Journal of Constructional Steel Research 66 (2010) 767–771

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

journal homepage: www.elsevier.com/locate/jcsr

Charged system search for optimum grillage system design using the LRFD-AISC
code
A. Kaveh a,∗ , S. Talatahari b
a
Centre of Excellence for Fundamental Studies in Structural Engineering, Iran University of Science and Technology, Narmak, Tehran-16, Iran
b
Department of Civil Engineering, University of Tabriz, Tabriz, Iran

article info abstract

Article history: Grillage systems are widely used in structures to cover large areas in bridge decks, ship hulls and floors.
Received 27 September 2009 In this paper, the charged system search (CSS) algorithm is utilized to obtain the optimum design of
Accepted 13 January 2010 grillage systems. This algorithm is inspired by the Coulomb and Gauss laws of electrostatics in physics
and the governing laws of motion from Newtonian mechanics. The cross-sectional properties of beams
Keywords: are considered as the design variables. Comparison of the results with those of some previous studies
Charged system search
shows the robustness of the new algorithm.
Heuristic optimization algorithm
Grillage systems
© 2010 Elsevier Ltd. All rights reserved.
Warping

1. Introduction
Meta-heuristic algorithms are more suitable than conventional
methods for structural optimum design due to their capability of
exploring and finding promising regions in the search space in an
affordable time [1]. Meta-heuristic algorithms tend to perform well
for most engineering optimization problems [2–7]. This is because
these methods refrain from simplifying or making assumptions
about the original form.
One of the recent meta-heuristic optimization techniques is
the charged system search (CSS), inspired by the governing laws
of electrostatics in physics and the governing laws of motion
from the Newtonian mechanics [8]. The CSS utilizes a number of
solution candidates which are called charged particles (CPs). Each
CP is treated as a charged sphere and it can exert an electrical
force on the other agents (CPs) according to the Coulomb and
Gauss laws of electrostatics. The resultant force acting on each CP
creates an acceleration in consideration of Newton’s second law.
Finally, utilizing Newtonian mechanics, the position of each CP is
determined at any time based on its previous position, velocity and
acceleration in the space. In this study, the CSS is used to determine
the optimum design of grillage systems.
Grillage systems are widely used in structures to cover large
areas in bridge decks, ship hulls and floors. The aim of the
optimum design of a typical grillage system is to determine
the cross-sectional properties of its longitudinal and transverse
beams such that the response of the system under the external
loading is within the allowable limits defined by the code of
practice, while the weight or the cost of the system becomes
the minimum. In one of the early studies, the optimum design
problem is formulated by treating the moments of inertia of the
beams and joint displacements as design variables [9]. Stiffness,
stress, displacement and size constraints are included in the
design formulation. The effect of warping is taken into account
in the computation of the stresses in the members and a genetic
algorithm based method is developed for the optimum design
of grillage systems which selects the optimum sections for the
grillage elements from a set of standard universal beam sections
and finds the optimum number of required beams for the grillage
system [10]. A harmony search algorithm is utilized to determine
the optimum W-sections for the members of grillage system from
the set of LRFD-AISC sections in [11]. Also, the effect of the spacing
between the longitudinal and transverse beams of a grillage system
is investigated in [12] using a harmony search. In these studies,
the deflection limitations and the allowable stress constraints are
considered in the formulation of the design problem.
Here, the ability of the CSS in finding the optimum cross-
sectional properties of longitudinal and transverse beams of some
grillages is studied. The results are compared to those of the
harmony search and genetic algorithm to illustrate the efficiency
of the present approach.
2. Charged system search

∗ Corresponding author. Tel.: +98 21 77240104; fax: +98 21 73223106. The charged system search is based on electrostatic and
E-mail address: alikaveh@iust.ac.ir (A. Kaveh). Newtonian mechanics laws. The Coulomb and Gauss laws provide
0143-974X/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2010.01.007
768 A. Kaveh, S. Talatahari / Journal of Constructional Steel Research 66 (2010) 767–771

the magnitude of the electric field at a point inside and outside a probability function:
charged insulating solid sphere, respectively, as follows [13]:
fit (i) − fitbest

1 > rand ∨ fit (j) > fit (i)
 
ke qi pij = fit (j) − fit (i) (9)
rij if rij < a


a3
 0 else
Eij = ke qi (1)


 r2 if rij ≥ a and calculate the attracting force vector for each CP as follows:
ij !
X qi qi
where ke is a constant known as the Coulomb constant; rij is the Fj = qj rij · i1 + · i2 pij (Xi − Xj )
separation of the centre of sphere and the selected point; qi is the i,i6=j
a3 rij2
magnitude of the charge; and a is the radius of the charged sphere.
j = 1, 2, . . . , N
*
Using the principle of superposition, the resulting electric force due
to N charged spheres is equal to [8] i1 = 1, i2 = 0 ⇔ rij < a (10)
! i1 = 0, i2 = 1 ⇔ rij ≥ a
N
X qi qi ri − rj where Fj is the resultant force affecting the jth CP.
Fj = ke qj rij · i1 + · i2
i=1
a3 rij2 kri − rj k Step 2. Solution construction. Move each CP to the new position and
find its velocity using the following equations:
i1 = 1, i2 = 0 ⇔ rij < a

(2)
i1 = 0, i2 = 1 ⇔ rij ≥ a. Fj
Xj,new = randj1 · ka · · 1t 2 + randj2 · kv · Vj,old · 1t + Xj,old (11)
Also, according to Newtonian mechanics, we have [13] mj
Xj,new − Xj,old
1r = rnew − rold (3) Vj,new = (12)
rnew − rold
1t
v= (4) where randj1 and randj2 are two random numbers uniformly
1t
distributed in the range (0, 1). mj is the mass of the CPs, which
vnew − vold
a= (5) is equal to qj in this paper. The mass concept may be useful for
1t developing a multi-objective CSS. 1t is the time step, and it is set
where rold and rnew are the initial and final position of a particle, to 1. ka is the acceleration coefficient; kv is the velocity coefficient
respectively; v is the velocity of the particle; and a is the to control the influence of the previous velocity. Here, ka and kv are
acceleration of the particle. Combining the above equations and taken as 0.5.
using Newton’s second law, the displacement of any object as a Step 3. CP position correction. If each CP exits from the allowable
function of time is obtained as search space, correct its position using the HS-based handling
approach as described for the HPSACO algorithm [14,15].
1 F
rnew = · 1t 2 + vold · 1t + rold . (6) Step 4. CP ranking. Evaluate and compare the values of the fitness
2m
function for the new CPs; and sort them in an increasing order.
Inspired by the above electrostatic and Newtonian mechanics laws,
Step 5. CM updating. If some new CP vectors are better than the
the pseudo-code of the CSS algorithm is presented as follows [8].
worst ones in the CM, in terms of their objective function values,
Level 1: Initialization include the better vectors in the CM and exclude the worst ones
Step 1. Initialization. Initialize the parameters of the CSS algorithm. from the CM.
Initialize an array of charged particles (CPs) with random positions. Level 3: Controlling the terminating criterion
The initial velocities of the CPs are taken as zero. Each CP has Repeat the search level steps until a terminating criterion is
a charge of magnitude (q) defined considering the quality of its satisfied.
solution as
fit (i) − fit w orst 3. Optimum design of grillage systems
qi = i = 1, 2, . . . , N (7)
fitbest − fit w orst
3.1. Objective function
where fitbest and fitworst are the best and the worst fitness of all
the particles; fit(i) represents the fitness of agent i. The separation The optimum design problem of a grillage system is to reach
distance rij between two charged particles is defined as a set of design variables corresponding to the minimum weight
satisfying the behavioral and performance limitations which are
kXi − Xj k adopted from the Load and Resistance Factor Design, American
rij = (8)
k(Xi + Xj )/2 − Xbest k + ε Institute of Steel Construction (LRFD-AISC) [16]. This can be
expressed as
where Xi and Xj are the positions of the ith and jth CPs,
respectively; Xbest is the position of the best current CP; and ε is Find A = [A1 , A2 , . . . , Ang ]
a small positive number to avoid singularities. Ai ∈ Di
Step 2. CP ranking. Evaluate the values of the fitness function for the
nm
X (13)
CPs, compare with each other and sort them in increasing order. to minimize W (A) = γi · Ai · li
i=1
Step 3. CM creation. Store the number of the first CPs equal to
charged memory size (CMS) and their related values of the fitness where A is a set of design variables (cross-sectional areas); ng is the
functions in the charged memory (CM). number of design groups; Di is the allowable set of values for the
design variable Ai given in the W-section list of LRFD-AISC; W (A) is
Level 2: Search the weight of the structure; nm is the number of members making
Step 1. Attracting force determination. Determine the probability up the structure; γi is the material density of member i; li is the
of moving each CP toward the others considering the following length of member i.
 A. Kaveh, S. Talatahari / Journal of Constructional Steel Research 66 (2010) 767–771 769

3.2. Design constraints to satisfy LRFD-AISC In Eq. (18), Mcr is the buckling moment; Fcr is the critical stress;
λ = bf /(2tf ) for I-shaped member flanges, in which bf and tf are
The following constraints are required to be imposed for grillage the width and the thickness of the flange; λ = h/tw for a beam
systems according to LRFD-AISC provisions [16]. web, in which h = d − 2k plus allowance for undersize inside fillet
Maximum displacement limitations: at compression flange for rolled I-shaped sections; d is the depth
of the section; k is the distance from the outer face of the flange to
δi
≤ 1, i = 1, 2, . . . , nj. (14) the web toe of the fillet; tw is the web thickness.
δiu λp and λr are given in table LRFD-B5.1 of the code as
The strength constraints:  s
E
0.38 F

for compression flange

Mu,i 
≤ 1, i = 1, 2, . . . , nm

(15)
φb Mn,i λr = s y (20)
 E
3.76 F

Vu,i  for the web
≤ 1, i = 1, 2, . . . , nm

(16) y
φ v V n ,i  s
where δi is the displacement of joint i and δiu is its upper bound; E
0.83 F − F

for compression flange


Mu,i is the required flexural strength in member i; Mn,i denotes 
the nominal flexural strength; φb is flexural resistance reduction λp = s y r
(21)
factor (φb = 0.90); Vu,i is the factored service load shear for E

5.70 F

 for the web
member i; Vn,i is the nominal strength in shear; and φv represents

y
the resistance factor for shear given as 0.9.
For a steel grillage system when the elements are subjected in which E is the modulus of elasticity and Fy is the yield stress
to torsional moments, warping takes place, and if the warping of steel. It is apparent that Mn is computed for the flange and for
is restrained, it causes large values of normal stresses in the the web separately by using corresponding λ values. The smallest
sections. Hence it becomes necessary to consider the effect of of these is taken as the nominal moment strength of the section
warping in the analysis of grillage systems [10]. According to LRFD- under consideration.
AISC [16], if the effect of warping is taken into account, using the The nominal shear strength of a rolled compact and non-
above constraints may be simplistic, and therefore the following compact W-section is computed from the data given in LRFD-AISC-
combinational constraint is considered instead of Eq. (15): F2.2 as follows:
σbx σby σw  s
± ± ± ≤ 1, i = 1, 2, . . . , nm (17) h E
0.6Fyw Aw ≤ 2.45

φb Fcr 0.9Fy 0.9Fy


t F
sw yw



where Fcr is the critical flexible stress; σb is the normal stress due


t E

w
to bending about either the x-axis or the y-axis; σw is the normal 1.47Fyw Aw h F




stress due to warping. yw
Vn = s s (22)
The number of degrees of freedom for a grillage joint without  E h E
2.45 < ≤ 3.07

considering the warping is three. The stiffness matrix of a grillage


Fyw tw Fyw


element with three degrees of freedom at each joint is available in 
 s
.

2
standard textbooks. The number of degrees of freedom of a joint in 4 52Et E h

w

3.07 < ≤ 260.

A
a grillage system where the effect of warping is taken into account
 w 2 h Fyw tw
becomes four, and the related stiffness matrix is given in detail in
Ref. [10].
4. Design examples
3.3. The nominal strength of members
In this section, two grillage systems are optimized utilizing the
In order to compute the nominal moment strength Mn of a new algorithm. The final results are then compared to the solutions
laterally supported beam, it is necessary first to determine whether of other advanced heuristic methods to demonstrate the efficiency
the beam is compact, non-compact or slender. In compact sections, of this work. The yield stress of materials is 250 MPa. The values
local buckling of the compression flange and the web does not of 205 kN/mm2 and 81 kN/mm2 are used for the modulus of
occur before a plastic hinge develops in the cross section. On the elasticity and the shear modulus, respectively. The discrete set
other hand, in practically compact sections, local buckling of the from which the design algorithm selects the sectional designations
compression flange or web may occur after the first yield has for grillage members is considered to be the complete set of 272 W-
taken place at the outer fiber of the flanges. The nominal moment sections as given in LRFD-AISC. For the CSS algorithm, a population
strength is given in the following, as defined in LRFD-AISC: of 20 CPs is used. 250 iterations are selected as the maximum
number of iterations. The algorithms are coded in Matlab and the
M = Zx Fy ≤ 1.5Sx Fy λ ≤ λp

 p

λ − λr
systems are analyzed using the direct stiffness method.
Mn = MP − (MP − Mr ) λp < λ ≤ λr (18)
 λ r − λp 4.1. Example 1
Mcr = Sx Fcr ≤ Mp λ > λr


where Mp is the plastic moment; Z is the plastic section modulus;
and Mr is the limiting buckling moment, given as
Mr = (Fy − Fr )Sx
where Fr is the compressive residual stress in the flange, which is
given as 69 MPa for rolled shapes in the code; and Sx is the section
modulus.
The grillage system shown in Fig. 1 has 40 members which are
collected in four groups. The outer and inner longitudinal beams
are considered to be group 1 and group 2, respectively, while the
(19)
outer and inner transverse beams are taken as group 3 and group 4.
The external loading is equal to 200 kN in each unsupported node.
The vertical displacements of the four middle joints are restricted
to 25 mm.
770 A. Kaveh, S. Talatahari / Journal of Constructional Steel Research 66 (2010) 767–771

Fig. 3. A 60-element grillage system.

Fig. 1. A 40-element grillage system.

× 104
1.8

1.6

1.4
Weight (kg)

1.2

0.8
Fig. 4. A 112-element grillage system.
0.6
0 50 100 150 200 250 considered, having 2 m and 1.5 m beams, as shown in Figs. 3 and 4,
Iterations where the system has 60 members and 112 members, respectively.
The total external load is distributed to the joints of the grillage
Fig. 2. The convergence history for the 40-element grillage system.
system as point loads. Their values are calculated according to
Table 1
beam spacing. Firstly, the members of grillage structures are
Comparison of the optimal design for the first example. collected in two groups. The longitudinal beams are considered to
belong to group 1 and transverse beams are taken as group 2. The
Element group Optimal W-shaped sections
vertical displacements of middle joints are restricted to 25 mm.
GA [11] HS [11] CSS
The optimum sectional designations obtained by CSS are given in
1 W150 × 13.5 W410 × 46.1 W410 × 46.1 Table 2. For the 60-member grillage system, CSS finds a system
2 W610 × 92 W460 × 52 W460 × 52
3 W410 × 46.1 W200 × 15 W150 × 13
of weight 11,358 kg, while a weight of 14,384 kg is found by the
4 W840 × 226 W1000 × 222 W1000×222 harmony search algorithm. For the 112-member grillage system,
Weight (kg) 8087.91 7198.20 7168.04 the result of the CSS algorithm is 13,519 kg, which is 16.5% lighter
δ u (mm) 24.7 23.2 23.7 than the design by the harmony search and is 12.1% heavier than
Maximum strength ratio 0.83 0.99 0.99 that of the 60-member grillage system. In all these designs, the
displacements have been dominant.
The minimum weight of the design by CSS is 7168.04 kg, In order to have a better design, four group designs are
while it is 7198.2 kg and 8087.91 kg for the harmony search and considered as described for the previous example. The weight of
genetic algorithm, respectively [11]. The optimum design obtained the 60-member grillage system with four groups is 9251 kg, which
for the system is given in Table 1. CSS can find the optimum is 22.7% lighter than this system with two groups. For the 112-
design after 2960 iterations, while the harmony search obtains member grillage system, it is 11,548 kg, which is 14.5% lighter than
the optimum design after 4558 grillage analyses and the genetic the value with two groups. When the effect of warping is taken
algorithm requires 27,200 grillage analyses to reach the final into account, these values change to 13,477 kg and 15,519 kg for
solution [11]. The design history curve for the CSS algorithm is 60- and 112-member grillage systems, respectively. It is shown
plotted in Fig. 2. The maximum vertical displacement is 23.7 mm, that the effect of warping causes up to 45% and 35% increase in
while the maximum value of the strength ratio is 0.99. In this the overall weight of the structure for these grillages, respectively.
case, the strength constraints are dominant in the design problem The optimum results of these grillage systems obtained by CSS are
instead of serviceability constraints. summarized in Table 3. Fig. 5 shows the convergence history for
these two grillage systems. Both systems need only about 3000
4.2. Example 2 analyses to reach optimum results.

For this example, a 12 m × 12 m square area is considered. The 5. Concluding remarks

design problem is to set up a grillage system that is supposed to
carry 15 kN/m2 uniformly distributed load (the total load is 2160
kN). The grillage system that can be used to cover the area will
have the longitudinal beams of length 12 m and the transverse
beams are of length 12 m. Two different grillage systems are
Charged system search as a new meta-heuristic optimization
algorithm is presented to optimize grillage systems. The algorithm
contains three levels: initialization, search, and controlling the
terminating criterion. In the initialization level, the parameters
 A. Kaveh, S. Talatahari / Journal of Constructional Steel Research 66 (2010) 767–771 771

Table 2
Comparison of the optimal design for the second example with two groups.
Element group Optimal W-shaped sections
60-member grillage system 112-member grillage system
HS [12] CSS HS [12] CSS

1 W200 × 22.5 W150 × 13.5 W690 × 170 W150 × 13.5

2 W690 × 217 W840 × 176 W200 × 22.5 W770 × 147
Weight (kg) 14,384 11,358 16,198 13,519
δ u (mm) 25.0 24.2 24.1 24.3
Maximum strength ratio 0.48 0.54 0.73 0.45

Table 3
Comparison of the optimal design for the second example with four groups.
Element group Optimal W-shaped sections
60-member grillage system 112-member grillage system
Without warping With warping Without warping With warping

1 W150 × 13.5 W840 × 210 W150 × 13.5 W200 × 35.5

2 W910 × 201 W840 × 175 W840 × 176 W530 × 138
3 W300 × 21 W254 × 18 W150 × 13.5 W400 × 53.5
4 W300 × 32.5 W254 × 45.5 W300 × 21 W530 × 87.5
Weight (kg) 9251 13,477 11,548 15,519
δ u (mm) 24.3 18.1 24.4 16.2
Maximum strength ratio 0.99 0.99 0.75 0.97

2 60-members grillage system design certainly becomes unsafe. Comparing the results of CSS
112-member grillage system for the two grillage systems considered with those form other
heuristic algorithms such as the harmony search and the genetic
Current weight/Best weight

1.8
algorithm shows a good balance between the exploration and
exploitation abilities of the CSS; hence its superior performance
1.6 becomes evident.

Acknowledgement
1.4
The first author is grateful to Iran National Science Foundation
1.2 for the support.

1 References
0 50 100 150 200
Iterations
Fig. 5. The convergence history for the second example.
of the CSS algorithm, the primary location of the CPs and their
initial velocities are defined. Also in this level, a memory to store
a number of the best CPs is introduced. The search level starts
after the initialization level, where each CP moves toward the
others considering the probability function, the magnitude of the
attracting force vector, and the previous velocities. The moving
process is defined in a way that it not only can perform more
investigation in the search space, but can also improve the results.
To fulfill this goal, some laws of physics containing the Coulomb
and Gauss laws, and the governing laws of motion from Newtonian
mechanics are utilized. The last level consists of controlling the
termination.
Grillage systems are widely used in structures to cover large
areas. Finding the optimum design of a grillage system consists
of determining the cross-sectional properties of the longitudinal
and transverse beams such that the response of the system under
external loading remains within the allowable limits and the
weight of the system becomes minimum. Here, the CSS algorithm
is utilized to determine the optimum W-sections for the members
of a grillage system when the LRFD-AISC is utilized as the code
of practice and the deflection limitations and the allowable stress
constraints are considered as the constraints. Also, the effect of
warping is taken into account in the computation of the stresses.
It is shown that warping plays a significant role in the optimum
design of grillage systems, and if this effect is ignored, the optimum
250 [1] Kaveh A, Talatahari S. An improved ant colony optimization for constrained
engineering design problems. Engineering Computations 2010;27(1):155–82.
[2] Prendes Gero M, Bello García A, del Coz Díaz JJ. A modified elitist genetic
algorithm applied to the design optimization of complex steel structures.
Journal of Constructional Steel Research 2005;61(2):265–80.
[3] Kaveh A, Talatahari S. Optimal design of Schwedler and ribbed domes
via hybrid Big Bang–Big Crunch algorithm. Journal of Constructional Steel
Research 2010;66(3):412–9.
[4] Prendes Gero M, Bello García A, del Coz Díaz JJ. Design optimization of
3D steel structures: Genetic algorithms vs. classical techniques. Journal of
Constructional Steel Research 2006;62(12):1303–9.
[5] Gholizadeha S, Salajegheh E. Optimal seismic design of steel structures by
an efficient soft computing based algorithm. Journal of Constructional Steel
Research 2010;66(1):85–95.
[6] Camp C, Bichon J. Design of space trusses using ant colony optimization.
Journal of Structural Engineering, ASCE 2004;130(5):741–51.
[7] Lee KS, Geem ZW. A new structural optimization method based on the
harmony search algorithm. Computers and Structures 2004;82:781–98.
[8] Kaveh A, Talatahari S. A novel heuristic optimization method: Charged system
search. Acta Mechanica. doi:10.1007/s00707-009-0270-4.
[9] Saka MP. Optimum design of grillages including warping. In: Proc. of
international symposium on optimum structural design. 1981.
[10] Saka MP, Daloglub A, Malhasc F. Optimum spacing design of grillage systems
using a genetic algorithm. Advances in Engineering Software 2000;31:863–73.
[11] Saka MP, Erdal F. Harmony search based algorithm for the optimum design of
grillage systems to LRFD-AISC. Structural and Multidisciplinary Optimization
2009;38:25–41.
[12] Erdal F, Saka MP. Effect of beam spacing in the harmony search based optimum
design of grillages. Asian Journal of Civil Engineering 2008;9(3):215–28.
[13] Halliday D, Resnick R, Walker J. Fundamentals of Physics. 8th ed. John Wiley
and Sons; 2008.
[14] Kaveh A, Talatahari S. Particle swarm optimizer, ant colony strategy and
harmony search scheme hybridized for optimization of truss structures.
Computers and Structures 2009;87(5–6):267–83.
[15] Kaveh A, Talatahari S. A Particle swarm ant colony optimization algorithm
for truss structures with discrete variables. Journal of Constructional Steel
Research 2009;65(8–9):1558–68.
[16] LRFD-AISC. Manual of Steel Construction, Load and resistance factor design.
Metric conversion of the second edition, Chicago: AISC. vols. 1 & 2. 1999.