World Applied Sciences Journal 13 (12): 2484-2494, 2011

ISSN 1818-4952
© IDOSI Publications, 2011

Minimum Cost Design of Concrete Slabs using Particle Swarm Optimization

with time Varying Acceleration Coefficients

Behrouz Ahmadi-Nedushan and Hesam Varaee

Department of Civil Engineering, Faculty of Engineering, Yazd University,

Pejoohesh Street Safa-ieh, P.O. Box: 89195-741, Yazd, Iran

Abstract: This article deals with cost optimization of one-way concrete slabs according to the most recent
American Concrete Institute code of practice (ACI 318-M08). The objective is to minimize the total cost of
the slab including costs of concrete and reinforcement bars while satisfying all the design requirements.
Particle Swarm Optimization (PSO) is used for solving the constrained optimization problem. As PSO is
designed for unconstrained optimization problems, a multi-stage dynamic penalty was also implemented to
solve the constrained optimization problem. Cost optimization of four different slabs with different support
conditions are illustrated and the results of the optimum design results are compared with exis ting methods
in the literature. A sensitivity analysis of optimal designs was also performed by optimizing the four
examples for different span lengths between 2 to 5 meters to investigate the effect of span length on
optimal costs and optimal reinforcement ratios. The results demonstrate that PSO is a promising method in
design optimization of structural elements.

Key words: Concrete slab • optimization • penalty function • particle swarm optimization • sensitivity
analysis

INTRODUCTION Engineers are faced with challenge of designing

efficient and cost-effective systems without
Design is an iterative process. The designer’s compromising the integrity of the system. The
experience, intuition and ingenuity are required in the conventional design process depends on the designer’s
design of systems in most fields of engineering intuition, experience and skill. This presence of a
(aerospace, automotive, civil, chemical, industrial, human element can sometimes lead to erroneous
electrical, mechanical, hydraulic and transportation). results in the synthesis of complex systems.
Iterative implies that several trial designs are analyzed Furthermore, the conventional design process can lead
one after another until an acceptable design is obtained. to uneconomical designs and can involve a lot of
In the design process, the designer estimates a trial calendar time. The conventional design leads to safe
design of the system based on experience, intuition, or designs, but the economy of the design is very much
some mathematical analysis. The trial design is linked to experience of the structural designer. Scarcity
analyzed to determine if it is satisfies all the and the need for efficiency in today’s competitive world
requirements. If it is, the design process is terminated have forced engineers to evince greater interest in
and the design is accepted as final design [1]. economical and optimized designs.
In conventional design of structural systems and Another design method, which is more systematic,
elements, most procedures first the adopt the cross- is the optimization design process, in which the trial
section dimensions and material grades based on design is analyzed to determine if it is the best.
common practice. Once the structure is defined, the Depending on the specifications, “best” can have
structure is analyzed to determine the stress resultants. different connotations for different systems. In general,
For concrete structure, the procedure is followed it implies cost-effective, efficient, reliable and durable
by the computation of reinforcement that satisfy systems. The optimum design process forces the
the limit states prescribed by concrete codes. designer to use rigorous formulation of the design
Should the dimensions or material grades be problem by defining explicitly a set of design variables,
insufficient, the structure is redefined on a trial an objective function to be optimized and the imposed
and error basis [2]. constraint functions [1].
Corresponding Author: Dr. Behrouz Ahmadi-Nedushan, Department of Civil Engineering, Faculty of Engineering, Yazd
University, Pejoohesh Street Safa-ieh, P.O.BOX 89195-741, Yazd, Iran
2484
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

A large number of articles have been published on for steel dimensioning of a one-way slab, simply
optimization of structures. In the great majority, the supported on one side and fixed on the other side are
objective is to minimize the weight of the structure reported [17].
[3, 4]. While weight of a structure constitutes a Recently, few articles based on concrete codes of
significant part of the cost, a minimum weight design is practice (i.e ACI) have been published. Sahab et al.
not necessarily the minimum cost design. (2005a, 2005b) presented a hybrid method based on
Especially, in optimal design reinforce concrete genetic algorithm for cost optimization of reinforced
elements and structures, the minimum weight design is concrete flat slab buildings according to the Britis h
not always the cheapest design. Minimizing the cost is Code of Practice (BS8110) [11, 12]. Ahmadkhanlou
more realistic and a more useful goal for saving natural and Adeli (2005) used a neural dynamics model for
resources. optimal cost design of reinforced concrete slabs
For a more widespread use of structural according to ACI, 1999 code privisions. They
optimization algorithms in design office practice, they formulated the optimization problem as a mixed
must be formulated as cost optimization and applied to integer-discrete variable optimization problem with
realistic structures subjected to the actual constraints of three design variables: thickness of slab, steel bar
commonly used design codes such as the American diameter and bar spacing [10].
Concrete Institute Code (ACI 318-M08) [5]. In this article, optimal cost design of one way slabs
State-of-the-art reviews of articles on cost based on the most recent ACI code (ACI 318-M08) in
optimization of concrete structures are presented in [6]. investigated. Particle swarm optimization, one of the
Recent examples of cost optimization of real life most recent evolutionary algorithms is used for this
structures subjected to design code constraints can be purpose. The constrained optimization problem is
found in [7-13]. tackled through the minimization of a non-stationary
The early work on optimization of Reinforced multi-stage assignment penalty function. Cost
Concrete (RC) slabs was based on many simplifying optimization of four different slabs with different
assumptions. Brown (1975) formulates the optimum support conditions are illustrated and the results of the
cost design of one-way concrete slabs as a single- optimum design results are compared with existing
variable optimization problem to find the optimum methods in the literature. A sensitivity analysis of
thickness of one-way slabs for uniformly loaded simply optimized designs was also performed by optimizing
supported slabs considering flexural deformations only the four examples for different span lengths.
and making other simplifying assumptions [14]. The reminder of the article is organized as follows:
Brondum-Nielsen (1985) presents a method for The problem formulation is given in Section 2. Section
minimizing the cost of reinforcement in reinforced 3 describes the PSO algorithm and its discrete version.
concrete shells, folded plates, walls and slabs by Section 4 contains four illustrative examples of one-
minimizing the summation of the forces in the steel way slabs with different support conditions and in
reinforcement in two perpendicular directions. An Section 5 the summary and conclusions are presented.
academic example is presented without considering any
code of practice in defining the constraints [15]. FORMULATION OF THE
Hanna and Senouci (1995) present an iterative MINIMUM COST DESIGN
procedure for minimum cost design of all-wood
concrete slab forms. Four design variables: type and Cost function: In concrete structures, at least three
thickness of sheathing, size and spacing of joists, type different cost items should be considered in
and size of stringers and the type of the wood shore optimization: costs of concrete, steel and the formwork.
were considered in the formulation. No formal The total cost function can be defined as
mathematical optimization is used. They report cost
savings as high as about 10% compared with traditional (1)
methods based on the use of charts and tables [16].
Tabatabai and Mosalam (2001) integrated three where CC, Cf and Cf are the costs of concrete,
computer programs: a commercial program for reinforcement bars and formwork and finishing
nonlinear analysis of structures, a program for finite materials, respectively. The formwork cost does not
element discretization of reinforced concrete structures vary significantly for any given locality and
and minimization of its reinforcement content and a consequently can be dropped from the formulation [10].
program for data commu nication between the first The concrete cost is defined by:
two programs for determination of the optimum
reinforcement in beams and one-way slabs. The results (2)
2485
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

Table 1: Moment coefficient of continuous slabs

Exterior span Interior span
-------------------------------------- ----------------------------------------
Support Middle Support Support Middle Support
-1/24 +1/14 -1/10 -1/11 +1/16 -1/11

Table 2: Maximum moment coefficient, k, used for design of RC

slabs
Simply One end Both ends
Fig. 1: Typical cross-section of RC slab
supported continuous continuous Cantilever

where L, b, h and C1C are the span length, span width, 1/8 1/10 1/11 1/2

thickness of slab (Fig. 1) and cost of concrete per unit

volume, respectively. The reinforcement cost is maximum value of the moment coefficient for any
computed as: given span is used, for four different support
conditions: simply-supported, continuous in one end
(3) and simply-supported at the other end, continuous at
both ends and cantilever (Table 2). In Eq. (6), w the
where ws , A s and C1r are the unit weight of steel, cross factored uniformly distributed load including the dead
load and the self-weight of slab is calculated as:
sectional area of reinforcement bars and cost of
reinforcement bars per unit weight. A s is calculated by:
(7)

where DL and LL are dead load of floor excluding the

(4) self-weight of slab and the live load, respectively. DLs
represents the self-weight of slab and is equal to:
where db and s are diameter of reinforcement bars and
their spacing, respectively. (8)

Constraints: All the design constraints imposed by the where wc is the weight of concrete per unit volume.
most recent ACI code (ACI 318-M08) are considered. For design, the ACI code allows the use of an
The constraints include flexural constraints, shear equivalent rectangular compressive stress distribution
constraints, serviceability constraints and deflection (stress block) to replace the more exact concrete stress
constraints. distribution. In the equivalent rectangular stress block,
an average stress of 0.85f′ c is used with a rectangle of
Flexural constraint: Flexural resistance of the slab depth a = β1 c where β1 is defined as follows:
must be greater the implied moments by the loads. This
constraint is presented in the following form:

(5)

where M u and Mn are ultimate design moment and (9)

nominal bending moment, respectively. The strength
The nominal bending moment, M n , is then
reduction factor f is calculated based on net tensile
calculated as follows:
strain in in the extreme tension steel at nominal strength
and is varied in the range of 0.65 to 0.9 for compression
controlled or tension controlled state (10)
The ultimate design moment is calculated as follows:
where fy is the specified yield strength of reinforcement
(6) bars and a is the equivalent depth of the concrete
compression stress block calculated from (Fig. 1)
where ln is the clear span length and k is a moment
coefficient that depends on the type of slab supports for
continuous slabs as presented in Table 1. In Eq. (6), the (11)
2486
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

Table 3: Minimum thickness of solid one-way slabs according to ACI

Minimum thickness, h
------------------------------------------------------------------------------------------------------------------------------------------
Simply supported One end continuous Both end continuous Cantilever
Member Members not supporting or attached to partitions or other construction likely to be damaged by large deflections
Solid one way slabs L/20 L/24 L/28 L/10
Notes: Values given shall be used directly for members with normal weight concrete and grade 420 reinforcement. for other conditions, the
values shall be modified as follows:
For lightweight Concrete having equilibrium density, wc, In the range of 1440 to 1840 kg/m 3 , the values shall be modified by Eq. (20) but not less
than 1.09
For fy other than 420 MPa, the values shall be multiplied by Eq. (21)

where f′ c is the specified compressive strength of (a) Slabs where Grade 280 or 350 deformed bars are
concrete. used 0.0020
(b) Slabs where Grade 420 deformed bars or welded
Shear constraint: The shear constraint is represnted in wire reinforcement are used..0.0018
the following form: (c) Slabs where reinforcement with yield stress
exceeding 420 MPa measured at a yield strain of
(12) 0.35 percent is used 0.0018 × 420/ fy

where Vu and Vc are the ultimate factored shear force In slabs, primary flexural reinforcement shall
and the nominal shear strength of concrete. The shear is not be spaced farther apart than three times the slab
carried entirely by concrete since no stirrup is used. The thickness, nor farther apart than 450 mm.
ultimate factored shear force and the nominal shear
strength of concrete are given by: (16)

ACI 318-M08 also specifies that the minimum

(13) clear spacing between parallel bars in a layer shall be
d b , but not less than 25 mm.

(14) (17)

kv is a number specified by ACI with values of 0.5, Deflection constraints: The ACI 318-M08 specifies a
0.575, 0.575 and 1.0 for examples 1 to 4 respectively. minimum slab thickness (h min) of L/20, L/24, L/28, or
L/10 for different support conditions (Table 3), with an
Serviceability constraints: The serviceability absolute minimum thickness of 1.5 inch (38.1 mm). In
constraints are presented in terms of limits on the steel order to take into account the effect of the weight of
reinforcement ratio and the bar spacing. concrete and the reinforcement yield strength, the
ACI 318-M08 requires that the net tensile strain numbers in Table 3 must be multiplied by the following
et at nominal strength of non-prestressed flexural modification factors:
members and non-prestressed members with factored
axial compressive load less than 0.10fcAg, shall not be (18)
less than 0.004.

(19)

(15) Constraint normalization: In numerical calculations,

it is desirable to normalize all the constraint functions
ACI also specifies a minimum amount of [1]. This normalization speeds up the convergence and
reinforcement and requires that for structural slabs, prevents undue dominance of any particular constraint
minimum As in the direction of the span shall provide since different constraints involve different orders of
at least the following ratios of reinforcement area to magnitude. The normalized constraints are introduced
gross concrete area, but not less than 0.0014: by the following equations:
2487
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

ACI provides eleven different bar sizes starting from

bar size #3 with diameter of 0.375 inch (0.953 cm) to
bar size #18 with diameter of 2.257 inch (5.733 cm).
(20) Therefore, the problem is formulated as a discrete
nonlinear programming problem expressed as:

minimize f(x)
(21)
subjected to the following constraints:

(22)
(29)
(23)
where nd is the total number of discrete design variables
(equal to three in this article) and Di is the set of
discrete values for the ith variable.
(24)
Particle swarm optimization: The PSO algorithm was
first proposed in 1995 by Kennedy and Eberhart. It is
(25)
based on the premise that social sharing of information
among members of a species offers an evolutionary
advantage [18]. Recently, the PSO has been
(26)
applied and proven useful on a few structural
engineering applications such as optimal truss
design [19, 20], structural damage detection [21]
(27)
among others. A number of advantages with respect to
other Evolutionary algorithms make PSO an ideal
OPTIMIZATION PROBLEM FORMULATION
AND SOLUTION APPROACH candidate for engineering optimization problems.
The algorithm is robust and well suited to handle
Discrete optimization formulation: First, a continuous non-linear, non-convex design spaces with
variable optimization problem is defined as: discontinuities. Furthermore, its easiness of
implementation makes it more attractive as it does
minimize f(x) = Ct not require specific domain knowledge information,
internal transformation of variables or other
subjected to the following inequality constraints: manipulations to handle constraints [22].
In PSO, a number of simple entities (the particles)
(28) are placed in the search space of some problem or
function and each particle evaluates the objective
where x is the vector of continuous design variables, function at its current location. Each particle then
f(x) is the cost function defined by Eq. (1) and m is the determines its movement through the search space by
number of inequality constraints (equal to 8 in this combining some aspect of the history of its own current
problem, as defined by Eqs. 20-27). Three design and best (best-fitness) locations with those of one or
variable are considered in the problem definition: the more members of the swarm, with some random
thickness of slab (h), the diameter of reinforcement bars perturbations. The next iteration takes place after all
(d b ) and the spacing of reinforcement bars (s). The particles have been re-located. Eventually the swarm as
thickness of the concrete slab and spacing of the a whole, like a flock of birds collectively foraging for
reinforcement bars can be considered as discrete food, is likely to move close to an optimum of the
variables, considering the common practice of using fitness function.
multiple integers of centimeters in the SI system, or a Each individual in the particle swarm is composed
multiple of 1/8 in or 1/4 in inch the US customary of three D-dimensional vectors: the current position ,
system. The bar diameters are treated as discrete the previous best position and the velocity where D
variables as their values must be assigned from a is the dimensionality of the search space (i.e. number of
limited number of commercially available bar sizes. design variables).
2488
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

The current position can be considered as a annealing scheme was used for the setting of the
set of coordinates describing a point in space. On each parameter ω, where ω decreases linearly from ωi to ωf
iteration of the algorithm, the current position is over the whole run
evaluated as a problem solution. If that position is better
than any that has been found so far, then the
coordinates are stored in the second vector, . The (34)
value of the best function result so far is stored in a
variable often called pbesti (for “previous best”), for where ωi and ωf are the values of parameter ω at the
comparison on later iterations. The objective is to keep start and end of the search., iter is the current iteration
finding better positions and updating and pbesti . New number and MAXITER is the number of maximum
points are chosen by adding coordinates to and the allowable iterations.
algorithm operates by adjusting , which can
effectively be seen as a step size [23]. The velocity of The constraint handling approach: Different
each particle is iteratively adjusted so that the particle constraint-handling techniques have been used over the
performs a stochastically oscillation around and years to handle linear and nonlinear inequality
constraints in evolutionary algorithms. An excellent
locations. The new velocity of each particle is
survey on constraint handling techniques is written by
calculated as follows:
Coello [26].
The search space in constrained optimization
problems consists of feasible and infeasible points. In
feasible points all the constraints are met. In contrast, in
(32) infeasible points at least one of constraints is violated.
The most common constraint-handling approach is the
where represents the current velocity of a design use of a penalty function for penalizing infeasible
variable. The superscript d stands for the dth particle, points. In this approach, the constrained problem is
the subscript i indicates the ith design variable and t is transformed to an unconstrained one, by penalizing the
the iteration number. c1 and c2 are two positive infeasible points and building a single objective
constants called acceleration coefficients, ω is the function, which in turn is minimized using an
inertia factor and r1 and r2 are two independent random unconstrained optimization algorithm
numbers uniformly distributed in the range of [0, 1]. Penalty functions can be categorized into two main
After the velocity is updated, the new position of each divisions: stationary and non-stationary. Stationary or
particle for the next generation is calculated according static penalty functions use fixed penalty values
to the following equation: throughout the minimization, where in contrast, in non-
stationary penalty functions, the penalty values are
(33) dynamically modified. In the literature, results obtained
using non-stationary penalty functions are almost
The particle is then evaluated according to its new always superior to those obtained through stationary
position and and are updated at each generation. functions [27, 28].
A penalty function can be defined as:
This process is repeated until a user-defined stopping
criterion is reached.
(35)
The original PSO algorithm is designed for
optimization problems with continuous variables. As
where f(x) is the original objective function; h(k) is a
design variables of slab optimization problems are
dynamically modified penalty value, k is the algorithm
discrete, we used An approach for tackling discrete
current iteration number; and H(k) is a penalty factor,
optimization problems by PSO which is based on
defined as:
the truncation of the real values to their nearest
integer [24, 25].
The original PSO algorithm lacks exploitation and
is generally slow at late stages of optimization. An (36)
improved version of PSO algorithms is adapted in this where
article. The objective of this modification is to try to
avoid premature convergence in the early stages
of the search and to facilitate convergence to the The function qi (x) is a relative violated function of
global optima in the the final stages of the search. An the constraints; θ(q i (x)) is a multi-segment assignment
2489
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

function; γ(q i (x)) is a power of the penalty function; and Table 4: Common data used in design examples
g i (x) are the constraint functions. fy 40 ksi (275.8 Mpa) b 1 ft (0.3048 m)
ws 490 Ib/ft 3 (77 KN/m 3 ) L 13 ft (3.96 m)
RESULTS
f′c 3 ksi (20.68 Mpa) DL 10 Ib/ft 2 (0.48 KN/m 2 )
Illustrative design examples: Four examples of one- wc 150 Ib/ft 3 (23.6 KN/m 3 ) LL 40 Ib/ft 2 (2.39 KN/m 2 )
way reinforced concrete slabs with different support Cover 3/4 in (19.05 mm) C r $1300/Ton (short)($1.43/kg)
1

conditions are optimized in this section. Example 1 is a

simply supported slab at both ends. Example 2 has a
simple support at one end and a continuous support at Table 5: Concrete cost, C 1c , according to its specified compressive
the other end. Exa mple 3 it is part of a multi-space strength
reinforced concrete slabs and therefore can be f′c (psi) C c ($/cyb)
1

considered continuous at both ends (). Example 4 is a

one-way cantilever reinforced concrete slab. 2000 (14 Mpa) 71.5
The common data for the examples are presented 2500 (17 Mpa) 64.0
in Table 4. These examples were previously optimized 3000 (21 Mpa) 76.0
with neural dynamic model using an earlier version of 3500 (24 Mpa) 78.0
ACI code (1999) edition as the design code [10]. The 4000 (28 Mpa) 81.5
cost of reinforcement steel, C1r is set to $1.43/Kg. The 4500 (31 Mpa) 83.0
cost of concrete is a function of concrete strength as 5000 (34 Mpa) 84.5
noted in Table 5, per Means [29]. For variables h and s, 6000 (41 Mpa) 96.5
practical values are assumed to be a multiple of 1/4 and 8000 (55 Mpa) 158.0
1/2 inch respectively. 10000 (17 Mpa) 224.0
Penalty parameters of equation 34 are the same 1.0 $/cyb = 0.122 $/m 3
values reported in [27]. Specifically, if qi (x)<1,
then γ((q i (x)) = 1, otherwise γ(q i (x)) = 2. Moreover, It should be noted that ultimate bending moments is
if q i (x))<0.001 then θ(q i (x)) = 10, else, if also much higher in example 4 (Table 2).
0.001<q i (x)<0.1 then θ(q i (x)) = 20, else, if qi (x)<1 The values of normalized constraint functions for
then θ(q i (x)) = 100, otherwise θ(q i (x)) = 300. h(k), the four examples are presented in Table 7. It can be noted
dynamically modified penalty value in equation 36, that all constraint function values are negative and
was set to therefore all constraints are met at optimal points. In all
Following parameters were used in the PSO: c1 = examples, the constraints representing moment capacity
c2 = 2. An annealing scheme was used for the ω-setting (g 1 ) and minimum slab thickness (g 7 ) are active. For
of the PSO, where ω decreases linearly from ωi = 0.9 to examples, 2 and 3, (g 8 ) is also active as a minimum bar
ωf = 0.4 over the whole run. The size of the swarm was size reinforcement is used in the optimal designs. It
set equal to 30 The PSO algorithm ran for 1000 should be noted that small deviations of active
iterations for each example and 30 runs were performed constraints from zero are due to existence of discrete
for each example problem. variables.
The optimum values for these examples and
associated costs are shown in Table 6. Results of neural Parametric study: This section describes the
dynamic model on the same examples using an earlier parametric study of one-way slabs with different
version of ACI code are also presented. The direct support conditions for various practical span lengths.
comparison of results is difficult as the optimal designs The goal of the parametric study was to investigate the
are based of different ACI code versions; however, it effect of slab span length on optimal design variables,
can be noted that the values obtained by PSO algorithm the cost components and optimal rein forcement ratio ρ,
are generally very similar to neural dynamics for all ratio of A s to bd. Four previously described examples
examples are optimized with different span lengths, ranging from
It can be seen that example 4 (cantilever slab) has 7 to 16 ft (2 to 5 meters), using the same procedure
the maximum cost (59.31$) among all examples. This is previously mentioned.
expected as the required minimum thickness for Results of corresponding optimal values of design
deflection control of cantilever slab is 1/10 of the span variables, concrete and reinforcement costs and total
length which is much higher than minimum thickness slab cost are presented at Table 8. Total cost obtained
requirements of other examples: 1/20, 1/24 and 1.28 of for four examples for different span lengths are
span lengths for examples 1 to 3 respectively (Table 3). presented in Fig. 2. Cantilever slab (example 4) has the
2490
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

Table 6: Cost optimization results for examples 1-4 (length = 13 ft)

Ahmadkhanlou & Adeli [10] This paper
---------------------------------------------------------------------------------------------------------------------------------------------
h (in) db (in) s (in) Total cost ($) h (in) db (in) s (in) Total cost ($)
Example 1 6.75 3/8 6.5 26.45 6.25 1/2 9.00 26.57
Example 2 5.75 3/8 7.0 22.98 5.25 3/8 5.50 22.76
Example 3 4.75 3/8 7.0 19.93 4.50 3/8 5.50 20.64
Example 4 13.50 3/8 2.0 60.22 12.50 5/8 12.50 59.31
1 in = 25.4 mm

Table 7: Normalized constraints for four examples (length = 13 ft)

Normalized constraints
-----------------------------------------------------------------------------------------------------------------------------------------------
g1 (Eq. 22) g2 (Eq. 23) g3 (Eq. 24) g4 (Eq. 25) g5 (Eq. 26) g6 (Eq. 27)6 g7 (Eq. 28) g8 (Eq. 29)
Example 1 -0.012 -0.771 -0.889 -0.745 -0.492 -5.064 -0.009 -0.333
Example 2 -0.035 -0.706 -0.875 -0.913 -0.651 -3.046 -0.017 0.000
Example 3 0.000 -0.667 -0.845 -1.231 -0.593 -3.046 -0.017 0.000
Example 4 -0.019 -0.683 -0.853 -1.462 -0.718 -2.105 -0.009 -0.669

Fig. 2: Optimal total cost of four examples for different Fig. 3: Optimal reinforcement ratios of four examples
span lengths for different span lengths

maximum total cost among all examples. Among the that cantilever slabs are rarely used in high span
remaining examples which are also more common and lengths and for most practical span lengths (7-10 ft),
practical slab configurations (examples 1-3), the simply the optimum ratios are in the very tight range of
supported slab has the maximum cost. It can be seen 0.43% to 0.47%.
that as expected in all examples the cost increases at Adeli (2001) used a neural dynamics model for
higher span lengths. presented examples. In their work, they obtained the
For different span lengths, optimum reinforcement solution in two stages. In the first stage, the neural
ratios corresponding to minimum cost designs of dynamics model was used to obtain an optimum
four examples are shown in Table 8 and Fig. 3. solution assuming continuous variables. In order to find
Optimum ratios for example 1 are very close to 0.4%. practical discrete values for the design variables, the
verying between 0.38% and 0.44%. For example 2, second stage of optimization was performed by
the optimum ratios vary between 0.44% and 0.63%. formulating the problem as a mixed integer-discrete
It can be observed that for most span lengths (9 ft optimization problem.
to 16 ft), the ratios are in the tight range of 0.44% to The main goal of this article has been to present a
0.48%. The optimum reinforcement ratios of simple algorithm that can be efficiently used in optimal
example 3 vary between 0.55% and 0.94%. For most design of engineering problems. In the approach
span lengths (9-16 ft), the ratios are in the very presented in this article, the optimal values of design
tight range of 0.55% to 0.60%. For the cantilever variables are obtained directly and in a single step,
slab (example 4), the optimum reinforcement ratios which in turn simplify the design for practicing
vary between 0.44% and 0.63%. It should be noted engineers.
2491
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

Table 8: Cost optimization results for examples 1-4 for different lengths (Span width = 1 ft)
Span Cost of Cost of Total Reinforcement
length (ft) h (in) s (in) db (in) concrete (CC) reinforcement (Cr) cost ($) Ratio (%)
Ex.1 7 3.50 10.5 3/8 5.75 1.95 7.70 0.41%
8 4.00 9.5 3/8 7.51 2.46 9.97 0.38%
9 4.50 8.0 3/8 9.50 3.29 12.79 0.39%
10 5.00 13.0 1/2 11.73 4.00 15.73 0.38%
11 5.25 11.0 1/2 13.55 5.20 18.75 0.42%
12 5.75 10.0 1/2 16.19 6.24 22.43 0.41%
13 6.25 9.0 1/2 19.06 7.51 26.57 0.42%
14 6.75 8.0 1/2 22.17 9.10 31.27 0.43%
15 7.25 7.5 1/2 25.51 10.40 35.91 0.42%
16 7.75 10.5 5/8 29.09 12.42 41.51 0.44%
Ex 2 7 3.00 8.5 3/8 4.93 2.41 7.34 0.63%
8 3.25 9.5 3/8 6.10 2.46 8.56 0.50%
9 3.75 8.5 3/8 7.92 3.10 11.01 0.46%
10 4.00 7.5 3/8 9.38 3.90 13.28 0.48%
11 4.50 7.0 3/8 11.61 4.60 16.21 0.44%
12 5.00 11.0 1/2 14.07 5.67 19.75 0.45%
13 5.25 5.5 3/8 16.01 6.91 22.92 0.47%
14 5.75 9.0 1/2 18.88 8.09 26.97 0.46%
15 6.00 4.5 3/8 21.11 9.75 30.86 0.48%
16 6.50 7.5 1/2 24.40 11.09 35.49 0.48%
Ex 3 7 2.50 7.5 3/8 4.10 2.73 6.84 0.94%
8 2.75 8.0 3/8 5.16 2.93 8.09 0.76%
9 3.25 8.0 3/8 6.86 3.29 10.15 0.60%
10 3.50 7.0 3/8 8.21 4.18 12.39 0.62%
11 3.75 6.5 3/8 9.68 4.95 14.63 0.60%
12 4.25 6.0 3/8 11.96 5.85 17.81 0.56%
13 4.50 5.5 3/8 13.72 6.91 20.64 0.56%
14 5.00 9.0 1/2 16.42 8.09 24.51 0.55%
15 5.25 4.5 3/8 18.47 9.75 28.22 0.57%
16 5.50 4.0 3/8 20.64 11.70 32.34 0.61%
Ex 4 7 6.75 8.0 1/2 11.08 4.55 15.63 0.43%
8 7.75 10.5 5/8 14.54 6.21 20.75 0.44%
9 8.75 9.0 5/8 18.47 8.15 26.62 0.44%
10 9.75 11.0 3/4 22.87 10.69 33.56 0.47%
11 10.50 6.5 5/8 27.09 13.79 40.89 0.50%
12 11.50 11.0 7/8 32.37 17.34 49.71 0.53%
13 12.50 5.0 5/8 38.12 21.19 59.31 0.54%
14 13.50 6.5 3/4 44.33 25.34 69.67 0.55%
15 14.50 4.0 5/8 51.02 30.57 81.59 0.57%
16 15.25 11.5 9/8 57.23 36.95 94.19 0.63%

SUMMARY AND CONCLUSIONS requirements are used to consider all imposed design
constraints including criteria for strength, ductility and
This article presents the cost optimization of serviceability among others. a multi-stage dynamic
one-way slabs with different support conditions penalty was also implemented to solve the constrained
using the PSO algorithm, one of the most recent optimization problem
evolutionary algorithms. The total cost of the slab was The optimization results of four different slab
used as the objective function and the ACI-M08 design configurations demonstrate that PSO is a promising
2492
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

method in design optimization of one-way reinforce 12. Sahab, M.G., A.F. Ashour and V.V. Toropov,
concrete slabs. It should be noted that the presented 2005. A hybrid genetic algorithm for reinforced
results are obtained based on prices in U.S. Obviously concrete flat slab buildings. Computers and
the results depends of on the relative cost of concrete Structures, 83: 551-559.
and reinforcement and therefore are location dependant. 13. Atabay, S., 2009. Cost optimization of three-
However, the presented algorithm can be applied in dimensional beamless reinforced concrete shear-
design offices in any location by using the relevant unit wall systems via genetic algorithm. Expert Systems
costs of concrete and reinforcement. This, in turn, and Applications, 36: 3555-3561.
generally reduces the cost of the construction and saves 14. Brown, R.H., 1975. Minimum cost selection of
the natural resources. one-way slab thickness. Journal of the Structural
Division, ASCE, 101: 2585-2590.
REFERENCES
15. Brondum-Nielsen, T., 1985. Optimization of
reinforcement in shells, folded plates, walls and
1. Arora, J.S., 2004. Introduction to optimum design.
slabs. Proceedings of American Concrete Institute,
2nd Edn. Elsevier.
82 (3): 304-309.
2. González-Vidosa, F., V. Yepes, J. Alcalá, M.
Carrera, C. Perea and I. Payá-Zaforteza, 2008. 16. Hanna, A.S. and A.B. Senouci, 1995. Design
Optimization of Reinforced Concrete Structures optimization of concrete slab forms. Journal of
by Simulated Annealing. In: Cher Ming Tan. Construction Engineering and Management,
Editor. Simulated Annealing, Vienna, Austria, ASCE, 121(2): 215-221.
pp: 307-320. 17. Tabatabai, S.M.R. and K.M. Mosalam, 2001.
3. Pezeshk, S., 1998. Design of framed structures: An Computational platform for non-linear
integrated non-linear analysis and optimal analysis/optimal design of reinforced concrete
minimum weight design. International Journal of structures. Engineering Computations, 18 (5/6):
Numerical Methods in Engineering, 41: 459-471. 726-743.
4. Chung, T.T. and T.C. Sun, 1994. Optimization for 18. Kennedy, J. and R. Eberhart, 1995. Particle
flexural reinforced concrete beams with static swarm optimization. In: IEEE international
nonlinear response. Structural Optimization, conference on neural networks, Piscataway, NJ,
8: 174-180. 4: 1942-1948.
5. Adeli, H. and K.C. Sarma, 2006. Cost optimization 19. Perez, R.E. and K. Behdinan, 2007. Particle swarm
of structures: Fuzzy logic, genetic algorithms and approach for structural design optimization.
parallel computing. John Wiley and Sons. Ltd. Computers and Structures, 85: 1579-1588.
6. Sarma, K. and H. Adeli, 1998. Cost optimization of 20. Li, L.J., Z.B. Huang and F. Liu, 2009. A heuristic
concrete structures. Journal of Structural particle swarm optimization method for truss
Engineering, ASCE, 124 (5): 570-578. structures with discrete variables. Computers and
7. Sarma, K. and H. Adeli, 2000. Cost optimization Structures, 87: 435-443.
of steel structures. Engineering Optimization, 21. Begambre, O. and J.E. Laier, 2009. A hybrid
32 (6): 777-802. Particle Swarm Optimization-Simplex algorithm
8. Sarma, K. and H. Adeli, 2001. Bi-level parallel (PSOS) for structural damage identification.
genetic algorithms for optimization of large steel
Advances in Engineering Software, 40: 883-891.
structures. Compututer-Aided Civil and
22. Perez, R.E. and K. Behdinan, 2007. Particle swarm
Infrastructure Engineering, 16 (5): 295-304.
optimization in structural design. In F.T.S. Chan
9. Adeli, H. and H. Kim, 2001. Cost optimization of
and M.K. Tiwari Eds. Swarm Intelligence: Focus
composite floors using the neural dynamics
on Ant and Particle Swarm Optimization, Itech
model. Communications in Numerical Methods in
Education and Publishing, Vienna, Austria,
Engineering, 17: 771-787.
10. Ahmadkhanlou, F. and H. Adeli, 2005. Optimum pp: 373-394
cost design of reinforced concrete slabs using 23. Poli, R., J. Kennedy and T. Blackwell, 2007.
neural dynamic model. Engineering Applications Particle swarm optimization: An overview. Swarm
of Artificial Intelligence, 18: 65-72. Intelligence, 1: 33-57.
11. Sahab, M.G., A.F. Ashour and V.V. Toropov, 24. Husseinzadeh Kashan, A. and B. Karimi, 2009. A
2005. Cost optimization of reinforced concrete discrete particle swarm optimization algorithm
flat slab buildings. Engineering Structures, for scheduling parallel machines. Computers and
27: 313-322. Industrial Engineering, 56: 216-223.
2493
 World Appl. Sci. J., 13 (12): 2484-2494, 2011

25. Laskari, E.C., K.E. Parsopoulos and M.N. Vrahatis, 28. Joines, J.A. and C.R. Houck, 1994. On the Use of
2002. Particle swarm optimization for integer Non-Stationary Penalty Functions to Solve
programming. Proceedings of the IEEE Congress Nonlinear Constrained Optimization Problems
on Evolutionary Computation, Honolulu, pp: with GA's. Proceeding of IEEE International
1582-1587. Conference on Evolutionary Computations,
26. Coello, C.A.C., 2002. Theoretical and numerical pp: 579-585.
constraint-handling techniques used with 29. Means, Co, 2002. Means Building Construction
evolutionary algorithms: A survey of the state of Cost Data. Kingston, MA.
the art. Computer Methods in Applied Mechanics 30. Shi, Y. and R.C. Eberhart, 2001. Fuzzy Adaptive
and Engineering, 191: 1245-1287. Particle Swarm Optimization. Proceedings of the
27. Parsopoulos, K.E. and M.N. Verahatis, 2002. Congress on Evolutionary Computation, Seoul,
Particle swarm optimization method for Korea, pp: 101-106.
constrained optimization problems. In Sincak,
P., J. Vascak, V. Kvasnicka and J. Pospichal, Eds.
Intelligent Technologies -Theory and Application:
New Trends in Intelligent Technologies. Frontiers
in Artificial Intelligence and Applications, IOS
Press, 76: 214-20.

2494