KSCE Journal of Civil Engineering (2013) 17(4):769-776 Structural Engineering

DOI 10.1007/s12205-013-0355-4
www.springer.com/12205

Overall Cost Optimization of Prestressed Concrete Bridge

using Genetic Algorithm
Zekeriya Aydin* and Yusuf Ayvaz**
Received November 10, 2011/Revised May 11, 2012/Accepted August 2, 2012

···································································································································································································································

Abstract
Overall cost optimization of Prestressed Concrete (PC) bridges is investigated in this study. The purpose of this study is to
determine the optimum span number and optimum cross-sectional properties of multi-span bridges. Considered bridge superstructure
is constituted by adjacent simply supported pretensioned prestressed I-girders. Also, considered bridge substructure is constituted by
single-column piers and rectangular spread footings. Span number, cross-section dimensions of prestressed girders and the area of
Prestressing Steel (PS) are considered as design variables. PC girders, piers and footings are designed according to AASHTO
Standard Specifications for Highway Bridges. A modified hybrid Genetic Algorithm (GA) is used for the optimum design. Working
stress, ultimate strength, ductility limits, deflection, and geometry constraints are considered. Total cost of the bridge is taken as
optimality criterion. A computer program is coded to perform optimum design and numerical examples from the application are
designed. One of these examples is used in this paper. It is concluded that GA can be effectively used in the overall cost optimization
of PC bridges.
Keywords: prestressed concrete, genetic algorithm, optimization, bridge
···································································································································································································································

1. Introduction tensioned PC I girders and reinforced concrete slab. Cohn and

Lounis (1994), Lounis et al. (1997) studied the cost optimization
Qualified structural design must satisfy required safety and of PC girder bridges. Sirca and Adeli (2005) presented a method
minimum cost. Cost minimization is especially important for for the minimum cost optimization of ptrestressed I girder bridge
large-scale structures. PC girder bridges are generally large-scale superstructures.
constructions. Therefore, cost minimization of PC girder bridges None of the studies above used GA. Optimum design of girder
is an important research area. In fact, all bridge projects are number and girder shape of PC bridge superstructures is realized
designed as minimum-cost in application. But, trial and error using GA by Aydin and Ayvaz (2010). But, number of the piers
method is generally used to design minimum-cost bridge in is not taken as a design variable in that study. On the other hand,
application. Trial and error method requires preliminary study optimum design of bridge longitudinal section, topology of
and experience. On the other hand, optimization methods give bridge cross section, and shape of the girder are realized in one
minimum-cost design exactly and directly in one step. step using GA in this study.
There are studies about shape optimization of PC bridge
superstructures in literature. Totres et al. (1966) realized one of 2. Overall Optimum Design of Pretensioned PC
preliminary studies on PC girder bridges. Cost optimization of I-Girder Bridges
PC highway bridges were performed using linear programming
technique, and number of the girders, height of the girders, In this study, overall cost optimization of PC bridges is
prestressing force, eccentricity of prestressing force were taken investigated. Considered bridge superstructure is constituted by
as design variables in that study. Jones (1985) developed a adjacent simply supported pretensioned prestressed I-shaped
program to design simply supported PC beams as minimum- longitudinal girders. Also, considered bridge substructure is
cost. Yu et al. (1986) presented the optimum design of PC box constituted by H-shaped single piers and rectangular spread
shaped bridge beams using general geometric programming. footings. Such structures are generally used for middle and short
Lounis and Cohn (1993) realized minimum cost design of span bridges. If middle span and short span bridges are constructed
middle and short span highway bridges constituted by post- to cross a valley, the heights of the piers are determined

o
*Assistant Professor, Dept. of Civil Engineering, Namik Kemal University, 59860 Çorlu, Tekirdag, Turkey (Corresponding Author, E-mail: zay-
[email protected])
**Professor, Dept. of Civil Engineering, Yildiz Technical University, 34210 Esenler, I· stanbul, Turkey (E-mail: [email protected])

− 769 −
 Zekeriya Ayd i n and Yusuf Ayvaz

according to shape of the valley. Therefore, shape of the valley is The bridge girders are considered as adjacent. So, slab thickness
considered as a design parameter in this study. The other and slab reinforcement are not taken as a design variable.
important design parameters are the total length of the bridge, the Similarly, cross section properties of the abutments are also not
width of the bridge, live load, material properties and unit prices taken as design variables. Therefore, the costs of the slab and
of the materials. abutments are not considered in the calculation of objective
The purpose of this study is to find out the minimum total cost function.
of the bridge. The total cost of the bridge includes the cost of the
PC girders, the cost of the transverse beams, the cost of the piers, 2.2 Design Variables and Design Parameters
and the cost of the foundations. Optimum span number, optimum Design variables considered in this study are given below:
PC girder number in bridge cross-section, optimum shape of the Number of piers (np)
●

PC girders and optimum number of the PS tendons are Number of PC girders (ng)
●

investigated to minimize the total cost. The total of 33 constraints Height of PC I-girder (h)
●

is considered according to AASHTO Standard Specifications for Bottom flange width (bb)
●

Highway Bridges. Web thickness (bw)

●

Optimum span number, girder number and shape of girders are Top flange thickness (tt)
●

determined in one step. It is very difficult to solve such a Bottom flange thickness (tb)
●

complicated problem with usual optimization techniques. GA is Top sloped thickness (pt)
●

used to solve this optimum design problem. A computer Bottom sloped thickness (pb)
●

program is coded in BASIC to perform the optimum design The number of PS tendons in each girder (nps)
●

mentioned above. Several design examples from application are

A typical cross-section of PC I-girder section is given in
optimized using the computer program coded. One of these
Fig. 1 to demonstrate some of design variables. A typical
examples is used in this paper to demonstrate the efficiency of
longitudinal section of a PC bridge for single span is given in
the coded program.
Fig. 2.
Each engineering optimization problem is determined by
Design parameters considered in this study are the total length
objective function, design variables, design parameters and
of bridge, the width of deck, the width of motorway, the number
constraints. These characteristics of optimization problem
of traffic lanes, the thickness of slab, the distance between top
considered in this study are explained below.
flanges of PC girders, material properties, the unit prices of
2.1 Objective Function
In this study, optimum design criterion is the total cost of the
bridge. It is supposed that the total cost of the bridge includes the
cost of PC girders, cast in place concrete transverse beams, cast
in place concrete piers, cast in place concrete foundations. The
objective is to make the cost minimum without violation of
constraints. So, the objective function of the optimum design
problem can be formulated as:
CT = Cg + Ctb + Cp + Cf (1)
where Cg, Ctb, Cp and Cf are the costs of girders, transverse
beams, piers, and foundations, respectively. These costs are
calculated as a function of prices that are given below: Fig. 1. Typical Cross-section of PC I-girder
Cg = f(uppc , upps) (2)
Ctb = f(upbc, ups) (3)
Cp = f(upbc, ups) (4)
Cf = f(upc , ups , uplc , upf , upex) (5)
where uppc and upps are the unit prices of precast concrete and PS,
respectively, upc and ups are the unit prices of ordinary concrete
and reinforcing steel, upbc is the unit price of cast-in-place
concrete for bridges, uplc, upf, upex are the unit prices of lean
concrete, formwork and excavation, respectively. The unit prices
of precast concrete (uppc) and cast-in-place concrete for bridges Fig. 2. Typical Longitudinal Section of PC Bridges for a Single
(upbc) include the price of formwork. Span

− 770 − KSCE Journal of Civil Engineering

 Overall Cost Optimization of Prestressed Concrete Bridge using Genetic Algorithm

materials, loads, soil bearing capacity, site altitudes, and the other q
g31 = ----------- – 1 (8)
necessary parameters. qallow
Where q is the maximum stress for the soil under the spread
2.3 Constraints
footing, and qallow is the allowable stress of the soil. Ultimate
Constraints for bridge superstructure are investigated according
shear strength constraints and ultimate flexural strength constraints
to AASHTO Standard Specifications (AASHTO, 2002). The
for the footings are given below, respectively.
total of 28 constraints is considered for superstructure. These
constraints can be categorized into six groups as given below: Vu, f
g32 = ----------
-–1 (9)
1) Flexural working stress constraints for girders φ Vc, f
2) Ultimate flexural strength constraints for girders
Mu, f
3) Shear constraints (working stress and ultimate strength) for g33 = -----------
-–1 (10)
φ Mn, f
girders
4) Deflection constraints Where Vu,f, and Mu,f are the factored shear force and moment at
5) Ductility constraints the footings, respectively, Vn,f and Mn,f are the nominal shear
6) Geometrical constraints strength and the nominal moment strength of the footings,
respectively. In the equations given above, φ is the strength
Detailed information about these constrains can be found in the
reduction factor. The value of φ is taken as 0.70 for the Eq. (6),
studies realized by Aydin and Ayvaz (2010), and Aydin (2006).
0.85 for the Eq. (7) and 9, 0.90 for the Eq. (10).
The total of 5 constraints for bridge substructure is also considered
according to AASHTO Standard Specifications (AASHTO, 2002).
3. Optimum Design with Genetic Algorithm
These constraints are given below in addition to the total of 28
constrains mentioned above:
GA is an optimization method that uses the principles of
1) Ultimate flexural strength constraints for piers
Darwin’s natural selection theory (Holland, 1975; Goldberg,
2) Ultimate shear strength constraints for piers
1989). GAs use a group of initial solutions and evaluate them to
3) Soil bearing capacity constraint
reach an optimum solution. These initial solutions are selected as
4) Ultimate shear strength constraints for footings
randomly. Solutions are categorized according to their fitness.
5) Ultimate flexural strength constraints for footings
The fitness of each solution is determined according to its
Cross-section dimensions of piers are not taken as design closeness to the optimum solution. Then, the genetic process
variables. Instead of this, cross-section of pier is selected begins. Reproduction, crossover, and mutation operators are
proportionally with those of application project. Selected cross- applied to the population at each iteration of genetic process.
section is checked according to ultimate flexural strength constraints This process continues until the required convergence is
o
and ultimate shear strength constraints. The cross-section of pier achieved (Goldberg, 1989; Ayvaz and Aydin, 2000; Dalog lu and
is increased proportionally in the case of inadequacy. Ultimate Aydin, 1999).
flexural strength constraints for piers are calculated as: GAs are widely used to solve structural engineering problems.
Several of such studies are realized by Rajeev and Krishnamoorthy
( Pu, Mu )
g29 = ------------------------
-–1 (6) (1998), Jenkins (1992), Rajan (1995), Pezeshk et al. (2000),
ϕ ( Pn, Mn )
Sarma and Adeli (2000), Saka (2007), Kumaran et al. (2010),
where Pu and Mu are the factored axial load and moment at the Degertekin et al. (2008) and Kameshki and Saka (2001).
piers, Pn and Mn are the nominal axial load strength and nominal In this study, a modified hybrid GA is used for the optimum
moment strength of the piers, respectively. Ultimate shear design. Three kinds of crossover and mutation operator are used.
strength constraints for piers can be written as: The major modification in GA is the use of controlled mutation.
The best solutions are reached at less iteration if controlled
Vu
g30 = ----------------------
-–1 (7) mutation is used.
φ ( Vc + Vs )
where Vu is the factored shear force at the pier section, Vc and Vs 3.1 Penalized Objective Function
are the shear strengths of piers provided by the concrete and the GAs are suitable for unconstrained optimization problems.
shear reinforcement, respectively. Therefore, a penalized objective function is used to convert the
Dimensions of spread footing are also not taken as design constrained problem to an unconstrained problem. The penalized
variables. Horizontal dimensions of footings are determined objective function is constituted by adding a penalty function to
according to soil bearing capacity constraint. The thickness of the objective function. So, the penalized objective function can
foundation is determined according to shear strength constraints. be expressed as:
Then, reinforcement of foundation is designed according to
Φ = CT ⋅ [ 1 + K ⋅ P ] (11)
ultimate flexural strength constraints. Thus, soil bearing capacity
constraint can be calculated as: where K is a penalty coefficient that depends on the kind of the

Vol. 17, No. 4 / May 2013 − 771 −

 Zekeriya Ayd i n and Yusuf Ayvaz

problem. It is important to determine suitable value of K to reach penalized objective function among the solution strings in each
the best solutions in a short time. P is the penalty function, and generation, respectively. A fitness factor is described to decide
can be given as: which solutions are removed from generation. The fitness factor
n for ith solution string is calculated as:
P= ∑ pi (12)
Fc, i = Fi ⁄ Fort (15)
i=1

where n is the total number of constraints, pi is the violation where Fort is the average of the fitness values. Thus, the fittest
factor of the ith constraint and can be calculated as: solutions are placed twice instead of the weakest solutions in the
mating pool by reproduction operator. The solutions in the
pi = gi for gi > 0 , pi = 0 for gi ≤ 0 (13)
mating pool are known as parent solutions (Goldberg, 1989;
where gi is the normalized value of ith constraint. Aydin, 2006).

3.2 Coding and Decoding of Solutions 3.4 Crossover Operator

A predetermined probable values set of design variables is The reproduction operator does not create new solutions for
used for optimization in GAs. These sets are named as the design population, but the solutions in the population must be changed
variable values sets. In this study, binary coding technique is and generated to reach an optimum solution. The crossover
used for coding the solutions. Thus, solutions are represented by operator is used to create new solutions that depend on the
strings which are constituted by the numbers “0” and “1”. These existing solutions. The crossover operation is applied by exchanging
strings are obtained by coding row numbers of design variable characters of strings between the mated solutions in the mating
values into binary representation. After the genetic process, pool. Mated solutions for crossover are generally named as
coded solutions are decoded reversely (Goldberg, 1989). parent solutions. Three kinds of crossover method are used in
this study. They are single-point crossover, multi-point crossover,
3.3 Reproduction Operator and uniform crossover (Goldberg, 1989; Aydin, 2006). Detailed
The worst solutions are taken out from population and the best information about crossover operator can be found in the studies
solutions are copied instead of them by reproduction operator. The realized by Aydin and Ayvaz (2010), and Aydin (2006).
worst and the best solutions are determined according to their
fitness values. Fitness value of the ith solution can be calculated as: 3.5 Mutation Operator
Mutation operator also creates new solutions in population.
Fi = ( Φmax + Φmin ) – Φi (14)
One character of a few solution strings is changed in a population
where Φmax, and Φmin are the largest and smallest values of the by mutation operator. Mutated solutions and mutated characters

Table 1. Design Parameters of the Bridge

Total length of bridge 177.70 m
Width of deck 12.00 m
Width of motorway 9.50 m
General properties of the bridge Number of traffic lanes 3
Thickness of slab 21 cm
Distance between top flanges of PC girders 4.5 cm
Distance between slab and girders 1 cm
Cast-in-place concrete C25
Precast concrete C40
Reinforcing steel S420
Material properties
Ultimate stress of PS 1898 MPa
Modulus of elasticity of PS 207300 MPa
Diameter of PS tendons 15.24 mm
Truck load category H30-S24
Unit weight of coating 20 kN/m3
Thickness of coating 10 cm
Loads
Thickness of sidewalk 30 cm
Weight of parapet and precast 30 N/cm
Load of pedestrian 0.003 MPa
Average of annual relative humidity % 40
Other properties Minimum spacing between prestressing tendons 6 cm
Soil bearing capacity 0.35 MPa

− 772 − KSCE Journal of Civil Engineering

 Overall Cost Optimization of Prestressed Concrete Bridge using Genetic Algorithm

are selected randomly. The mutation rate is generally determined Table 2. Unit Prices
to be between 0.1% and 1%. Three kinds of the mutation Production Unit Unit price ($)
operator are used in this study. They are standard mutation, PC (uppc) m3 900
controlled mutation, and hybrid mutation (Goldberg, 1989; PS (upps) ton 9500
Aydin, 2006). Detailed information about mutation operator can Cast-in-place concrete for bridges (upbc) m3 220
be found in the studies realized by Aydin and Ayvaz (2010), and Ordinary cast-in-place concrete (upc) m3 80
Aydin (2006). Lean concrete (uplc) m3 80
Reinforcing steel (ups) ton 1500
4. Numerical Example Formwork (upf) m2 35
Excavation (upex) m3 42
Several numerical examples are designed to demonstrate
efficiency of the program. One of these examples is given here.
Selected bridge has been previously designed by Emay variable values. There are 16 possible values for the 2nd, 3rd, and
International Eng. Cons. Co. (Emay, 1998) and constructed in 4th design variables, so that a four-character string must be used
Turkey. Design parameters of the bridge for this example are to demonstrate each of these design variable values. Similarly,
taken as the same with application project and given in Table 1. there are 32 possible values for the 10th design variable, so that a
The unit prices of the productions are taken according to five character string must be used for this design variable.
Turkish unit price criteria and given in Table 2. The site Therefore, each of the solution strings must have 35 characters.
altitudes of the valley are also considered as the same with In other words, 35 number characters (“1” or “0”) must be
those of application project. determined randomly to constitute each initial solution. Design
Design variables values sets for ten design variables are given variable values sets are taken to be the same with those of Aydin
in Table 3. There are 8 possible values for the 1st, 5th, 6th, 7th, 8th, and Ayvaz (2010).
and 9th design variables, as shown in Table 3, so that a three- Several of the optimum design parameters for GA are
character string must be used to demonstrate each of these design determined as follows for this example.

Table 3. Design Variable Values Sets

Design variable Number of values Selected values
Number of piers (np) 8 23456789
Number of PC girder (ng) 16 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Height of beam (hb) (cm) 16 85 90 95 100 105 110 115 120 125 130 135 140 145 150 155 160
Bottom flange width (bb) (cm) 16 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Web thickness (bw) (cm) 8 15 16 17 18 19 20 22.5 25
Top flange thickness (tt) (cm) 8 10 11 12 13 14 15 17.5 20
Bottom flange thickness (tb) (cm) 8 15 20 25 30 35 40 45 50
Top sloped thickness (pt) (cm) 8 5 6 7 8 9 10 12 12.5 15
Bottom sloped thickness (pb) (cm) 8 5 6 7 8 9 10 12 12.5 15
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Number of PS tendons (nps) 32
23 24 25 26 27 28 29 30 31 32 33 34 35 36

Table 4. Optimum Design Parameters and Penalized Objective Functions

Design Type of crossover Type of mutation Penalized objective
Iteration number
no operator operator function ($)
1 Single point Standard No convergence 2,692,943
2 Single point Controlled 474 2,692,943
3 Single point Hybrid 156 2,894,343
4 Single point No mutation 123 2,711,469
5 Multi point Standard No convergence 2,713,446
6 Multi point Controlled 527 2,909,295
7 Multi point Hybrid 120 2,689,540
8 Multi point No mutation 146 2,869,209
9 Uniform Standard No convergence 2,736,037
10 Uniform Controlled 408 2,686,504
11 Uniform Hybrid 193 2,704,110
12 Uniform No mutation 164 2,689,417

Vol. 17, No. 4 / May 2013 − 773 −

 Zekeriya Ayd i n and Yusuf Ayvaz

Table 5. Optimum Values of Design Variables and Cost

hb bb bw tt tb pt pb CT
Solution np ng nps
(cm) (cm) (cm) (cm) (cm) (cm) (cm) ($)
This study 6 7 145 35 16 10 30 6 7 18 2,686,504
Application project 5 11 120 75 20 10 18 8 10 24 3,075,547

The number of solution in generation ··························· : 36 10th design in Fig. 3.

Penalty coefficient ·························································· : 0.6 Cross-section of the PC girder, cross-section of the bridge,
Convergence criterion ···················································· : %60 cross-section of pier, and longitudinal section of the bridge are
Rate of mutation ····························································· : %0.3 given for the application project and for the optimum design in
Maximum number of iteration ······································· : 2000
Optimum design is realized several times for different types of
crossover and mutation operators. The results of these designs
are given in Table 4.
It can be seen from Table 4 that the optimum solution is
reached at the 10th design by using uniform crossover and
controlled mutation. Optimum values of design variables and the
total cost reached at the 10th design is given in Table 5. The
values of application project are also given in Table 5 to compare
the solutions. The total cost obtained in this study is %12.6 less
than the application project. The penalty functions of all the final Fig. 5. Cross-section of the Bridge: (a) For the Application Project,
solutions are zero. This means that all of the final solutions in (b) For this Study
Table 4 are code-compliant. Therefore, the solutions are usable
according to AASHTO Standard Specifications for Highway
Bridges. The change of average of the penalized objective
functions according to iteration number is demonstrated for the

Fig. 6. Cross-section of the Piers: (a) For the Application Project,

(b) For this Study (Dimensions in Centimeters)

Fig. 3. Variation of Average of the Penalized Objective Functions

for the10th Design

Fig. 4. Cross Section of the Girder: (a) For the Application Project, Fig. 7. Longitudinal Section of the Bridge: (a) For the Application
(b) For this Study (Dimensions in Centimeters) Project, (b) For this Study (Dimensions in Meters)

− 774 − KSCE Journal of Civil Engineering

 Overall Cost Optimization of Prestressed Concrete Bridge using Genetic Algorithm

Table 6. Footing Dimensions for the Application Project and for this Study
Dimension
Solution 1st footing 2nd footing 3rd footing 4th footing 5th footing 6th footing
(m)
B 12.00 12.50 12.50 12.50 12.50 11.00
This study L 15.00 19.00 19.00 18.00 15.00 13.00
H 2.45 2.45 2.45 2.45 2.45 2.45
B 13.00 13.20 13.20 13.00 12.00
Application project L 16.50 20.40 20.40 16.50 15.00
H 2.50 2.50 2.50 2.50 2.50

Fig. 4-Fig. 6, and Fig. 7, respectively. Footing dimensions for the n: Total number of constraints
application project and for the optimum design are given in Table 6. P: Penalty function
pi: Violation factor of the ith constraint
5. Conclusions Pn: Nominal axial load strength of the piers
Pu: Factored axial load at the piers
1. Cost optimization of PC I-girder bridges is realized by using q: Maximum stress for the soil under the spread footing
GA. It is concluded that the program coded in this study can qallow: Allowable stress of the soil
be efficiently used in the optimizations of PC bridges. upc: Unit price of ordinary concrete
2. The optimum number of piers, the optimum number of PC upbc: Unit price of cast-in-place concrete for bridges
girders and optimum shape of PC girder are found out at the upex: Unit price of excavation
same time. Although it is very difficult to solve such a com- upf: Unit price of formwork
plicated problem with usual optimization techniques, it is uplc: Unit price of lean concrete
solved easily by GA. uppc: Unit price of precast concrete
3. The solutions reached by GA are realistic and constructible ups: Unit price of reinforcing steel
without modifying because of using discrete design vari- upps: Unit price of prestressing steel
ables and probable design variable values sets. Vc: Shear strength of piers provided by the concrete
4. Shape of the site is considered, and heights of the piers are Vn,f: Nominal shear strength of the footings
determined according to site altitude. Therefore, optimum Vs: Shear strength of piers provided by the shear
number of piers is determined as more realistic. reinforcement
5. The solution obtained by the proposed GA is up to %12.6 Vu: Factored shear force at the pier section
more economical than the application project. Vu,f,: Factored shear force at the footings
6. In this study, various crossover and mutation types are used. φ: Strength reduction factor
The best solutions are obtained by using the uniform cross- Φ: Penalized objective function
over and the controlled mutation. Φmax: The largest value of the penalized objective function in
7. Constraints can easily be adapted to the problem using the each generation
penalty function and design variable value sets. Φmin: The smallest value of the penalized objective function in
each generation
Notations
References
Cf: Cost of foundations
AASHTO (2002). Standard specifications for highway bridges, American
Cg: Cost of girders
Association of State Highway and Transportation Officials, Washington,
Cp: Cost of piers D.C.
CT: Total cost (Objective function) Aydin, Z. (2006). Optimum design of prestressed concrete bridge
Ctb: Cost of transverse beams girders using genetic algorithm, PhD Thesis, Karadeniz Technical
Fc,i: Fitness factor for the ith solution string University, Turkey.
Fi: Fitness value of the ith solution Aydin, Z. and Ayvaz, Y. (2010). “Optimum topology and shape design
Fort: Average of the fitness values of prestressed concrete bridge girders using a genetic algorithm.”
gi: Normalized value of ith constraint Struct. Multidiscip. Optim., Vol. 41, No. 1, pp. 151-162.
Ayvaz, Y. and Aydin, Z. (2000). “Optimum design of trusses using a
K: Penalty coefficient
genetic algorithm.” Proceedings of the Second International Conference
Mn: Nominal moment strength of the piers on Engineering using Metaphors from Nature, Leuven, Belgium, pp.
Mn,f: Nominal moment strength of the footings 159-168.
Mu: Factored moment at the piers Cohn, M. Z. and Lounis, Z. (1994). “Optimal design of structural
Mu,f: Factored moment at the footings concrete bridge systems.” J. Struct. Eng., Vol. 120, No. 9, pp. 2653-

Vol. 17, No. 4 / May 2013 − 775 −

 Zekeriya Ayd i n and Yusuf Ayvaz

2674. concrete bridge girder systems.” PCI J., Vol. 38, No. 4, pp. 60-78.
o
Daloglu, A. and Aydin, Z. (1999). “Kafes sistemlerin uygulamaya Lounis, Z., Mirza, M. S., and Cohn, M. Z. (1997). “Segmental and
yönelik optimum tasarimi.” Mühendislik Bilimleri Dergisi, Vol. 5, conventional precast prestressed concrete I-bridge girders.” J.
No. 1, pp. 951-957 (in Turkish). Bridge. Eng., Vol. 2, No. 3, pp. 73-82.
Degertekin, S. O., Saka, M. P., and Hayalioglu, M. S. (2008). “Optimal Pezeshk, S., Camp, C. V., and Clem, D. (2000). “Design of nonlinear
load and resistance factor design of geometrically nonlinear steel framed structures using genetic optimization.” J. Struct. Eng., Vol.
space frames via tabu search and genetic algorithm.” Eng. Struct., 126, No. 3, pp. 382-389.
Vol. 30, No. 1, pp. 197-205. Rajan, S. D. (1995). “Sizing, shapes and topology design optimization
Emay International Eng. Cons. Co. (1998). Uzunçayir baraji varyanti of trusses using genetic algorithm.” J. Struct. Eng., Vol. 121, No. 10,
yolu dinar köprüsü uygulama projesi hesaplari, I·stanbul (in pp. 1480-1486.
Turkish). Rajeev, S. and Krishnamoorthy, C. S. (1998). “Genetic algorithm-based
Goldberg, D. E. (1989). Genetic algorithms in search, optimization and methodology for design of reinforced concrete frames.” Comput.-
machine learning, Addison-Wesley Publishing Company, Inc., New Aided. Civ. Infrastruct. Eng., Vol. 13, No. 1, pp. 63-74.
York, N. Y. Saka, M. P. (2007). “Optimum topological design of geometrically
Holland, J. H. (1975). Adaptation in natural and artificial systems, nonlinear single layer latticed domes using coupled genetic algorithm.”
University of Michigan Press, Ann Arbor, Mich. Comput. Struct., Vol. 85, Nos. 21-22, pp. 1635-1646.
Jenkins, W. M. (1992). “Plane frame optimum design environment Sarma, K. C. and Adeli, H. (2000). “Fuzzy discrete multicriteria cost
based on genetic algorithm.” J. Struct. Eng., Vol. 118, No. 11, pp. optimization of steel structures.” J. Struct. Eng., Vol. 126, No. 11,
3103-3112. pp. 1339-1347.
Jones, H. L. (1985). “Minimum cost prestressed concrete beam design.” Sirca, G. F. and Adeli, H. (2005). “Cost optimization of prestressed
J. Struct. Eng., Vol. 111, No. 11, pp. 2464-2478. concrete bridges.” J. Struct. Eng., Vol. 131, No. 3, pp. 380-388.
Kameshki, E. S. and Saka, M. P. (2001). “Optimal design of nonlinear Totres, G. G. B., Brotchie, J. F., and Cornell, C. A. (1966). “A program
steel frames with semi-rigid connections using a genetic algorithm.” for the optimum design of prestressed concrete highway bridges.”
Comput. Struct., Vol. 79, No. 17, pp. 1593-1604. PCI J., Vol. 11, No. 3, pp. 63-71.
Kumaran, S. S., Muthukumaran, S., and Vinodh, S. (2010). “Optimization Yu, C. H., Das Gupta, N. C., and Paul, H. (1986). “Optimization of
of friction welding of tube to tube plate using an external tool.” prestressed concrete bridge girders.” Eng. Optim., Vol. 10, No. 1, pp.
Struct. Multidiscip. Optim., Vol. 42, No. 3, pp. 449-457. 13-24.
Lounis, Z. and Cohn, M.Z. (1993). “Optimization of precast prestressed

− 776 − KSCE Journal of Civil Engineering