Arab J Sci Eng (2011) 36:1179–1187

DOI 10.1007/s13369-011-0113-1

R E S E A R C H A RT I C L E - C I V I L E N G I N E E R I N G

A. Kaveh · A. Shakouri Mahmud Abadi

Cost Optimization of Reinforced Concrete One-Way Ribbed

Slabs Using Harmony Search Algorithm

Received: 16 July 2009 / Accepted: 25 May 2010 / Published online: 19 October 2011
© King Fahd University of Petroleum and Minerals 2011

Abstract Cost optimization of a reinforced concrete one-way joist floor system consisting of a hollow slab
is presented in this paper. The cost of the system is considered to be the objective function, and the design
is based on the American Concrete Institute’s ACI 318-05 standard. This function is minimized, subject to
design constraints, using the harmony search algorithm. A numerical example is presented to illustrate the
performance of the algorithm and a sensitivity analysis is performed. A parametric study is also conducted to
investigate the effects of the beam span and the loading on the cost optimization of the ribbed slab.
Keywords Reinforced concrete slab · One-way joist floor · Harmony search algorithm · Cost optimization

1 Introduction

A one-way joist floor system consists of hollow slabs with a total depth greater than that of solid slabs. The
system is the most economical for buildings where the superimposed loads are small and the spans are rela-
tively large, such as in schools, hospitals, and hotels. The concrete in the tension zone is ineffective; this area
is therefore left open between the ribs or filled with lightweight material to reduce the self-weight of the slab.
Although some mathematical programming based methods have been developed for optimum design prob-
lems, their application has been hindered by requirements for gradient information and the dependency of the
results on the starting node. In recent years, powerful meta-heuristic algorithms have been developed for
structural optimization, each having both advantages and disadvantages. These stochastic search techniques
use ideas adopted from nature, and do not suffer from the difficulties involved in mathematical programming.
A. Kaveh (B)
Centre of Excellence for Fundamental Studies in Structural Engineering,
Iran University of Science and Technology, Narmak, Tehran-16, Iran
E-mail: [email protected]; [email protected]; [email protected]

A. Shakouri Mahmud Abadi

Building and Housing Research Centre, Tehran-14, Iran

123
1180 Arab J Sci Eng (2011) 36:1179–1187

The basic idea behind these techniques is to simulate natural or behavioral phenomena. One of the more
recently developed algorithms of this type is the harmony search algorithm. This approach is based on the pro-
cess in improvisational musical performance that occurs when the musicians seek the best harmony between
themselves. This musical harmony search process is similar to an optimum design process seeking optimum
solutions. The harmony search is also capable of searching from a disjoint feasible domain.
In the process of musical production, a musician selects a number of different notes and then plays them
together to find out whether they give a pleasing harmony. The musician then tunes some of the notes to
achieve a better harmony. The harmony search algorithm works similarly to produce a candidate solution from
the design variables, and then checks to see if this candidate solution improves the objective function. This
candidate solution is then controlled, whether it satisfies the objective function or not, similar to the process of
finding out whether pleasant music has been obtained or not. Applications of the harmony search algorithm in
structural optimization problems are found in the works of Saka [1,2], Saka and Erdal [3], Kaveh and Shakouri
Mahmud Abadi [4,5] and Kaveh and Talatahari [6,7], and in the recent book by Geem [8]. In this study, the
harmony search algorithm is used to determine the optimum design of a reinforced concrete ribbed slab.

2 Design Variables

The six discrete design variables selected for modeling of the ribbed slab are shown in Fig. 1. These include
the thickness of the top slab (X1), the rib spacing (X2), the rib width at the lower end (X3), the rib width at
the top end (X4), the bar diameter (X5), and the rib depth (X6).

3 Optimum Design Process

Conventional design of the ribbed slabs consists of two phases:

1. Checking the dimensions according to the adopted code, e.g.:
• The ribs shall not be less than 100 mm (4 in.) in width and must have a depth of not more than 3.5
times the width;
• Clear spacing between the ribs shall not exceed 750 mm (30 in.);
• The thickness of the top slab shall not be less than 1/12 of the clear span between the ribs or 50 mm
(2 in.).
2. Checking of each component for strength and for steel reinforcement.
The harmony search algorithm, however, initiates the design process by selecting random values for the
thickness of the top slab (X1), the rib spacing (X2), the rib width at the lower end (X3), the rib width at the taper
end (X4), the bar diameter (X5), and the rib depth (X6). The algorithm then checks the dimensions according
to the ACI 318.05 standard, and if these dimensions satisfy the code criteria, the algorithm calculates the
required reinforcement and checks the strength. The harmony search algorithm tries to find the best value for
each design variable to minimize the objective function. The design process consists of six steps as follows:
Step 1. Values are selected for the harmony memory parameters (HMS (harmony memory size), HMCR
(harmony memory considering rate), and PAR (pitch adjusting rate)).

Fig. 1 A ribbed slab and the design variables considered

123
Arab J Sci Eng (2011) 36:1179–1187 1181

Step 2. The harmony memory matrix is initialized; values for the thickness of the top slab (X1), rib spacing
(X2), rib width at the lower end (X3), rib width at the top end (X4), bar diameter (X5), and rib depth (X6)
are chosen.
Step 3. A new harmony vector is improvised using the following rules: (1) random selection; (2) HM
consideration; and (3) pitch adjustment.
Step 4. With the values selected for X1, X2, X3, X4, X5 and X6, the algorithm checks the dimensions
according to the selected code, and if these dimensions satisfy the required criteria, the algorithm calculates
the necessary reinforcement and checks the strength.
Step 5. The value of the objective function is calculated for the newly selected design vector. If this value is
better than the result of the worst harmony vector in the harmony matrix, it is then included in the matrix,
while the worst value is taken out of the matrix.
Step 6. Steps 2 and 6 are repeated until the pre-selected maximum number of iterations is reached.

4 Objective Function

By minimizing a suitable and explicit cost function, one can reach an optimum solution for a concrete ribbed
slab. The optimal design of the concrete ribbed slab is determined by the minimum of the concrete and steel
reinforcement costs. The objective function can be expressed as follows:
Q = Vconc × (C1 + C2 ) + Wsteel × (C3 + C4 ). (1)

By considering 
Q = Q/(C1 + C2 ) we have:
  
 C3 + C4
minimized Q = Vconc + Wsteel /br (2)
C1 + C2
where Vconc and Wsteel are the volume of concrete and the weight of the reinforcement steel in the unit length
(m3 /m, kg/m), respectively; C1 and C3 are the costs of concrete and steel ($/kg for steel and $/m3 for concrete),
respectively; C2 and C4 are the costs of concreting and erecting the reinforcement, respectively. br is the
center-to-center distance of the ribs. Based on reviews and the cost estimation performed, a value of 0.04 for
C3 +C4
the coefficient C(C = ( C +C
)) was obtained.
1 2

5 Design Constraints

The formulation of the design problem is carried out according to the provisions provided in [9].

5.1 Flexural Constraint

The flexural constraint is presented in the following form:

Mu /(φb Mn ) ≤ 1 (3)
where Mu and Mn are the ultimate design moment and the nominal bending moment, respectively.

5.2 Shear Constraint

The shear constraint is presented in the following form:

Vu /(φv Vn ) ≤ 1 (4)
where Vu and Vc are the ultimate factored shear force and the nominal shear strength of the concrete, respec-
tively. Because no stirrup is used in the slab, the concrete should carry the entire shear. The shear strength, Vc ,
provided by the concrete for the ribs may be taken to be 10% greater than that of the beams. This is mainly
because of the interaction between the slab and the closely spaced ribs.

123
1182 Arab J Sci Eng (2011) 36:1179–1187

5.3 Serviceability Constraints

The serviceability constraints are presented in terms of the limits on the steel reinforcement ratio and the bar
spacing. The steel reinforcement ratio should satisfy the following constraint:

ρ ≤ ρmax = 0.75ρb (5)

The minimum shrinkage steel ratio, ρmin , in the slab is 0.002 for slabs in which grade 40 or 50 bars are used
and 0.0018 for slabs in which deformed bars of grade 60 are used. The bar spacing should satisfy the following
constraints:
• The minimum clear spacing between parallel bars in a layer, d b , should not be less than 25 mm.
• The maximum spacing between the bars ≤5 times the rib thickness ≤450 mm (18 in.).

5.4 Deflection Constraints

The thickness of the top slab should not be less than 1/12 of the clear span between the ribs or 50 mm (2 in.).
Based on the ACI code (ACI, 05), a minimum slab thickness hmin of L/16, L/18.5, L/21, or L/8 is required,
depending on the support conditions. Here L is the effective span length of the slab.

5.5 Other Constraints

The ribs should not be less than 100 mm in width, and should have a depth of no more than 3.5 times the
minimum width of the rib. Clear spacing between the ribs should not exceed 750 mm. A limit on the maximum
spacing of the ribs is required because of the special provisions permitting higher shear strengths and lower
concrete protection for the reinforcement of these relatively small repetitive members.

6 Harmony Search Algorithm

The method consists of five basic steps. The detailed explanation of these steps can be found in Lee and Geem
[10–12], and can be summarized as follows:
Step 1. Harmony search parameters are initialized.
Step 2. Harmony memory matrix is initialized.
Step 3. New harmony is improvised.
Step 4. Harmony memory matrix is updated
Step 5. Steps 3 and 4 are repeated until the termination criterion is satisfied.
These steps are described more fully as follows:
Step 1. A possible value range for each design variable of the optimum design problem is specified. A pool
is constructed by collecting these values together, from which the algorithm selects values for the design
variables. The number of solution vectors in the HMS, i.e., the size of the harmony memory matrix, the
HMCR, the PAR and the maximum number of searches are also selected in this step.
Step 2. The harmony memory matrix is initialized. Each row of the harmony memory matrix contains val-
ues of the design variables which are randomly selected as feasible solutions from the design pool for that
particular design variable. Hence, this matrix has ncolumns, where nis the total number of design variables
and HMS rows, which are selected in the first step. HMS is similar to the total number of individuals in the
population matrix of a genetic algorithm. The harmony memory matrix has the following pattern:
⎡ ⎤
x1,1 x2,1 . . . . . . xn−1,1 xn,1
⎢ x1,2 x2,2 . . . . . . xn−1,2 xn,2 ⎥
⎢ ⎥
⎢... ... ... ... ... ... ⎥
[H] = ⎢ ⎥ (6)
⎢... ... ... ... ... ... ⎥
⎣x ⎦
1,hms−1 x2,hms−1 . . . . . . xn−1,hms−1 xn,hms−1
x1,hms x2,hms . . . . . . xn−1,hms xn,hms

123
Arab J Sci Eng (2011) 36:1179–1187 1183

xi,j is the value of the ith design variable in the jth randomly selected feasible solution. It is worth mentioning
that the feasible designs satisfying the constraints are not the only designs that are inserted into the harmony
memory matrix.
Step 3. In generating a new harmony vector, the new value of the ith design variable can be chosen from any
discrete value within the range of ith column of the harmony memory matrix with a probability given by
the HMCR, varying between 0 and 1. In other words, the new value of xi can be one of the discrete values
of the vector {xi,1 , xi,2 , . . . , xi,hms }T with the probability given by the HMCR. The same applies to all of
the design variables. In the random selection, the new value of the ith design variable can also be chosen
randomly from the entire pool with a probability of 1 − HMCR. That is:

{xi,1 , xi,2 , . . . , xi,hms }T with probability H MC R

xinew = (7)
{x1 , x2 , . . . , xns }T with probability (1 − H MC R)

where ns is the total number of values for the design variables in the pool. If the new value of the design
variable is selected from those of the harmony memory matrix, this value is then checked as to whether it
should be pitch adjusted. This operation uses the PAR parameter that sets the rate of adjustment for the pitch
chosen from the harmony memory matrix as follows:

Yes with probability of PAR

Is xnew to be pitch-adjusted? (8)
i No with probability of (1 − PAR)

If we suppose that the new pitch-adjustment decision for xnewi from the test is “yes”, and if the value selected
for xnew
i from the harmony memory is the kth element in the general discrete set, then the neighboring value
i . This operation prevents stagnation and improves the harmony
k + 1 or k-1 is taken for the new value of xnew
memory diversity for a greater chance of reaching the global optimum.
Step 4. After selecting the new values for each design variable, the value of the objective function is cal-
culated for the new harmony vector. If this value is better than the worst harmony vector in the harmony
matrix, it is then included in the matrix, while the worst value is taken out of the matrix.
Step 5. Steps 3 and 4 are repeated until the termination criterion, which is the pre-selected maximum number
of cycles, is reached. This number is selected to be large enough such that within this number of design
cycles, no further improvement is observed in the objective function.

7 Illustrative Example

An example of a one-way RC ribbed slab simply supported at both ends is presented in this section. The
common data for the example is provided in Table 1.
The clear concrete cover is 20 mm. The design variables are shown in Table 2. A schematic view of the
concrete ribbed slab is shown in Fig 2. The harmony search algorithm parameters are taken to be:

HMS = 30, HMCR = 0.85, PAR = 0.45.

The results of the optimum design are provided in Table 3, and the design history convergence is shown
in Fig. 3.

Table 1 Common data as an example

fy 420 MPa
f c 28 MPa
DL 0.78 kN/m2
LL 4 kN/m2
L 6m
Cover 20 mm
ws 78.5 kN/m3
wc 24 kN/m3

123
1184 Arab J Sci Eng (2011) 36:1179–1187

Table 2 Design variables

Value (cm)
Slab thickness 2.5, 5, 7.5, 10
Rib spacing 40, 42.5, 45,…, 72.5, 75
Rib width at lower end 10, 12.5,…, 22.5, 25
Rib width at taper end 10, 12.5,…, 27.5, 30
Bar diameter 1, 1.2, 1.4, 1.6, 1.8, 2
Rib depth 15, 17.5,…, 72.5, 75

Fig. 2 Schematic view of a concrete ribbed slab

Table 3 Results of the optimum design

Slab thickness Rib spacing Rib width Rib width Bar diameter Rib depth Value of objective
at lower end at lower end function
5 cm 60 cm 10 cm 10 cm 1.4 cm 35 cm 1.3626

Fig. 3 History of the convergence for the design

8 Parametric Study

A parametric study is also conducted to investigate the effects of the beam spans and loadings on the cost
optimization of a ribbed slab. Table 4 summarizes the beam spans and the loadings considered in the study.
Table 5 summarizes the design results obtained in the case study using the present model. As expected, the top
slab thickness, the rib depth and the rib width at the lower end all increase with the beam span. Variations of
the total cost and the beam span for three different loadings are shown in Fig. 4. These curves have a non-linear

123
Arab J Sci Eng (2011) 36:1179–1187 1185

Table 4 Parametric study

DL (kN/m2 ) LL (kN/m2 ) Span (m)

7.8 40 4 6 8 10
10 35 4 6 8 10
12 20 4 6 8 10

Table 5 Results of the parametric study

Case DL LL Span Slab Rib Rib Rib Bar Rib Value of Value of
(kN/m)2 (kN)/m2 (m) thickness spacing width width diameter depth objective objective
(cm) (cm) at at (cm) (cm) function function
lower taper /length
end end of span
(cm) (cm)
1–1 7.8 40 4 5 55 10 30 1.6 25 0.2219 0.8876
2–1 6 5 60 10 10 1.4 35 0.2271 1.3626
3–1 8 7.5 67.5 12.5 12.5 1.4 42.5 0.2685 2.148
4–1 10 10 75 15 15 1.6 52.5 0.3446 3.446
1–2 10 35 4 5 60 10 27.5 1.6 25 0.2164 0.8656
2–2 6 5 60 10 10 1.4 32.5 0.2207 1.3242
3–2 8 7.5 72.5 12.5 12.5 1.4 42.5 0.2586 2.0688
4–2 10 10 75 15 15 1.6 52.5 0.3446 3.446
1–3 12 20 4 5 60 10 30 1.6 20 0.1978 0.7912
2–3 6 5 55 10 10 1.2 35 0.2104 1.2624
3–3 8 7.5 75 12.5 12.5 1.4 42.5 0.2541 2.0328
4–3 10 10 75 15 15 1.6 52.5 0.3446 3.446

Fig. 4 Objective function versus span length

Table 6 Polynomial best-fit equations

Case Objective function/span

1 −0.0003L3 + 0.0196L2 − 0.0518L + 0.2544
2 0.0024L3 + 0.0023L2 − 0.0195L + 0.2312
3 0.0026L3 − 0.002L2 − 0.0053L + 0.2006

trend that increases with the beam span. Table 6 summarizes the third-order polynomial fitting between the
beam costs and the spans, which can be used to give an initial estimate of the total cost under a given span
length or given loading combination.
For a span of 10 m, all three cases produce the same optimum design, as shown in Fig. 4. This is because,
based on the ACI code, the clear spacing between the ribs should not exceed 750 mm, and in the 10 m span
cases this constraint bounds the optimum results.

123
1186 Arab J Sci Eng (2011) 36:1179–1187

Table 7 Harmony search parameters used for the sensitivity analysis

Case HMS HMCR PAR

1 25 0.85 0.4
2 25 0.9 0.45
3 30 0.95 0.3
4 30 0.75 0.25
5 35 0.85 0.35

Fig. 5 Design history for cases C1, C2, C3, C4 and C5

9 Sensitivity Analysis for the Harmony Search Parameters

The results of this study show that the harmony search (HS) algorithm can be used efficiently for slab optimi-
zation, provided that a sensitivity analysis is performed to obtain the best values for the HS parameters.
In this section, a sensitivity analysis is performed for the harmony search parameters involved in this study.
The parameters of the sensitivity analyses carried out to determine the appropriate values of the harmony
search parameters are given in Table 7. The results are shown in Fig. 5.
Sensitivity analysis of the HS parameters, consisting of the HMS, HMCR and PAR parameters, is per-
formed to select suitable parameter values. After carrying out a sensitivity analysis with different values of the
HS parameters for five cases, C1 to C5, it is found that the values of 25 for HMS, 0.85 for HMCR and 0.4 for
PAR, correspond to the best performance for this optimization algorithm (case C1). The other cases reached
the optimal solution with higher numbers of iterations.

10 Conclusions

In this paper, optimization is performed using the recently developed harmony search method. The results
obtained show that this method is a powerful and efficient approach for finding the optimum solution to struc-
tural optimization problems. It is found that the newly framed algorithm for the optimal design of a concrete
ribbed slab is quite robust and efficient.
The proposed optimum design model enables structural designers to generate and evaluate optimal or
near-optimal design solutions. The algorithm consists of three modules:
(1) A design module that performs the design of the ribbed slab;
(2) A cost module that computes the total cost of the ribbed slab;
(3) An optimization module that searches for optimal design alternatives.
The main aim of this paper is to present a simple and efficient algorithm which can be used in practical
engineering problems. This simple approach can be used in many other engineering design problems to reduce
the construction costs.

123
Arab J Sci Eng (2011) 36:1179–1187 1187

The optimum cost design of the ribbed slabs from the structural design viewpoint consists of determining
the appropriate thickness of the top slab (X1), the rib spacing (X2), the rib width at the lower end (X3), the
rib width at the top end (X4), the bar diameter (X5) and the rib depth (X6) for a slab. The design is performed
optimally from the available variables list such that with the optimum set of variables, the design of the slab
is within the limitations imposed by the code of practice and at the minimum cost.
The results of parametric study show that the slab thickness, the rib spacing and the rib depth will increase
with increasing span length. For spans with lengths of more than 9 m, the ACI code limitations bound the
optimal results for various loading values and lead to the same result. In spans with lengths of less than 9 m,
changing the loading affects the optimal result and leads to different optimal designs for each loading condition.
The thickness of the top slab (X1) and the rib spacing (X2) have the greatest influence on the cost, and the
bar diameter has the lowest influence on the cost.

Acknowledgments The first author is grateful to the Iran National Science Foundation for support.

References

1. Saka, M.P.: Optimum geometry design of geodesic domes using harmony search algorithm. Adv. Struct. Eng. 10, 595–606
(2007)
2. Saka, M.P.: Optimum design of steel sway frames to BS5950 using harmony search algorithm. J. Construct. Steel Res. 65,
36–43 (2009)
3. Saka, M.P.; Erdal, F.: Harmony search based algorithm for the optimum design of grillage systems to LRFD-AISC. Struct.
Multidiscip. Optim. 38, 25–41 (2009)
4. Kaveh, A.; Shakouri Mahmud Abadi, A.: Cost optimization of composite floor system using an improved harmony search
algorithm. J. Construct. Steel Res. 66, 664–669 (2010)
5. Kaveh, A.; Shakouri Mahmud Abadi, A.: Harmony search algorithm for optimum design of slab formwork. Iranian J. Sci.
Technol. 34, 335–351 (2010)
6. Kaveh, A.; Talatahari, S.: Particle swarm optimizer, ant colony strategy and harmony search scheme hybridized for optimi-
zation of truss structures. Comput. Struct. 87, 267–283 (2009)
7. Kaveh, A.; Talatahari, S.: An efficient hybrid algorithm based on harmony search, particle swarm and ant colony strategies
for optimal design of structures. In: Geem, Z.W. (ed.) Harmony Search Algorithms for Structural Design. Springer, Berlin
(2009)
8. Geem, Z.W.: Harmony Search Algorithms for Structural Design. Springer, Berlin (2009)
9. ACI Committee 318, Building Code Requirements for Structural Concrete (ACI 318-05) and Commentary (318R-05).
American Concrete Institute, Farmington Hills, Mich. (2005)
10. Geem, Z.W.; Kim, J.H.; Loganathan G.V.: A new heuristic optimization algorithm: harmony search. Simulation 76, 60–68
(2001)
11. Lee. K.S.; Geem, Z.W.: A new structural optimization method based on harmony search algorithm. Comput. Struct. 82,
781–798 (2004)
12. Lee, K.S.; Geem, Z.W.: A new meta-heuristic algorithm for continuous engineering optimization: harmony search theory
and practice. Computer Methods Appl Mech Eng. 194, 3902–3933 (2005)

123