Scientia Iranica A (2018) 25(1), 162{173

Sharif University of Technology

Scientia Iranica
Transactions A: Civil Engineering
http://scientiairanica.sharif.edu

Optimum design of castellated beams: Eects of

composite action and semi-rigid connections
A. Kaveha; and M.H. Ghafarib
a. Centre of Excellence for Fundamental Studies in Structural Engineering, Iran University of Science and Technology, Narmak,
Tehran, P.O. Box 16846-13114, Iran.
b. School of Civil Engineering, Iran University of Science and Technology, Narmak, Tehran, P.O. Box 16846-13114, Iran.
Received 15 May 2016; received in revised form 4 October 2016; accepted 10 December 2016

KEYWORDS Abstract. Castellated beams and composite action of beams are widely applicable
Structural methods to increase the capacity of the beams. Semi-rigid connections can also redistribute
optimization; internal moments in order to attain a better distribution. Combination of these methods
Semi-rigid helps to optimize the cost of the beam. In this study, some meta-heuristic algorithms
connections; consisting of the particle swarm optimization, colliding bodies optimization, and enhanced
End lled castellated colliding bodies optimization are used for optimization of semi-rigid jointed composite
beams; castellated beams. Prole section, cutting depth, cutting angle, holes spacing, number of
Composite beams; lled end holes of the castellated beams, and rigidity of connection are considered as the
Colliding Bodies optimization variables. Constraints include the construction, moment, shear, de ection,
Optimization (CBO); and vibration limitations. Eects of partial xity and commercial cutting shape of a
Enhanced Colliding castellated beam for a practical range of beam spans and loading types are studied through
Bodies Optimization three numerical examples. The eciency of three meta-heuristic algorithms is compared.
(ECBO);
Particle Swarm
Optimization (PSO). © 2018 Sharif University of Technology. All rights reserved.

1. Introduction that limit their usage. For example, in order to face the
secondary shear eect of the castellated beams, some
Optimizing the structural elements to achieve the best end hole should be lled.
economical and serviceable result is one of the most One of the rst studies that evaluated the eect
desirable aims of the structural designers. A typical of web opening on composite beams was performed by
example of such a trend is the eorts of researchers Redwood and Cho [1]. They also surveyed the failure
to eciently increase the capacity of beams. Castel- modes of the hexagonal composite castellated beams.
lated beams and composite beams can increase beams Several tests were performed by Jackson [2] showing
moment of inertia and this eect increases the bending that the AISC design guide procedures for composite
moment capacity of the beams. However, both of them prismatic beams could be used for calculating the
have some limitations and produce secondary eects natural frequency of the beams.
The eect of edge constraint component on ex-
*. Corresponding author. Tel.: +98 21 44202710; ural strength of composite beam was studied by Wang
Fax: +98 21 77240398 and Li [3]. Ellakani and Tablia [4] developed a
E-mail address: alikaveh@iust.ac.ir (A. Kaveh) numerical model for static and free vibration analysis
of elastic composite beams with end shear restraint.
doi: 10.24200/sci.2017.4195 They found that the end shear restraint played an
 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173 163

important role when the composite beam interaction 20% of the total cost of an optimized steel frame
was not complete. structure.
There have been several eorts for cost optimiza- The main objective of the present paper is to
tion of composite beam. Morton and Webber [5] study the composite action, semi rigidity of joint, and
optimized the composite I beams. They found that end holes lling on optimization of the castellated
the xed-end conditions led to lighter beams. Senouci beams. The optimization algorithms used in this paper
and Al-Ansari [6] used the genetic algorithm for opti- consist of the colliding body optimization, enhanced
mizations. colliding body optimization, and particle swarm op-
Optimization of castellated beams with several timization. Optimization variables are prole section,
variables has been performed by many researchers. cutting depth (dh ), cutting angle (), holes spacing (s),
Sorkhabi et al. [7] optimized the castellated beams number of lled end holes of the castellated beam, and
by PSO and Genetic Algorithm (GA) and found that rigidity of the connections (Rj ).
PSO was better than GA. Kaveh and Shokohi [8,9] The present paper is organized as follows: In the
optimized castellated beam with hexagonal and cellular next section, design of the castellated beams is intro-
opening by CBO and Charged System Search (CSS) duced. In Section 3, design of the composite beams is
meta-heuristic algorithms. They also considered the provided. Semi-rigid connections are introduced brie y
eect of end lled plates [10]. in Section 4. Some dependence features of semi rigidity,
Semi-rigid composite connections have been mod- composite action, and castellated beams are discussed
eled and tested by some researchers in recent years. in Section 5. Three optimization algorithms that are
Modeling of beam to girder semi-rigid composite used in this study are brie y introduced in Section 6.
connection with angles, including the eects of con- Based on these sections, problem denition is presented
crete tension stiness, was studied by Oliveira and in Section 7. In Section 8, some examples are optimized
Batista [11]. They calibrated a modied theoretical using dierent metaheuristics, and nally Section 9
model based on the component method and found a concludes the paper.
good agreement between the experimental and theo-
retical results. They observed the role of the concrete 2. Design of castellated beams
before and after crack development and stabilization
and the shear lag eect in the slab. Fu et al. [12] mod- Increasing beam depth without using additional ma-
eled a 3D nite element model of semi-rigid composite terial by cuttings and welding beams is known as the
end-plated connections. They presented the eects of castellated beam. This method produces a beam web
longitudinal bar, thickness of the endplate, thickness of and shape of the holes may be dierent depending on
beam ange, etc. on capacity of the connection. Gil et the cutting procedures. Hexagon and circle are two
al. [13] validated the design procedures experimentally well-known and common shapes of the cutting. To
and numerically. They found that the load level in the avoid the stress concentration at angles of hexagonal
minor column axis had no in uence on the behavior of shape, sinusoid cutting can be performed instead of
the major one. Rex and Easterling [14] developed a hexagonal openings. Also, for both shapes of holes,
simple method of approximating the moment-rotation secondary plates can be placed between two parts of
behavior of composite beam-girder connection. They the beam and increased depth of the beam. In the
presented the eect of pre-loading on the moment present study, simple hexagonal shape of castellated
rotation behavior. beam is considered, because this results in better design
Redistribution of internal member forces by semi- in optimization [8-10].
rigid connections can help designers to decrease the Determining the strength of the castellated beams
total cost of the building. Simoes [15] was one of is more complex than that of the standard beams.
the rst researchers who optimized frame elements by Interaction between shear and moment, horizontal
semi-rigid connections. He used segmental method shear and radial moment are the main dierences
utilizing linear programming to solve optimization between typical beam design process and castellated
problems. Kameshki and Saka [16] utilized GA to beam design procedure.
optimize non-linear steel frame with semi-rigid con- Unbraced length of the beams plays an important
nections. They considered P
eect in their role in their exural capacity. By considering the
analysis. Ramires et al. [17] optimized the composite concrete on the top of the beam, unbraced length can
and steel endplate semi-rigid joints by GA. They be considered as zero and lateral-torsional buckling can
found that the joint cost could be decreased by 10% be assumed not to occur.
by tuning the stiness of the connections. Ali et In the following, the design procedure of the
al. [18] used the multi-stage design optimization by GA castellated beams is provided based on Load and
for semi-rigid steel frames, and their results showed Resistance Factor Design (LRFD) method of the AISC
that the cost of joints might constitute more than 360-10 [19].
164 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173

Figure 1. Details of a composite castellated beam.

2.1. Flexural capacity must satisfy the following equations in AISC 360-
Section plastic modulus of a beam identies the mo- 10 [19]:
ment capacity of the beam and the maximum moment
under applied load should satisfy the following condi- AW = ds  tw ; (4)
tion in AISC 360-10 [19]:
Vu <  v Vn w = v  0:6Fy AW Cv ; (5)
Mu < b Mn = b Fy Znet - st ; (1)
where Aw is the area of the net section web; tw is
where Mu is the ultimate moment of beam; Mn is the the thickness of the web; ds is height of the internal
nominal moment capacity of the beam; Zn st is the castellated beam; Vu is the ultimate shear force; Vn w
plastic module of steel net section; b is the bending is the nominal web shear capacity of net section; v
reduction factor; and Fy is the yield strength of steel. is the shear reduction factor; and Cv is the web shear
This equation is related to the general beam coecient.
section. At the holes, Vierendeel mechanism identies Second, vertical shear capacity of the tee beams
the exural demand of the beam. In Vierendeel must satisfy the following equations:
mechanism, two virtual hinges at top and bottom tee
beams between two holes are considered. Vertical Atee = dt  tw ; (6)
shear forces of these two tee beams produce secondary
moment that is added to essential moment. Because of Vu
< v Vn tee = v  0:6Fy Atee Cv ; (7)
linear distribution of moment at depth of the beam, at 2
points \a" and \c" depicted in Figure 1, sum of the two where Atee is the area of each tee section and Vn tee is
moments may be critical. Therefore, exural capacity the nominal web shear capacity of tee section.
of the tee beam should satisfy the following equation Third, horizontal shear capacity of post web must
in AISC 360-10 [19]: satisfy the following equations:
V e Ahe = e  tw ;
mu = u ; (2) (8)
4
Mu m Vu  Qg
+ u < b Fy ; (3) Vh =
Ig
 s < v Vn p = v  0:6Fy Ahe Cv ; (9)
Znet - st Ztee
where mu is the ultimate moment of the secondary where Vh is the horizontal shear at post web; Qg and Ig
shear; Vu is the ultimate shear force; e is the web post are the rst and second moments of inertia of general
length; Mu is the ultimate moment; Znet-st is the plastic section, respectively; s is the spacing between the holes
module of steel net section; and Ztee is the plastic (Figure 1); Vn p is the nominal shear capacity of post
module of steel tee section. b for concrete and steel is web; Ahe is web post horizontal shear area; and v and
considered to be equal to 0.9. Cv are equal to 1.
By lling the hole, Vierendeel mechanism can be Because of the penetration of web welding, weld-
neglected. ing capacity is not critical at horizontal shear check.
By lling the hole, horizontal shear can be controlled.
2.2. Shear capacity
Three types of shear forces must be controlled in 2.3. Web post buckling
castellated beam. Due to the horizontal shear, which was discussed above,
First, overall shear capacity of the general section web post plate performs similar to a cantilevered beam
 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173 165

without ange, and at one side of it, compression forces Shear studs are designed for transforming shear
produce instability by out of plane buckling [20]. The forces between steel and concrete completely. Shear
radial moment and capacity of the web post, according and strength of the steel channel anchors, without
to Structural Stability Research Council (SSRC), must considering compression steel eect, are determined
satisfy the following equations: according to AISC 360-10 as follows [19]:
M

M 2

Qu = min(0:85Fc be hc ; As Fy ) < Nc v Qn = Nc  v
Cb = 1:75 + 1:05 1 + 0:3 1 < 2:3;
M2 M2 p
 0:3(ts f + 0:5ts w )La Fc Ec ; (11)
t
rT = pw ; where Fc and Ec are compression strength and elastic-
12
ity module of concrete, respectively; be and hc are the
22 E eective width and height of concrete, respectively; As
Cc = ; is the steel section area; ts f , ts w , and La are ange
Fy
thickness, web thickness, and width of the channel
Lb = 2dh ; shear studs, respectively. Considering linear shear
2
diagram, Nc is half of the total number of shear studs
 2 3
Lb and v is equal to 0.75.
3 Vh tan  rT In positive moment area, compression stress in top
frb = 2 < b Frb = 6
41 2
7
5 b Fy ;
4 tw  e 2Cc Cb (10) of the concrete must be less than allowable compression
stress of concrete and tensile stress in the bottom of
where , e, and dh are the cutting angle, hole pure steel must be less than the allowable tensile stress of
distance, and cutting depth of castellated beam (Fig- steel:
ure 1), respectively; tw is the thickness of the web; M1 Mu < min(Mn con ; Mn st )
and M2 are the moment at each beam end; Es is steel
module of elasticity; and b is equal to 0.9, similar to = min(b 0:7Fc Zn com top ; b Fy Zn com bot );
the moment equation. (12)
By lling the hole, radial moment can be con-
trolled. where Mn con and Mn st are the nominal moment ca-
pacities of composite beam according to concrete part
3. Design of composite beams and steel part of the beam, respectively; Zn com bot
and Zn com top are section modules of bottom part
In composite beam with complete interaction, the and top part of the composite net section, respectively.
center line should be found and then moment of inertia b for concrete and steel is considered to be equal
of the composite section should be calculated. In this to 0.9 (AISC 360-10) [19].
study, temporary shores are considered for use during In negative moment area, cracked concrete cannot
construction and only composite section is designed for sustain tensile stress and only steel section should be
the total live and dead loads. Eective width of the considered as presented in Eq. (1).
concrete slab should not exceed the limits as described The available shear strength of the composite
in AISC 360-10 [19]: beams will be determined based upon the properties
of the steel section alone as presented in Eq. (5), AISC
1. One-eighth of the beam span, center-to-center of 360-10 [19].
the connections;
2. One-half of the distance to the centerline of the 4. Semi-rigid connection
adjacent beam;
3. The distance to the edge of the slab. Considering partial xity in connection and redistribu-
tion of internal member forces by semi-rigid connec-
According to the height to web thickness ratio, tion can help designers to decrease the total cost of
Mn should be determined from the superposition of building, Figure 2. Modeling, analyzing, and testing
elastic stresses or from the plastic stress distribution of this type of connections have been the subject of
on the composite section. In this study, superposition many research eorts. Linear, bi-linear, and tri-linear
of elastic stresses is considered because behavior of the moment-rotation models can be adopted to describe
composite castellated beam plastic is unpredictable. the connection behavior. The component method is
In order to consider the eect of dierential applied to dene and estimate the properties of the
shrinkage and creep on a composite steel-concrete connection [22]. It is usual to consider the web angles
structure, eective length can be divided by 3 [21]. subjected to plastic behavior under construction loads
166 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173

for the distributed load is dened as:

1 1
Rj = =
1 + 2SEjsLIdef
T
1 + 2Es Idef
Es Areb (d )2 L
s T

Leff

1
= ; (15)
1+ 2Idef Leff
Areb (ds )2 LT 
where Idef is the eective moment inertia for de ection
calculation.
According to this equation, the properties of rebar
can be determined. But, in the present study, only
optimal xity factor is calculated.
Figure 2. Model of a semi-rigid connection. The common shrinkage and temperature rein-
forcements parallel to beam can help to have sucient
partial xity. If there is no sucient reinforcement,
that do not contribute to the initial stiness of the partial xity will be limited.
composite joint. Initial stiness of joint (Sj;ini ) can be Moment resistance of the semi-rigid joint must be
estimated by: controlled as:
P P
Mi Fi  di Fs  ds Mu < b Mj Rd = b Fy reb Areb ds ; (16)
Sj;ini = = =
d d d
where Mj:Rd is the moment resistance of semi-rigid
Es Areb d d
Leff r  dr EA joint.
= = s reb (dr )2 ; (13)
d Le
5. Semi-rigid composite castellated beam
where Mi , Fi , and di are the moment, force, and
distance, respectively, and is rotation of semi-rigid 5.1. De ection of semi-rigid composite
joint. Areb , dr , and Le are the area, distance, and castellated beam
eective length of the rebars, respectively. Simple equations govern the elastic exural de ection
This equation is based on small rotation of con- of a typical beam. In the castellated beam, shear defor-
nection and calculates the force and moment of the mation should be considered. According to Jackson [2],
eective component. real moment of inertia for de ection is close to the
In this study, the bilinear model presented in Mur- composite moment of inertia of net section. However,
ray et al. [23] is considered for semi-rigid connection, AISC design guide for oor vibration indicates that
Figure 3. According to Silva et al. [22], the factor  for the shear studs are not suciently sti to justify the
beam to beam joint is considered 3: fully transformed moment of inertia assumption for the
composite beams; the following equation determines
Sj;ini the eective moment of inertia [23]:
Sj = ; (14)

(I 0:85In )
where Sj is the eective stiness of semi-rigid connec- Idef = 0:85In + com ; (17)
4
tion.
After calculating Sj , from the basic equation for where Icom and In are the composite section and net
calculating rotation of the beam, the xity factor, Rj , section moments of inertia for de ection calculation.
Also, the coecient 0.85 identies the shear eect for
typical open web beam.
In negative moment area, three parts are impor-
tant for calculating the de ection:
1. De ection from changing end slope of the beam:
def1 = tan  Lneg : (18)
2. De ection of semi-cantilevered beam at negative
moment part:
W Lneg 4 ( W L2pos )Lneg 3
def2 = + : (19)
Figure 3. Bending moment-rotation curve of a joint [23]. 8Es In 3Es In
 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173 167

3. De ection of semi-pinned beam at positive moment where Sb is beam spacing and hc is concrete height.
part: (DLF)max values for various natural frequencies are
presented in [24].
5W Lpos 4 The required damping ratio for the specied
def3 = ; (20)
384Es Idef amplitude and frequency is determined by [24]:
Lpos Dreq = 35Ao f + 2:5 < Dall = 0:035: (27)
def = def1 + def2 + def3 < defall = ; (21)
360
For a damping ratio above 3.5%, designer must
where Lneg and Lpos are the lengths of the negative either identify an exact source of damping or articially
and positive parts of the beam, respectively. provide additional damping to be sure that the oor
For concentrated loading, similar calculation must system is satisfactory [24].
be performed.
6. Optimization algorithms
5.2. Vibration of semi-rigid composite
castellated beam There are many methods for solving the optimization
Frequency of the load carrying system of a oor is problems. Mathematical programming based methods
the most important factor to identify the level of solve these problems when the objective function and
serviceability of the oor. Frequency is related to constraints have linear or non-linear relationship and
the stiness of the beam, boundary condition, and gradients of the functions are easily accessible. Because
distribution of the mass. of complexity of the structural optimization, many
The maximum initial amplitude of the beam is the problems in this eld cannot be easily handled by
other important parameter to identify the serviceability mathematical programming methods. After 1988,
level of a oor. meta-heuristic based methods were gradually devel-
Frequency of beam can be estimated by [24]: oped and increasing the power of the computers helped
the researchers to solve optimization problems more
def1 + def2 eciently. Genetic Algorithm (GA) [25] and Particle
defvib = def3 1:5 + defcol ; (22) Swarm Optimization (PSO) [26] are the most common
+
v ones.
r u W r
1 Stiness 1 u def g In this paper, the PSO, CBO, and ECBO algo-
f= = t vib rithms are used. In order to have reasonable com-
2 Mass 2 W defvib
g parison, the number of iterations and the number of
r population sizes are selected to be the same in all three
1 g algorithms.
= ; (23)
2 defvib PSO is based on sharing information between
each pair of particles in the swarm and updating each
where W and g are the eective weight and gravity particle's position based on its memory and the data
acceleration, respectively. gained regarding other particles.
defcol is considered zero and W can be calculated
by 0.2 time the live load in addition to dead loads. 6.1. CBO and ECBO
The maximum initial amplitude (inch) of the Colliding Bodies Optimization (CBO) is a recently de-
beam is determined as [24]: veloped meta-heuristic algorithm [27]. In this method,
one object collides with another object and they move
Aot =(DLF)max towards a minimum energy level. The CBO is simple in
! concept and depends on no internal parameter. Each
0:6(Lpos  0:393)3 Colliding Body (CB) has a specied mass related to
 ;
48(Es  14:22  10 3 )(Idef  0:3934 ) (24) the tness function. In order to select pairs of objects
for collision, CBs are sorted according to their mass in
  a decreasing order and they are divided into two equal
Sb
Ne =2:97 0:0578  + 2:56  10 8 groups: stationary and moving. Moving objects collide
hc with stationary objects to improve their positions and
   3 push stationary objects towards better positions.
Lpos 4 L In order to improve the CBO to get faster and
 Idef
+ 0:0001
Sb
; (25)
more reliable solutions, Enhanced Colliding Bodies Op-
timization (ECBO) is developed by utilizing a memory
Aot to save a number of historically best CBs as well as by
A0 = : (26)
Ne using a mechanism to escape from local optima [28].
168 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173

7. Problem denition of the rolled steel prole in the standard steel section
list (British Standards) starting from 127  76  13 UB
7.1. Cost function and ending in 914  419  388 UB. Cutting angle is
Total cost function is dened in the following form: limited between 40 to 64 . Other bounds for variables
Costini = Costp + CostW + Costc + Costs : (28) are presented in the constraints.
In a real structure, it is not common to have
Each sub-cost is determined by multiplying the cor- dierent cutting shapes for one beam. Commercial
responding weight (for Costp and Costs ) or length cutting shapes for castellated beams vary in dierent
(Costw and Costc ) by appropriate coecient. Values countries. The characteristic properties of one of the
of these coecients are determined either by other most popular commercial shapes are as follows:
researchers [9] or by engineering experiences: 1. Cutting angle is equal to 63.4 ;
Pp = 2  ((s 2e) + (Hs 2dt )) ; (29) 2. Cutting depth is half of the beam height;
  3. Holes spacing is three times the cutting depth.
L
Nh = T ; (30) 7.3. Constraints
s
As mentioned in the previous section, there are some
Lw = (Nh + 1 2  Nfh )  e + Pp  Nfh  4; (31) design constraints as follows:
3
Costw = (Lw )  Cw ; (32) g1 = dh (H 2tf ); (36)
8 s
Costp = (As  LT )  Cp ; (33) g2 = (Hs 2tf ) 10(dt tf ); (37)
 
Nh  dh 2
Costc = 2  Lw + 2  + Pp  Nfh  4  Cc ; g3 = dh cot() e; (38)
Sin() (34) 3
g4 = e 2dh cot(); (39)
Costs = (La  Ass  Nss  s )  Cs ; (35)
where p, w, c, and s are prole, welding, cutting, and g5 = 2dh cot() + e 2dh ; (40)
shear studs, respectively. Nss and Ass are the number
and area of channel shear studs, respectively. , s, e, g6 = Mu b Mn st ; (41)
Hs , dh , and dt are represented in Figure 1. Nh and
Nfh are the total and lled hole numbers, respectively. Mu mu
g7 = + b Fy ; (42)
As described in pervious equations, the cost of Zn com bot Ztee com
lling end holes by plates is considered by adding their
weights, cutting length, and welding length to the total g8 = Mu b Mn con ; (43)
cost. Cost coecients are given in Table 1.
g9 = Mu Mj Rd ; (44)
7.2. Variables
In this study, 6 variables are used for nding optimum g10 = Vu  v Vn w; (45)
results consisting of prole section, cutting depth (dh ),
cutting angle (), holes spacing (s), number of lled Vu
g11 = v Vn tee ; (46)
end holes of the castellated beams, and rigidity of 2
the connection (Rj ). The minimum and maximum
magnitudes of the variables must be clear for avoiding g12 = Vh v Vn p ; (47)
non-acceptable results and for fast convergence to the
global optimum. Prole section is the sequence number g13 = frb b Frb ; (48)

Table 1. Cost coecients. g14 = def defall ; (49)

Coecient Value Unit g15 = Dreq Dall ; (50)
Cp 0.3 $ per each kg
Cw 1 $ per each m g1 to g5 are related to constructional procedure of
Cc 0.8 $ per each m
manufacturing the castellated beams [9]. Also, in order
to have a good comparison between constraints, they
Cs 2.4 $ per each kg are normalized.
 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173 169

7.4. Penalty function

In order to avoid non-acceptable results, penalty func-
tion increases the cost of results that do not satisfy
the constraints. The constraint is satised when gi
is greater than zero. Thus, sum of the non-satisfying
constraints can show the degree of non-acceptability.
According to the above sentences, penalty factor is Figure 4. Schematics of the beam in Example 1.
dened by the following equation:
8.1. Example 1
PF = 10NAC ; (51) A simply supported beam with 4 m span, shown in
NAC = sum(gi > 0); (52) Figure 4, is selected as the rst design example. The
beam is subjected to 5 kN/m dead load, including
Costn = Costini  PF; (53) its own weight, and concentrated live load of 50 kN.
Eective width and height of concrete are 150 cm
where Costn and Costini are the nal and initial costs, and 10 cm, respectively. Number of iterations and
respectively. population size are considered 80 and 100, respectively,
Normalizing constraints also helps to have good for all the utilized algorithms.
results when they are summed together.
8.2. Example 2
8. Design examples A simply supported beam with 9 m span, shown in
Figure 5, is selected as the second design example. The
In order to compare the fabrication cost of the semi- beam is subjected to 40 kN/m dead load, including
rigid joint composite castellated beams with those of its own weight, and two concentrated live loads of
dierent methods, three examples are selected. Modu- 50 kN. Eective width and height of concrete are
lus of elasticity is taken 205 kN/mm2 and grade 50 is 150 cm and 15 cm, respectively. For all the considered
selected for the steel of the beam, which has the design algorithms, the number of iterations and population
strength of 355 MPa. Concrete design strength is equal size are considered 120 and 150, respectively.
to 25 MPa. The combinations for live and dead loads Table 2 summarizes the main objectives of the
are dened as those in ASCE [29]. performed optimizations. Four methods are presented
Table 2. Comparison of the optimum designs for the considered examples.
Combination Castellated Composite Semi-rigid Filled hole Cost % Critical
type beam beam connection at end ($) constraint
Example 1
Kaveh and p { {
p 96.04 1.24
Shokohi [10]
1
p {
p 1.00 HS FM
2
p {
p{ {
77.72
77.53 1.00 HS
3
p p 0.87 DE
4
p p{ {
p 67.74
1.00 HS
5
p p{ p 77.53
0.87 DE
6
p p p{ 67.74
0.87 DE
7
p p p p- 67.74
61.61 0.79 FM DE
Example 2
Kaveh and p { {
p 1031 1.59
Shokohi [10]
1
p {
p 1.00 DE RM FM
2
p {
p{ {
647
613 0.95 HS RM FM
3
p p { { 605 0.94 HS RM DE
4
p p p 0.76 HS RM
5
p p{ p 489
0.94 HS RM DE
6
p p p{ 605
0.94 HS RM DE
7
p p p p{ 605
431 0.67 HS RM DE
 HS: Horizontal Shear, DE: De ection, FM: Flexural Moment, RM: Radial Moment.
170 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173

8.3. Example 3
In this example, a simply supported composite beam
with dierent spans (600, 750, and 900 cm) is subjected
to 50 kN/m to 100 kN/m dead load, including its own
weight, having no concentrated load. Eective width of
the concrete is 150 cm and concrete heights are 15 cm
Figure 5. Schematics of the beam in Example 2. and 10 cm. In order to study the eect of partial xity
and use of commercial cutting shape, the following
at each column
p and their combinations are marked at conditions are considered.
each row by . Net cost and relative cost are displayed
in two columns. Over 90% ratio constraint is called 8.3.1. Optimum value of partial xity
critical constraint and the ratio constraints are shown For neglecting the eect of cutting shape and focusing
in the last columns. on the partial xity, a commercial shape of the castel-
The results of the examples are shown in Table 3. lated beams is considered. Also, two holes of each end
Figure 6 shows the variation of cost versus the number are considered to be lled. Thus, only partial xity
of iterations for Example 2. and beam section are variables. Due to the reduction
in the number of variables, the maximum number of
iterations and population size are considered 30 and
20, respectively. Here, only ECBO algorithm is used
to nd the optimum result.
Figure 7 shows that the optimum partial xity can
be changed from 40 to 100 percent, and there is no clear
relationship among xity and beam span, intensity of
distributed load, and concrete thickness.
8.3.2. Cost saving of the use of partial xity
connection
Similar to the previous section, to ignore the eect of
cutting shape and to focus on the eect of partial xity,
commercial shape of castellated beam is considered.
Also, two holes of each end are considered to be
lled. Hence, only partial xity and beam section
are variables. For each choice of span and load,
Figure 6. Variation of cost versus the number of two optimum results are found; one for the partial
iterations in Example 2. xity condition and the other for simple connection

Figure 7. Optimum partial xity percent versus the intensity of distributed load with 10 cm and 15 cm concrete
thicknesses in Example 3.
Table 3. Results of the examples.
Section Cutting Cutting Holes Composite Number of lled Partial Cost
depth angle number end holes xity
Example 1 152  89  16 6.06 cm 60  26
p 1 0.22 61.6$
Example 2 457  152  52 14.3 cm 64 20
p 4 0.65 431.4$
 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173 171

Figure 8. Cost saving percent of using partially xed connections versus the intensity of distributed load with 10 cm and
15 cm concrete thicknesses in Example 3.

Figure 9. Cost saving percent of using non-commercial cutting shape versus the intensity of distributed load with 10 cm
and 15 cm concrete thicknesses in Example 3.

condition. Relative dierence between the costs of 1. Semi-rigid joint composite castellated beam can be
two conditions is considered as the amount of cost viewed as the best choice in the rst three examples,
saving. Algorithm parameters are similar to those in and it can decrease the cost by 21% to 35%;
the previous case. 2. Eect of partial xity cost saving is estimated 5 to
Figure 8 shows that reduction in optimum cost 25 percent;
can be changed from 5 to 25 percent.
3. In many problems, horizontal shear at the end of
8.3.3. Cost saving of using non-commercial cutting beam controls the optimization problem. Filling
shape the end holes can improve these constraints and
Dierence between the costs of commercial shape it works better in some problems with distributed
and non-commercial shape of the castellated beam is load. Also, lling the end holes, as far as the
considered as the cost saving of using non-commercial rst non-lled holes are placed in positive moment
shape of the castellated beam. All variables in non- length of the beam, can adequately control the
commercial shape of castellated beam are considered. horizontal shear and radial moment;
Number of lled holes is limited to 2. Only ECBO
algorithm is used and maximum numbers of iterations 4. The ECBO algorithm seems to result in better
and population size are considered 100 and 80. design in a higher number of iterations. This
Figure 9 shows that the percentage of cost saving eciency is more sensible when the problem has
can be changed from 6 to 30. wide variable bounds and it is more complex;
5. Use of commercial cutting shape for composite
castellated beam can increase the cost by 6 to 30
9. Discussion and conclusions percent.
In this paper, the PSO, CBO, and ECBO algorithms
were utilized to optimize the process of nding the best Acknowledgement
property of semi-rigid jointed composite castellated
beam. The rst author is grateful to Iran National Science
Comparison of the observations is as follows: Foundation for the support.
172 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173

References 16. Kameshki, E. and Saka, M. \Optimum design of non-

linear steel frames with semi-rigid connections using a
1. Redwood, R. and Cho, S.H. \Design of steel and genetic algorithm", Comput. Struct., 79(17), pp. 1593-
composite beams with web openings", J. Construct 1604 (2001).
Steel Res., 25(1), pp. 23-41 (1993).
17. Ramires, F.B., de Andrade, S.A.L. and de Lima,
2. Jackson, R., Vibration and Flexural Strength Char- L.R.O. \Genetic algorithm optimization of composite
acteristics of Composite Castellated Beams, Virginia and steel endplate semi-rigid joints", Eng. Struct., 45,
Polytechnic Institute and State University, USA pp. 177-191 (2012).
(2012).
18. Ali, N.B.H., Sellami, M., Cutting-Decelle, A-F. and
3. Wang, Y. and Li, T. \Flexural bearing capacity Mangin, J.C \Multi-stage production cost optimiza-
research of composite beams with edge constraint tion of semi-rigid steel frames using genetic algo-
component", Adv Mater. Res., 639-640, pp. 807-811 rithms", Eng. Struct., 31(11), pp. 2766-2778 (2009).
(2013).
19. Committee, A., Specication for Structural Steel Build-
4. Ellakany, A.M. and Tablia, H.A. \A numerical model ings (ANSI/AISC 360-10), American Institute of Steel
for static and free vibration analysis of elastic compos- Construction, Chicago-Illinois, USA (2010).
ite beams with end shear restraint", Meccan., 45(4),
pp. 463-474 (2010). 20. Kerdal, D. and Nethercot, D. \Failure modes for
castellated beams", J. Construct. Steel Res., 4(4), pp.
5. Morton, S. and Webber, J. \Optimal design of a 295-315 (1984).
composite I-beam", Compos. Struct., 28(2), pp. 149-
168 (1994). 21. Roll, F., Eects of Dierential Shrinkage and Creep
on a Composite Steel-Concrete Structure, ACI Special
6. Senouci, A.B. and Al-Ansari, M.S. \Cost optimization Publication, 1971. 27 (2010).
of composite beams using genetic algorithms", Adv.
Eng. Softw., 40(11), pp. 1112-1118 (2009). 22. Silva, L.S., Sim~oes, R. and Gervasio, H. \Design
of steel structures: Eeurocode 3: design of steel
7. Sorkhabi, R.V., Naseri, A. and Naseri, M. \Opti- structures", Part 1-1: General Rules and Rules for
mization of the castellated beams by particle swarm Buildings, John Wiley & Sons (2012).
algorithms method", APCBEE Procedia, 9, pp. 381-
387 (2014). 23. Murray, T.M., Allen, D.E. and Ungar, E.E., Floor
Vibrations Due to Human Activity, American Institute
8. Kaveh, A. and Shokohi, F. \A hybrid optimization of Steel Construction (2003).
algorithm for the optimal design of laterally-supported
castellated beams", Scientia Iranica, 2(23), pp. 508- 24. Naeim, F., Design Practice to Prevent Floor Vibra-
519 (2016). tions, Steel Committee of California, USA (1991).
9. Kaveh, A. and Shokohi, F. \Cost optimization of 25. Golberg, D.E., Genetic Algorithms in Search, Op-
castellated beams using charged system search algo- timization, and Machine Learning, Addison Wesley
rithm", Iranian J. Sci. Technol., Trans. Civil Eng., (1989).
38(C1), pp. 235-249 (2014). 26. Eberhart, R.C. and Kennedy, J. \A new optimizer
10. Kaveh, A. and Shokohi, F. \Cost optimization of using particle swarm theory", in Proc. Sixth Int. Symp.
end-lled castellated beams using meta-heuristic algo- Micro Mach. Human Sci., New York, USA (1995).
rithms", Int. J. Optim. Civil Eng., 5(3), pp. 333-354 27. Kaveh, A. and Mahdavi, V.R. \Colliding bodies op-
(2015). timization: A novel meta-heuristic method", Comput.
11. De Oliveira, T.J.L. and De Miranda Batista, E. \Mod- Struct., 139, pp. 18-27 (2014).
elling beam-to-girder semi-rigid composite connection 28. Kaveh, A. and Ilchi Ghazaan, M. \Enhanced colliding
with angles including the eects of concrete tension bodies optimization for design problems with continu-
stiness", Eng. Struct., 31(8), pp. 1865-1879 (2009). ous and discrete variables", Adv. Eng. Softw., 77, pp.
12. Fu, F., Lam, D. and Ye, J. \Parametric study of semi- 66-75 (2014).
rigid composite connections with 3-D nite element 29. ASCE, Minimum Design Loads for Buildings and
approach", Eng. Struct., 29(6), pp. 888-898 (2007). Other Structures, American Society of Civil Engineers,
13. Gil, B., Go~ni, R. and Bayo, E. \Experimental Reston, Virginia, USA (2006).
and numerical validation of a new design for three-
dimensional semi-rigid composite joints", Eng. Struct.,
48, pp. 55-69 (2013). Biographies
14. Rex, C.O. and Easterling, W.S. \Partially restrained Ali Kaveh was born in 1948 in Tabriz, Iran. After
composite beam-girder connections", J. Construct. graduation from the Department of Civil Engineering
Steel Res., 58(5), pp. 1033-1060 (2002). at the University of Tabriz in 1969, he continued his
15. Simoes, L. \Optimization of frames with semi-rigid studies on Structures at Imperial College of Science
connections", Comput. Struct., 60(4), pp. 531-539 and Technology at London University, and received his
(1996). MS and PhD degrees in 1970 and 1974, respectively.
 A. Kaveh and M.H. Ghafari/Scientia Iranica, Transactions A: Civil Engineering 25 (2018) 162{173 173

He then joined the Iran University of Science and Mohammad Hossein Ghafari was born in 1989 in
Technology in Tehran, where he is presently Profes- Tehran. He obtained his BS degree in Civil Engineering
sor of Structural Engineering. Professor Kaveh is from University of Tehran, Iran, In 2012 and his
the author of 525 papers published in international MS degree in Earthquake Engineering from the same
journals and 150 papers presented at international university in 2014. At present, he is a PhD student
conferences. He has authored 23 books in Farsi and at Iran University of Science and Technology Tehran,
7 books in English published by Wiley, the Ameri- Iran, and he works on optimum design of composite
can Mechanical Society, Research Studies Press, and oors and steel moments frame via the meta-heuristic
Springer. methods.