforcing bars. Instead the reinforcement bar layout is defined and
these design variables relevant with reinforcement are expressed in terms of the

5
number of reinforcing bars and the diameter of reinforcing bars. By this way, design
process can reach directly constructible optimum designs. The details of the design
variables are shown in Figure 1. The total number of the beam design variables,
Nbdv, for one beam group changes according to number of span (bay) in frame. Total
number of beam design variables for one beam group is calculated by Eqn.8;

N bdv = 5 + 4nbay (8)

The total number of beam design variables when each beam group considered,
Ndvbeam , is computed by following equation;

Ndvbeam = nbeamgroup ⋅ N bdv (9)

where, Nbdv = the total number of the beam design variables for one beam group,
nbay= the number of bays in a frame, nbeam group = the total number of beam group.

Design Variables Number

Xi,1 The width and the height of beam cross-section 1
Xi,2 The area of the steel reinforcement that continues 1
through the top of all the beams *
Xi,3 The area of the steel reinforcement that continues 1
through the bottom of all the beams *
Xi,4 -Xi,k The area of the top steel reinforcement at spans of nbay
beams*
Xi, k+1 –Xi,m The area of the bottom steel reinforcement at spans of nbay
beams*
Xi,m+1- Xi,n The area of the top steel reinforcement at the supports* nbay+1
Xi,n+1- Xi,Nbdv The area of the bottom steel reinforcement at supports* nbay+1
th
i= number of beam group (i beam group), nbay ; number of spans (k = 3+ nbay , m= 3+2nbay, n=
4+3nbay, Nbdv= 5+4nbay )
* For the reinforcement design variables, the areas of steel reinforcement bars are defined in terms
of the number and diameter of the bars (nφd n; number of bars, d; diameter of bars) to obtain
constructable reinforcement areas.

Table 1: Cross-sectional and reinforcement design variables for beam groups.

For the cross-sectional and reinforcement design variables, the design variable
pools are created as shown in Table 2. For the reinforcement design variables, the
values are selected such that they give constructable reinforcement areas. In other
words, the number and diameter of the bar for a beam can be considered as one
design variable (nφd n; number of bars, d; diameter of bars). For this purpose, the
variable pool table is composed of a combination of number and diameter of bars.
For the reinforcement design variables, the design variable pool is created by the
combination of number and diameter of reinforcement bars as shown in Table 2. It is
apparent from the design pool that 61 different combination of number and diameter
of a bar can be arranged for each reinforcement design variables. The 6
reinforcement bar combination are used for the reinforcement continues through the

6
top (Xi,2) and bottom (Xi,3) of all the beams. In this optimum design problem,
material strengths, unit costs of materials, structural geometry, support conditions,
loading conditions, and cover details are pre-assigned as design parameters at the
beginning of the optimization process. However, the value of the dead load which
includes the self-weight of beam depending on the cross-sectional dimensions is
automatically updated during the design cycles.

# X1 (mm) # X2 – X3 # X4 – XNbdv
1 250/400 1 2φ12 1 NNR
2 250/450 2 2φ14 2 1φ12
3 250/500 3 3φ12 3 1φ14
4 250/550 4 2φ16 4 1φ16
5 250/600 5 3φ14 5 1φ18
6 300/400 6 3φ16 6 1φ20
7 300/450 7 1φ22
8 300/500 8 1φ24
9 300/550 9 1φ26
10 300/600 10 1φ28
11 300/650 11 1φ30
12 300/700 12 2φ12
13 350/500 13 2φ14
14 350/550 14 2φ16
15 350/600 15 2φ18
16 350/650 16 2φ20
17 350/700 17 2φ22
18 400/600 . .
19 400/650 . .
20 400/700 45 5φ18
21 400/750 46 5φ20
22 400/800 47 5φ22
23 450/700 48 5φ24
24 450/750 49 5φ26
25 450/800
26 450/850
27 450/900 55 6φ18
28 500/700 56 6φ20
29 500/750 57 6φ22
30 500/800 58 6φ24
31 500/850 59 6φ26
32 500/900 60 6φ28
61 6φ30
* NNR ; no need reinforcement

Table 2: Variable pool for beam design variables.

7
 Figure 1: The reinforcement bar layout and the design variables for ith beam group.

2.2.2 Column Design variables

Unlike the design of steel structures, there is infinite set of member size and amount
of steel reinforcement used in the design of reinforced concrete (RC) buildings. In
addition to the discrete and combinatorial nature of the RC sections, the restrictions
and reinforcement detailing specified in the design specifications make the optimum
design of RC buildings even more complicated.
The design variables for the construction of column section database are selected as
the dimensions of columns in x and y directions, the diameter of reinforcement bars
at the cross-section of column and numbers of reinforcement bars in both sides of
the column as shown in Figure 2. The lower-upper bounds and increments of these
variables are given in Table 3. The selected feasible column sections are given in
Table 4. The compressive strength of concrete and the yield strength of steel are
taken as 30MPa and 400MPa, respectively, and the cover of the concrete section is
taken as 50mm to evaluate the strength of sections and to select the sections in the
database. In the optimum design process, the interaction diagrams for the assigned
sections are computed for the given material strength properties and the detailing of
the section.

8
 Increment Number of
Design Lower Upper
possible values
Variables bound Bound (Step size) in the range
X1c b (mm) 300 500 50 mm 5
X2c d (mm) 400 1000 50 mm 13
X3c φ (mm) { 14,16,18,20,22,24,26,28,30} 9
X4c and X5c n1 and n2 0 7 1 8

Table 3: Design variable bounds for RC column design examples.

# d b φ n1 n2 Pn,max (kN) Pb (kN) Mb (kNm)

1 400 300 14 1 1 2288,78 851 130,97
2 400 300 14 1 2 2363,72 848,46 142,13
3 400 350 14 2 1 2695,22 991,85 149,3
4 400 350 14 1 2 2695,22 991 158,38
. . . . . . . . .
. . . . . . . . .
45 550 500 14 2 6 5307,57 2040,23 450,22
46 600 300 14 4 1 3508,11 1374,17 294,71
47 600 300 18 3 2 3850,72 1368,77 362,88
48 600 350 16 3 1 4068,07 1596,43 346,04
. . . . . . . . .
. . . . . . . . .
91 700 350 16 7 2 5137,63 1934,52 529,61
92 700 350 18 7 1 5299,76 1951,72 537,78
93 700 350 16 7 3 5235,52 1931,19 559,51
94 700 350 24 3 1 5382,35 1909,13 587,06
. . . . . . . . .
. . . . . . . . .
156 800 450 24 4 5 8389,7 2814,16 1205,06
157 800 500 20 5 2 8006,54 3154,92 943,66
158 800 500 24 5 5 9272,95 3154,95 1299,31
159 850 350 14 7 1 5680,51 2371,26 664,1
. . . . . . . . .
. . . . . . . . .
215 1000 500 20 7 2 9969,93 4041,03 1450,92
216 1000 500 20 5 3 9816,98 3998,45 1473,01
217 1000 500 18 5 5 9774,16 3977,65 1504,87
218 1000 500 22 6 4 10508,3 4025,99 1677,74
219 1000 500 28 5 2 10985,5 4071,49 1749,45

Table 4: Design variables database for column sections.

9
 Figure 2: The design variables for construction of column section database.

2.3 Design Constraints

The design constraints considered in the optimum design problem are strength,
serviceability, ductility and other side constraints that are implemented from ACI
318-05.

2.3.1 Constraints for Column Groups

Two types of design constraints are considered for the column members of the RC
frame. The first type includes those constraints on the axial load and moment
resistance capacities of the section, clear spacing limits between reinforcing bars,
and the minimum and the maximum percentage of steel allowed. The second type
consists of those constraints defining architectural requirements and good design and
detailing practices. These include the requirement of the minimum and the
maximum dimensions of column, the maximum aspect ratio of the section,
maximum number of reinforcing bars and other reinforcement requirements. The
constraints are explained and expressed in a normalized form as given Table 5.

10
Constraints for Column Groups
The maximum axial load Pd (i , j )
capacity of columns, Pn,max g1 ( x) = −1 ≤ 0 i = 1, N col j = 1, 6 (10)
Pn,max (i , j )
should be greater than the
factored axial design load where, i = number of the column (ith column), Ncol = total
acting on the column section, number of columns, j = load combination type.
Pd ;
For a column section with M d (i , j )
uni-axial bending, the g 2 ( x) = −1 ≤ 0 i = 1, N col j = 1, 6 (11)
M n @ Pu ( i , j )
moment carrying capacity of
column section, Mn, obtained where, i = number of the column (ith column), Ncol = total
for each factored axial design number of columns, j = load combination type.
load, Pd, should be greater
than the applied factored
design moment, Md ;
The percentage of ρ min
g3 ( x) = − 1 ≤ 0 and
longitudinal reinforcement ρi
steel, ρ, in a column section ρ
should be between minimum g 4 ( x) = i − 1 ≤ 0 i = 1, N col (12)
and maximum limits ρ max
permitted by design where, i = number of the column (ith column), Ncol = total
specification (ρmin= 0.01 and number of columns.
ρmax = 0.06) ;
The width b and the height d cd min
of a column section should g5 ( x) = −1 ≤ 0
bi
not be less than the minimum
dimensions limit value given cd
g 6 ( x) = min − 1 ≤ 0 i = 1, N col (13)
for columns (min. dimension, di
cdmin = 300mm); where, i = number of the column (ith column), Ncol = total
number of columns.
The ratio of shorter cdrmin
dimension of column section g 7 ( x) = (b / d ) − 1 ≤ 0 i = 1, N col (14)
i i
to longer one should be th
greater than permitted limit where, i = number of the column (i column), Ncol = total
number of columns.
(cdrmin = 0.40);

The minimum diameter of φmin

longitudinal reinforcing bars, g8 ( x) = φ − 1 ≤ 0 i = 1, N col (15)
i
Ø, in a column section should th
be greater than minimum bar where, i = number of the column (i column), Ncol = total
number of columns.
diameter, Ømin, specified by
design code;
The total number of longitudinal nrbi
reinforcing bars, nrb, in a g9 ( x) = −1 ≤ 0 i = 1, N col (16)
column section should be nrbmax
smaller than specified maximum where, i = number of the column (ith column), Ncol = total
number of reinforcing bars, number of columns.
nrbmax, for detailing practice
(nrbmax = 24);

11
Constraints for Column Groups (Cont.)
The minimum and maximum amin
clear spacings between g10 ( x ) = a − 1 ≤ 0 and
i
longitudinal bars, a, in a
a
column section should be g ( x) = i − 1 ≤ 0
11 i = 1, N col (17)
between minimum and amax
maximum limits, amin and where, i = number of the column (ith column), Ncol = total
amax, specified for detailing number of columns.
practice (amin=50 mm and
amax=150 mm);
The shear force capacity of Vd ( i , j )
column section, ØVn, should g12 ( x) = −1 ≤ 0 i = 1, N col j = 1, 6 (18)
φVn ( i , j )
be greater than applied
where, i = number of the column (ith column), Ncol = total
factored design shear
number of columns, j = load combination type.
force,Vd ;
Also, the shear force capacity
of column section, ØVn, g13 ( x ) =
{
min Vec( k ) Veb( k ) } −1 ≤ 0 k = 1, N column group (19)
should be greater than the φVn ( k )
minimum capacity shear where, k = number of the column group (kth column group),
forces, min{Vec , Veb }, based Ncol = total number of column groups.
on probable maximum
flexural strength of column,
Vec , and based on probable
maximum flexural strengths
at the ends of beams framed
into the top joint of column,
V eb ;
The factored design shear Vd ( i , j )
force acting on column g14 ( x ) = −1 ≤ 0 i = 1, N col j = 1, 6 (20)
φVmax( i , j )
section, Vd, should be less
than allowed maximum shear where, i = number of the column (ith column), Ncol = total
force capacity, ØVmax ; number of columns, j = load combination type.

The area of shear Ash ,min

reinforcement (ties), Ash , g15 ( x) = −1 ≤ 0 k = 1, N column ∈group j = 1,3
Ash ( i , j )
should be greater than
limitations on the minimum ⎧ s s
⎪0.062 f c bd f ≥ 0.35d f
area of shear reinforcement, ⎪ y y

Ash,min ; ⎪ (21)
⎪ ⎛ Ag ⎞ f′
Ash ,min → max ⎨ 0.3 sbc ⎜ − 1⎟ c
⎪ A
⎝ ch ⎠ fy
⎪ f
⎪ 0.09 s bc c
⎪⎩ f y

where, k = number of the column group (kth column group),

Ncolumn group = total number of column groups, j = three (top,
middle and bottom parts) shear design region of column , fc =
the compressive strength of concrete, fy = the yielding strength
of reinforcing steel, b and d = the width and height of column
section, s = the spacing between stirrups (ties), bc = the cross-
sectional dimension of column core measured center-to-center
of outer legs of the transverse reinforcement comprising area

12
Constraints for Column Groups (Cont.)
Ash , Ag = the gross area of section, Ach = the cross-sectional
area of member measured out-to-out of transverse
reinforcement.
The spacing between stirrups, smin
s, in the column should be g16 ( x) = s − 1 ≤ 0 k = 1, N column group j = 1,3 (22)
(i , j )
greater than minimum
where, k = number of the column group (kth column group),
spacing, smin (smin = 50 mm) N
column group = total number of column groups, j = three (top,
for constructional middle and bottom parts) shear design region of the column.
requirements;

At the top and bottom ends of s(i , j )

column members, the spacing g17 ( x) = −1 ≤ 0 k = 1, N column group j = 1, 2
smax,end
between stirrups, s, should be
less than maximum spacing ⎧ d
of shear reinforcement for ⎪ 4
end regions of column, ⎪
⎪ b (23)
smax,end ; smax,end → min ⎨
⎪ 4
⎪6φb
⎪
⎩
where, k = number of the column group (kth column group),
Ncolumn group = total number of column groups, j = two (top and
middle parts) shear design region of the column, b and d = the
width and height of column section, s = the spacing between
stirrups (ties), Øb = the diameter of longitudinal reinforcing
bars.
For the middle parts column s( i , middle )
members, the spacing g18 ( x ) = −1 ≤ 0 k = 1, N column group
between stirrups, s, should be smax, middle
less than maximum spacing ⎧ 6φ
of shear reinforcement, smax,middle → min ⎨ b (24)
smax,middle ; ⎩15 cm
where, k = number of the column group (kth column group),
Ncolumn group = total number of column groups, Øb = the diameter
of longitudinal reinforcing bars.
The length of top and bottom l0,min
shear regions, lo, should be g19 ( x ) = −1 ≤ 0 k = 1, N column group j = 1, 2
l0( i , j )
greater than allowable design
length, lo,min ; ⎧ larger of b and d
⎪ (25)
l0,min → max ⎨1/ 6 clear span of column
⎪ 45 cm
⎩
where, k = number of the column group (kth column group),
Ncolumn group = total number of column groups, j = two (top and
middle parts) shear design region of column, b and d = the
width and height of column section.

Table 5: Constraints are considered for the column members of the RC frame.

13
2.3.2 Constraints for Beam Groups

Constraints to be imposed for each beam group are based on strength, serviceability,
ductility and other side constraints. The normalized forms of all constraints
considered in optimum design problem are given Table 6.

Constraints for Beam Groups

For three critical sections M d ( i , j , k ,l )
(left end, middle part and g 20 ( x) = − 1 ≤ 0 i = 1, N beam j = 1,.., 6 k = 1,3 l = 1, 2
M n ( i , j , k ,l )
right end) of each beam, the
(26)
negative (with top steel in
tension) and positive (with where, i = number of the beam (ith beam), Nbeam = total number
bottom steel in tension) of beams, j = load combination type, k = three critical sections
moment carrying capacities for flexural design of beam, l= the negative moment and
of section for Mn, should be positive moment situations (top steel in tension or bottom steel
greater than the applied in tension).
factored design moments Md
The tension area of As ,min(i ,k ,l )
longitudinal reinforcement g 21 ( x) = −1 ≤ 0 i = 1, N beam k = 1,3 l = 1, 2
As (i ,k ,l )
steel bars in tension As, for
three critical sections in a
beam should satisfy the As (i ,k ,l )
g 22 ( x) = −1 ≤ 0 i = 1, N beam k = 1,3 l = 1, 2
minimum and the maximum As ,max( i ,k ,l )
requirements permitted by
design specification; ⎧ f′
⎪0.25 c bw h
⎪ fy
As ,min → max ⎨
⎪ 1.4 bw h
⎪ fy
⎩ (27)

As ,max = 0.025bw h (for the total area of top and bottom flexural bars )
where, i = number of the beam (ith beam), Nbeam = total number
of beams, k = three critical sections for flexural design of beam,
l= the negative moment and positive moment situations
(checking for top and bottom reinforcement steel area), bw and
h = the width and height of beam section, fc = the compressive
strength of concrete, fy = the yielding strength of reinforcing
steel.
At any end (support) of the 1 −
beams, the positive moment M n ,( i ,k )
- g 23 ( x ) = 2 + −1 ≤ 0 i = 1, N beam k = 1, 2 (28)
capacity Mn , (i.e., associated M n ,( i , k )
with the bottom steel) should
be greater than 1/2 of the
beam negative moment where, i = number of the beam (i.th beam), Nbeam = total
capacity Mn+, (i.e., associated number of beams, k = the left and right ends of beam.
with the top steel) at that end;
The positive flexural moment 1 −
strength, Mn,middle+, at span of M n ,( i , k )
g 24 ( x ) = +4 −1 ≤ 0 and
beams should be greater than M n ,( i , middle )
a quarter of negative and

14
Constraints for Beam Groups (Cont.)
positive flexural moment 1 +
M n ,( i ,k )
strengths at the ends of
g 25 ( x ) = 4 + −1 ≤ 0 i = 1, N beam k = 1, 2 (29)
beams, Mn,- and Mn+ ; M n ,( i ,middle )
where, i = number of the beam (i.th beam), Nbeam = total
number of beams, k = the left and right ends of beam.
The shear force capacity of Vd ( i , j ,k )
three regions (left and right g 26 ( x ) = −1 ≤ 0 i = 1, N beam j = 1, 6 k = 1,3
φVn (i , j ,k )
ends, and span) of beam,
ØVn, should be greater than (30)
applied factored design shear where, i = number of the beam (i.th beam), Nbeam = total
force, Vd ; number of beams, j = load combination type, k = three critical
region (left and right ends, and span) for shear design of beam.
Also, the shear force capacity
of design regions in a beam, g 27 ( x ) =
{
max Ve1( i ) Ve1( i ) } −1 ≤ 0 i = 1, N beam k = 1,3
ØVn, should be greater than φVn ( i ,k )
the probable shear forces (31)
based on probable maximum
flexural strengths with the
factored gravity loads at where, i = number of the beam (i.th beam), Nbeam = total
number of beams, k = three critical region (left and right ends,
beam ends of beam, max{Ve1
and span) for shear design of beam.
, Ve2 } ;
The factored design shear Vd ( i ,k )
forces at the middle and ends g 28 ( x) = −1 ≤ 0 i = 1, N beam k = 1,3 (32)
of beam, Vd, should be less
φVmax( i ,k )
than allowed maximum shear where, i = number of the beam (i.th beam), Nbeam = total
force capacity, ØVmax ; number of beams, k = three critical region (left and right ends,
and span) for shear design of beam.
The spacing between stirrups smin
s, in the middle and the ends g 29 ( x) = −1 ≤ 0 k = 1, N beam j = 1,3 (33)
s( i ,k )
of the beam should be greater
than the minimum spacing, where, i = number of the beam (i.th beam), Nbeam = total
smin=50mm for constructional number of beams, k = three critical region (left and right ends,
and span) for shear design of beam.
requirements;
At the left and right ends of s( i , j )
beam members, the spacing g30 ( x) = −1 ≤ 0 i = 1, N beam k = 1, 2
smax,end
between stirrups, s, should be
less than maximum spacing ⎧ h
limit of shear reinforcement ⎪ 4
⎪ (34)
for end regions of beam smax,end → min ⎨30 cm
smax,end ; ⎪ 24φ
⎪ b

⎩
where, i = number of the beam (ith beam), Nbeam = total number
of beams, k = three critical regions (left and right ends and the
span) for shear design of beam, h = the height of beam section,
Øb = the diameter of shear reinforcing bars (ties) .

15
Constraints for Beam Groups (Cont.)
Along the span of a beam s( i ,middle )
member, the spacing between g 31 ( x) = −1 ≤ 0 i = 1, N beam
stirrups s should be less smax,middle
than maximum spacing of the ⎧ h
shear reinforcement, smax,middle ⎪
⎪⎪ 2
(35)
smax,middle → min ⎨ 60 cm
⎪ A
⎪30 sv bw
⎪⎩ fy
where, i = number of the beam (ith beam), Nbeam = total number
of beams, h = the height of beam section, bw and h = the width
and height of column section, Asv = the area of shear
reinforcement (tie), fy = the yielding strength of reinforcing
steel.
The area of shear Asv ,min
reinforcement (ties) in beam g 32 ( x ) = −1 ≤ 0 i = 1, N beam k = 1, 3
sections (span and ends of
A sv ( i , k )

beam), Asv, should be greater s s

than limitations on the
Asv ,min = 0.062 f c bw h ≥ 0.35h (36)
fy fy
minimum area of shear
where, i = number of the beam (ith beam), Nbeam = total number
reinforcement, Asv,min ;
of beams, k = three critical regions (left and right ends, and
along the span) for shear design of a beam, fc = the
compressive strength of concrete, fy = the yielding strength of
reinforcing steel, bw and h = the width and height of beam
section, s = the spacing between stirrups (ties) .
The width of beams, bw, bw,min
should be greater than g33 ( x) = −1 ≤ 0 i = 1, Nbeam
allowable minimum width bw(i )
for beams, bw,min ; ⎧ 25 cm
⎪
bw,min → max ⎨ Lbeam ( i ) (37)
⎪
⎩ 50
where, i = number of the beam (ith beam), Nbeam = total number
of beams, Lbeam = the length of beam.
The height of beams, h, hmin
should be greater than g34 ( x) = −1 ≤ 0 i = 1, Nbeam
allowable minimum height
h(i )
for beams, h,min ; ⎧ Lbeam ( i )
⎪⎪ for one end continuous
18.5 (38)
hmin → ⎨
L
⎪ beam ( i ) for both end continuous
⎪⎩ 21
where, i = number of the beam (ith beam), Nbeam = total number
of beams, Lbeam = length of the beam.

Table 6: Constraints are considered for the beam members of the RC frame.

16
2.3.3 Constraints for Joints

Constraints for Joints

At frame joints, the width of bw(i )
beams, bw, should be smaller g35 ( x) = −1 ≤ 0 i = 1, N joint (39)
than the width of column, b, b(i )
framed into the ends joint of where, i = number of the joint (ith joint), Njoint = total number of
beam; joints.
The width of the top column, btop column(i )
b, should be equal or smaller g36 ( x) = −1 ≤ 0 i = 1, N joint (40)
than the width of the bottom bbottom column(i )
column at the column joints; where, i = number of the joint (ith joint), Njoint = total number of
joints.
The height of the top column, htop column(i )
h, should be equal or smaller g37 ( x) = −1 ≤ 0 i = 1, N joint (41)
than the width of the bottom hbottom column(i )
column at the column joints; where, i = number of the joint (ith joint), Njoint = total number of
joints.
The sum of nominal flexural 6
strengths of columns framing 5
∑ M nb(i)
into the joint, ΣMnc, should g ( x ) = −1 ≤ 0 i = 1, N joint (42)
∑
38
M nc (i )
be 1.2 times greater than the th
sum of nominal flexural where, i = number of the joint (i joint), Njoint = total number of
strengths of beams framing joints.
into same joint, ΣMnb,
Δi, the relative displacements ⎛ Δi ⎞
of stories, should be smaller ⎜ ⎟
than the allowable story drift, g ( x) = ⎝ hi ⎠ − 1 ≤ 0 i = 1, N story (43)
39
Δa, given in the design Δa
specification; where, i = number of the story (ith story), Nstory = total number
of stories, hi = the height of story from base level.

Table 7: Constraints are considered for the RC frame joints.

3 Harmony Search Method

The Harmony Search (HS) method is a meta-heuristic optimization algorithm, which
is proposed by Geem, Kim, and Loganathan [15]. The HS algorithm is inspired by
the underlying principles of the musicians’ improvisation of the perfect state of
harmony. The HS algorithm mimics the improvisation of music players for
searching a better harmony. The various possible combinations of the musical
pitches stored in the musicians’ memory, when they compose harmony. And the
musicians can find the fantastic harmony from their harmony memories [16,17].
In the adaptation of this musical improvisation process into the solution of
optimization problems, each musician corresponds to a decision variable and
possible notes correspond to the possible values of decision variables. The players
play any pitch within the possible range, and these sounds make one harmony vector

17
in music improvisation. These harmony vectors are stored in each player’s memory
if all the pitches make a good harmony, and the possibility to make a better harmony
is increased next time. Similarly in optimization problems, each decision variable
initially chooses any value within the possible range, and these values of decision
variables make one solution vector. If all the values of decision variables make a
good solution, these solution vectors are stored in the memory. Thus, the possibility
to make a good solution is also increased next time. The optimization procedure of
the Harmony Search meta-heuristic algorithm consists of Steps 1-5, as follows;

Step 1- The discrete optimization problem is specified as follows:

Minimize f(x) s.t. xi ∈ Xi , i= 1,2,…,N (44)

where; Xi is the set of possible candidate values for each decision variable i. A
design variable pool is constructed for each design variable. The number of solution
vectors in harmony memory matrix (HMS), harmony memory considering rate
(HMCR), pitch adjusting rate (PAR), and termination criterion (maximum number
of search) are also decided in this step.

Step 2- The ‘‘harmony memory’’ (HM) matrix is filled with randomly generated
solution vectors using the values in the design pool and sorted by the values of the
objective function, f(x). The Harmony Memory is filled with as many solution
vectors as harmony memory size (HMS). The harmony memory matrix has the
following form:

⎡ x11 x12 .. .. x1N −1 x1N ⎤

⎢ 2 ⎥
⎢ x1 x22 .. .. xN2 −1 xN2 ⎥
⎢ .. .. .. .. .. .. ⎥
[H ] = ⎢ ⎥
⎢ .. .. .. .. .. .. ⎥ (45)
⎢ x HMS −1 x2HMS −1 .. HMS −1
.. xN −1 xNHMS −1 ⎥
⎢ 1 HMS ⎥
⎣⎢ x1 x2HMS .. .. xNHMS
−1 xNHMS ⎦⎥

Each row of harmony memory matrix contains the values of design variables of
the optimization problem and presents one feasible solution vector. The design
variables are randomly selected from the design variable pool for that particular
design variable. Hence, this memory matrix has N columns (N; the total number of
design variables) and HMS rows (HMS; harmony memory size). These candidate
solutions are sorted such that the objective function value corresponding to the first
solution vector (first row) is the minimum or maximum.
Step 3- A new harmony vector, x ı = ( x1ı, x2ı, x3ı,…. xNı) is improvised. There are
three rules to choose one value for each decision variable; harmony memory
consideration rate (HMCR), pitch adjustment rate (PAR), and random selection.

18
 In memory consideration, the value of the first decision variable (x1ı) can be
chosen from any discrete value in the specified HM range { x11, x12, x13,…., x1HMS-1,
x1HMS } with the probability of HMCR which varies between 0 and 1. Values of the
other decision variables (xi ı) can be chosen in the same manner. However, there is
still a chance where the new value can be randomly chosen from a set of entire
possible values in the design variable pools with the probability of (1-HMCR). That
is
⎧ x ′ ∈ { x 1 , x 2 ,........, x HMS } with probability HMCR
′ ⎪ i i i i
xi ← ⎨
⎪⎩ xi′ ∈ X i with probability (1 − HMCR)

Any component of the new harmony vector, whose value was chosen from the
HM, is then examined to determine whether it should be pitch-adjusted. This
operation uses pitch adjusting parameter (PAR) that sets the rate of pitch-adjustment
decision as follows:
⎧ Yes with probability PAR
Pitch adjusting decision for xi′ ← ⎨
⎩ No with probability (1 − PAR )

If the pitch adjustment decision for xi ı is Yes , xi ı is replaced with xi(k) (the kth
element Xi ) , and the pitch-adjusted value of xi(k) becomes:

xi ı ← xi ( k ± 1)

The algorithm chooses - 1 or 1 for the neighbouring index m with the same
probability.
Step 4- After selecting the new values for each design variable the objective
function value is calculated for the new feasible harmony vector. If the new
Harmony vector is better than the worst harmony in the HM in terms of the
objective function value, the new harmony is included in the HM and the existing
worst harmony is excluded from the HM. The HM is then sorted by the objective
function value.
Step 5- The computations are terminated when the termination criterion is
satisfied. If not, Steps 3 and 4 are repeated. The termination criterion is selected
large enough such that within this number of cycles no further improvement is
observed in the objective function value.

4 Design Example
Two-bay, six-storey RC plane frame is designed by the algorithm presented in order
to demonstrate its efficiency. The design examples considered are solved several
times using with different set of harmony search parameters and among the optimum
frames obtained for each set, the best one is taken as the optimum design.

19
4.1 Two-Bay Six-Story RC Frame
A two-bay six-story RC frame given in Figure 3 is designed by Rajeev and
Krishnamoorthy [7], Camp and Pezeshk [8], Govindaraj and Ramasamy [12]. This
frame has 12 beams and 18 columns, which are collected in five design groups: two
groups for beams and three groups for columns. The compressive strength of
concrete and yield strength of steel are 20MPa and 415MPa, respectively. The frame
is loaded with the unfactored uniformly distributed load of 30kN/m and the lateral
force of 10kN is applied at each story level. The load combinations are not
considered in this practical optimum design problem to be able compare the results
with the ones given in [7]. The unit costs of concrete, steel and formwork are 735
Rs./m3, 7.1 Rs./kg and 54 Rs./m2, respectively as given in [7]. The unit weight of
concrete and steel is taken as 25kN/m3 and 78.5kN/m3, respectively. The input data
of the frame is given in Table 8. In this design example, there are 29 design
variables, 26 of which are beams (13 for each beam design group), the remaining 3
is for columns. The Harmony search method parameters HMS, HMCR, PAR are
selected as 45, 0.80, and 0.15, respectively. The cost of optimum design obtained as
43586.19Rs. The optimum values for design variables obtained are given in Table 9.
The harmony search algorithm design history is given in Figure 4. For this
optimization problem, the minimum costs of optimum designs are obtained as
26052Rs., 24959Rs., and 22966Rs. by Rajeev [7], Camp [8], Govindaraj [12],
respectively. In these studies, neither shear design constraints nor the requirements
of Special Seismic Moment Frames are considered. Furthermore the development
and hook lengths of reinforcement steel bars in the concrete members and the other
detailing issues are not considered in the calculation of the frame cost. Therefore, the
differences between obtained the optimum costs is based on different of design
philosophies which exist between this study and the ones available literature.

Input datas for two-bay six-storey frame

Compressive strength of concrete, fc (N/mm2) 20
Yielding stress of steel, fy (N/mm2) 415
Cover for beams, mm 25
Cover for columns, mm 25
Total number of beam groups 2
Total number of beams 12
Total number of column groups 3
Total number of columns 18
Lateral force for each story (kN) 10
Factored uniformly distributed load on beams (kN/m) 30
Harmony Search Algorithm Parameters
HMS 45
HMCR 0.80
PAR 0.15
Max. Iteration 100,000

Table 8: Input data for two-bay six-story RC frame example.

20
 g = 30 kN/m g = 30 kN/m

Beam group 1

Column gr. 3
Column gr. 1

Column gr. 2

4,00 m
g = 30 kN/m g = 30 kN/m

Beam group 2

Column gr. 3
Column gr. 1

Column gr. 2

4,00 m
g = 30 kN/m g = 30 kN/m

Beam group 2
Column gr. 1

Column gr. 2

Column gr. 3

4,00 m
g = 30 kN/m g = 30 kN/m

Beam group 2
Column gr. 2

Column gr. 3
Column gr. 1

4,00 m
g = 30 kN/m g = 30 kN/m

Beam group 2
Column gr. 2

Column gr. 3
Column gr. 1

4,00 m

g = 30 kN/m g = 30 kN/m

Beam group 2
Column gr. 2

Column gr. 3
Column gr. 1

4,00 m

6,00 4,00

Figure 3: Two-bay six-storey RC plane frame.

21
 Rs. 100x103
50000

Rs. 97x103

Rs. 94x103 49000

Rs. 91x103
48000

Rs. 88x103

Rs. 85x103 47000

Rs. 82x103
46000
Rs. 79x103

Rs. 76x103 45000

COST, Rs.

Rs. 73x103
44000
Rs. 70x103

Rs. 67x103
43000
20x103 30x103 40x103 50x103 60x103
Rs. 64x103

Rs. 61x103 Minimum Cost in Harmony Memory

Average Cost in Harmony Memory
Rs. 58x103 Maximum Cost in Harmony Memory

Rs. 55x103

Rs. 52x103

Rs. 49x103

Rs. 46x103

Rs. 43x103
0 20x103 40x103 60x103 80x103 100x103

Iteration Number

Figure 4: The design history for two-bay six-storey RC plane frame (load
combinations are not considered.

5 Conclusions
The optimum design problem of reinforced concrete special seismic moment frames
is formulated according to the provisions of ACI 318-05. In addition to the design
constraints specified in the design code architectural and reinforcement detailing
constraints are also considered in the design problem. It is shown that the harmony

22
 Optimum design results for two-bay six-storey frame
Optimum column design results
Column Column Column
Design Variables
Gr. 1 Gr. 2 Gr. 3
Section number in design pool (3) (82) (3)
The height of section, h (mm) 400 500 400
The height of section, b (mm) 350 300 350
The diameter of reinforcement bars, φ (mm) 14 14 14
The number of bars in x direction, n1 2 3 2
The number of bars in y direction, n2 1 2 1
Total reinforcement 10φ14 14φ14 10φ14
Optimum beam design results
Design Variables Beam Gr. 1 Beam Gr. 2
x1 25/40 (1) 25/40 (1)
x2 2φ12 (1) 2φ12 (1)
x3 2φ12 (1) 2φ12 (1)
x4 1φ16 (4) 1φ16 (4)
x5 1φ16 (4) 1φ14 (3)
x6 1φ12 (2) 1φ12 (2)
x7 1φ12 (2) 1φ12 (2)
x8 1φ14 (3) 2φ14 (7)
x9 NNR* (1) NNR* (1)
x10 1φ14 (3) 1φ16 (4)
x11 NNR* (1) NNR* (1)
x12 NNR* (1) NNR* (1)
*
x13 NNR (1) NNR* (1)
* th
NNR , no need reinforcement, ( ni ), the sequence number of i design variable in the design
variable pool.

Table 9: The optimum values of design variables for two-bay six-story RC frame.

search algorithm can efficiently be used to determine the solution of this optimum
design problem. It is noticed that three main parameters of the harmony search
algorithm namely, harmony memory size HMS, harmony memory considering rate
HMCR, and the pitch adjustment rate PAR play an important role on the optimum
designs obtained. It is also noticed that in the design example considered, the
harmony search algorithm parameters, the high HMCR, especially from 0.70 to
0.95, the small PAR, generally from 0.15 to 0.20, and HMS, from 40 to 60, yielded
good performance in the design optimization. Furthermore the unit costs of concrete,
steel, and formwork play an important role in determining of the optimum column
and beam section dimensions. The results obtained from the design examples are
shown that harmony search method is a reliable and robust technique that can be
effectively used in finding the optimum detailed design of reinforced concrete
frames. The optimum design algorithm developed arrives at rational and realistic
design solutions that are directly constructible. It is also shown that optimum design
opt reinforced concrete frame without considering special seismic constraints yields
unsafe design.

References
[1] X-S. Yang, “Nature-Inspired metaheuristic Algorithms”, Luniver Press, 2008.

23
[2] X-S. Yang, “Engineering Optimization: An Introduction with Metaheuristic
Applications”, John Wiley, 2010.
[3] M. P. Saka, “Optimum Design of Skeletal Structures: A Review”, Chapter 10
in Progress in Civil and Structural Engineering Computing, Ed. B. H. V.
Topping, 237-284, Saxe-Coburg Publications, 2003.
[4] M. P. Saka, “Optimum Design of Steel Frames using Stochastic Search
Techniques Based on Natural Phenoma: A Review”, Chapter 6 in Civil
Engineering Computations: Tools and Techniques, Ed: B. H. V. Topping, 105-
147, Saxe-Coburg Publications, 2007.
[5] C. Choi, H. Kwak, “Optimum RC member design with predetermined discrete
sections”, The Journal of Structural Engineering ASCE, Vol. 116(10), 2634-
2655, 1990.
[6] R.J. Balling, X. Yao, “Optimization of reinforced concrete frames”, The
Journal of Structural Engineering ASCE, Vol. 123(2), 193-202, 1997.
[7] S. Rajeev, C.S. Krishnamoorthy, “Genetic algorithm-based methodology for
design optimization of reinforced concrete frames”, Computer-Aided Civil
Infrastructral Engineering, Vol. 13(1), 63–74, 1998.
[8] C.V. Camp, S. Pezeshk, H. Hansson, “Flexural design of reinforced concrete
frames using a genetic algorithm”, The Journal of Structural Engineering
ASCE, Vol. 129(1), 105–115, 2003.
[9] C. Lee, J. Ahn, “Flexural design of reinforced concrete frames by genetic
algorithm”, The Journal of Structural Engineering ASCE, Vol. 129(6), 762–
774, 2003.
[10] H-G. Kwak, J. Kim, “Optimum design of reinforced concrete frames based on
predetermined section database”, Computer-Aided Design, Vol. 40(3), 396-
408, 2008.
[11] H-G. Kwak, J. Kim, “An integrated genetic algorithm complemented with
direct search for optimum design of RC frames”, Computer-Aided Design,
Vol. 41(7), 490-500, 2009.
[12] V. Govindaraj, J.V. Ramasamy, “Optimum detailed design of reinforced
concrete frames using genetic algorithms”, Engineering Optimization, Vol.
39(4), 471-494, 2007.
[13] ACI 318-05 “Building code requirements for structural concrete and
commentary”, ACI Committee 318, Structural Building Code, American
Concrete Institute, Farmington Hills, MI, USA, 2005.
[14] ASCE 7-05 “Minimum design loads for buildings and other structures”,
American Society of Civil Engineers, Virginia, USA, 2005.
[15] Z.W. Geem, J-H. Kim, G.V. Loganathan, “A new heuristic optimization
algorithm: Harmony search”, Simulation, Vol. 76(2), 60–68, 2001.
[16] K.S. Lee, Z.W. Geem, “A new structural optimization method based on the
harmony search algorithm”, Computers & Structures, Vol. 82(9-10), 781–798,
2004.
[17] K.S. Lee, Z.W. Geem, S.H. Lee, , K.W. Bae, “The harmony search algorithm
for discrete structural optimization”, Engineering Optimization, Vol. 37(7),
663–684, 2005.

24