ENGINEERING STRUCTURES AND TECHNOLOGIES

ISSN 2029-882X / eISSN 2029-8838

2016 8(4): 157–164
doi:10.3846/2029882X.2016.1250230

MULTI-OBJECTIVE OPTIMIZATION OF REINFORCED

CONCRETE FRAMES USING NSGA-II ALGORITHM

Mehdi BABAEI, Masoud MOLLAYI

Department of Civil Engineering, Faculty of Engineering, University of Zanjan, Zanjan, Iran
Received 25 April 2016; accepted 16 October 2016
Abstract. In recent decades, the use of genetic algorithm (GA) for optimization of structures has been
highly attractive in the study of concrete and steel structures aiming at weight optimization. However, it has
been challenging for multi-objective  optimization  to determine the trade-off between objective  functions
and to obtain the Pareto-front for reinforced concrete (RC) and steel structures. Among different methods
introduced for multi-objective optimization based on genetic algorithms, Non-Dominated Sorting Genetic
Algorithm II (NSGA II) is one of the most popular algorithms. In this paper, multi-objective optimization of
RC moment resisting frame structures considering two objective functions of cost and displacement are in-
troduced and examined. Three design models are optimized using the NSGA-II algorithm. Evaluation of op-
timal solutions and the algorithm process are discussed in details. Sections of beams and columns are con-
sidered  as design variables and the specifications of the American Concrete Institute (ACI) are employed
as the design constraints. Pareto-fronts for the objective space have been obtained for RC frame models of
four, eight and twelve floors.  The results indicate smooth Pareto-fronts and prove the speed and accuracy
of the method.
Keywords: NSGA-II, multi-objective optimization, moment resisting frame, reinforced concrete, weight,
displacement, Pareto front.

Introduction Optimization using genetic algorithms has been

welcomed in recent decades by many researchers. The
Considering the fact that reinforced concrete (RC)
structures compared to steel structures have more va- research conducted by  Hoit et  al. (1991) focused on
riety in materials, hence optimization of RC structures minimizing the weight of the structure using Lagrangi-
are more complex than steel structures. Among the an method and nonlinear techniques. In 1994, Adamu
variables included in the optimization of RC structures et al. (1994) published an article about optimization of
are: compressive strength of concrete, steel yield stress, the cost of RC beams. In 1997, Baling and Yao (1997)
size of cross sections, and percentage of reinforce- published a paper on the three-dimensional optimiza-
ments, shape and topology of the structure. To calcu- tion of RC frames using multi-layer methods. In 2003,
late an RC building structural cost, three parameters the article of Camp et al. (2003) was published about
including the required volume of concrete, weight of optimization of moment resisting frames using genetic
reinforcement steel, and formwork are important. It algorithms.
should be noted that labor costs should also be add- Cellular automata method was applied to opti-
ed to the above costs, although one can include labor mize the continuum structures (Sanaei, Babaei 2012),
costs in the cost of formwork, as it has been assumed reporting interesting results. Analysis and optimiza-
in this article. tion of elastic-plastic framing structures was investi-

Corresponding author:
M. Babaei E-mail: [email protected]
Copyright © 2016 Vilnius Gediminas Technical University (VGTU) Press
http://www.tandfonline.com/TESN 157
158 M. Babaei, M. Mollayi. Multi-objective optimization of reinforced concrete frames using NSGA-II algorithm

gated under complex constraints (Gervytė, Jarmola- 1. Optimization process

jeva 2013). The impact of ductility levels on the cost 1.1. Objective functions definition
of RC moment resisting frames for buildings with 5
In this paper, two different objective functions are pre-
to 15 floors was investigated (Babaei 2015b). Optimi-
sented. The first objective function  is the structural
zation of RC frames to determine the best topology
cost and the second criterion to be optimized is the
and optimal arrangement of columns have been stud-
maximum lateral displacement. For this purpose, the
ied in structures ranging from 5 to 10 floors (Babaei
cost is calculated on the basis of three parameters: vol-
2015a).  Similar studies were carried out to explore
ume of the required concrete, weight of the required
optimal topology for steel structures (Babaei, Jabbar
reinforcement, and the cost of formwork and labor.
2016; Babaei, Omidi 2015).
Both objective functions have to be minimized. There-
On the other hand, it is inevitable to consider
fore, the first objective function for cost minimization
more than one objective function for engineering op-
is introduced as follows:
timization problems. Therefore, multi-objective op-
Fc = f (Ps , P c , Pf ) , (1)
timization has been introduced and applied in many
fields of engineering practice (Deb 2001; Andersson where Fc is the objective function to minimize con-
2001; Ghosh, Dehuri 2004; Carrillo, Taboada 2012). struction cost, Ps is the cost of the required reinforce-
In recent years, many multi-objective optimization ment, Pc is the cost of the required concrete and Pf is
methods based on genetic algorithm have been pre- the cost of formwork and labor. 
sented, including weighted sum method (WSM) (Kim, Design variables are assumed to be the  dimen-
De Weck 2005), vector evaluated genetic algorithm sions of cross sections and the cross sectional area of
(VEGA) (Ghosh, Dehuri 2004), strength Pareto evolu- reinforcement bars. Standard problem formulation is
tionary algorithm (SPEA) (Sarker et al. 2002), Niched as follows:
Pareto genetic algorithm (NPGA) (Horn et al. 1994), Minimize F1 = f (Ps , Pc , Pf ) ; (2)

and non-dominated sorting genetic algorithm (NSGA) = F1 ∑ Cs LAs + Cc Lbh + C f Lp ; (3)
(Deb et al. 2000). element
In recent years, the powerful algorithms of NSGA Minimize F2 = max(roof displacement) , (4)
have been used for the optimization in various fields

including RC structures. Some of these studies  are: where, Cc is the cost of concrete per unit volume, Cs
two-objective optimization of CO2emissions and cost is the cost of reinforcement per unit volume, Cf is the
cost of formwork per unit area, L is the cross section
for composite building design using NSGA (Park et al.
length, As is the cross section area of reinforcement, b
2012), synthesis of truss structure designs by NSGA-II
is the cross sectional width of the beams or columns, h
and Node Sort algorithm (Stanković et al. 2012), and
is the cross sectional length of the beams or columns,
cost and CO2 emission optimization of steel reinforced
and p is the perimeter of the cross sections.
concrete columns in high-rise buildings (Park et  al.
2013). A comparative study for two meta-heuristic al-
1.2. Constraints and penalty functions definition
gorithms and big bang-big crunch for optimal design
of RC frames carried out and reported in the literature Almost all engineering optimization problems include
(Kaveh, Sabzi 2011, 2012). constraints, which have to be satisfied by each feasi-
The following section deals with the optimization ble design. In optimal design of RC moment resisting
process by introducing objective function, constraints, frames, there following constraints are applied.
code specifications and design variables for RC mo- The stress constraints are as:
ment resisting frames. Section 2 describes the NSGA- σi
≤ 1, i =
1 to n. (5)
II algorithm for optimal design. Section 3 includes test σia

models to illustrate the capability of the algorithm. Sig- In addition, the displacement constraints are:
nificant conclusions are drawn at the end of the paper.
δi ≤ δmax , i =1 to m , (6)

where si and σia are the existing and allowable bend-
ing stresses, respectively; di and dmax are the story and
allowable displacements, respectively.
Engineering Structures and Technologies, 2016, 8(4): 157–164 159

To satisfy all constraints in an optimization prob-

lem, one of the popular methods developed in the
literature is to define penalty functions. On the other
hand, from the construction point of view, columns
B(Mu, Pu)
of the lower levels must have larger dimensions than
the columns located in the upper floors, as well as, the
reinforcing bars and their number in  the upper col-

ЖPn
A(ЖMn, ЖPn)
umns must be less than or equal to the lower columns.
Other restrictions related to the design code require-
ments must also be satisfied. In the penalty approach,
if a constraint is not satisfied, the objective (cost or
displacement) functions are penalized. In this method,
in addition to simplifying the objective function, re-
strictions will be applied in such a way that the cost O
ЖMn
of structures or  their displacement is increased, so
automatically a penalized design would have a little Fig. 1. Penalty calculation in columns
chance to be selected as an optimal solution. Many
methods  are developed in the literature to introduce BO − AO
Column penalty = . (9)
penalty function. In this article, the penalty function AO
is applied as follows: For beams, it is less complex to apply penalties
Ci = aiVi; (7) than for columns. It needs only to calculate the ratio of
n the applying bending moment to the bending strength
Φ =∑ Ci . (8) of the section. When this ratio is less or equal to 1,
i =1
then there is no need for penalty, otherwise the ele-
In these equations Ci is the penalty function for
ment must be penalized. Reinforcement ratio limits for
each of the constraints, ai is  penalty factor of each
columns and beams, distance between bars, and other
constraint, Vi is the amount of each penalty, and j is
code regulations are imposed to all structural models
the total penalty. Based to the problem, penalty coef-
of this study.
ficient values​​are obtained through trial and error, and
are unique values for each of the structures (Carrillo,
2. NSGA-II algorithm
Taboada 2012).
Inspired from the nature, GA is one of the me-
1.3. ACI Specifications ta-heuristic algorithms that have been welcomed
In this study, the American Concrete Institute (ACI) by many researchers in recent decades. Genetic algo-
Code restrictions for each of the  elements  and all rithms are used to solve single objective optimization
the requirements for the number and location of rein- problems while NSGAs are developed to solve multi-
forcing bars are applied and satisfied through design- objective optimization problems. GA selects the most
ing the structural elements. Constraints are applied qualified chromosomes and sends them  to the next
using a penalty function as introduced in the previous generation. Through selection, crossover, and muta-
section. There are different methods for calculating tion operations, the algorithm finds the best solution.
penalties. Constraints need to be satisfied by two types Non-dominated sorting algorithm II (NSGA-II) algo-
of elements in a frame structure; columns and beams. rithm, suggested in 2000 by Deb and his colleagues
In a column there is an interaction between axial load (Deb et al. 2000), is an algorithm based on genetic al-
and bending moment and this interaction has to be gorithm (GA) for multi-objective optimization prob-
considered to evaluate the strength of a column, in- lems (Andersson 2001; Ghosh, Dehuri 2004; Deb et al.
stead of a separate design for axial force and bending 2000; Deb 2001). Compared to the first version of the
moment. To apply penalties to a column, which is un- algorithm, the second version of the NSGA is faster.
der axial load and bending moment, the following for- In both versions of the algorithm, all members of the
mula is used, using the description in Figure 1. population are compared to each other based on the
160 M. Babaei, M. Mollayi. Multi-objective optimization of reinforced concrete frames using NSGA-II algorithm

objective functions. In NSGA-II congestion distance is obtain reliable results, ten runs are implemented and
used as an important parameter to improve the algo- finally the best ones are shown in the figures.
rithm. This parameter provides an appropriate diver- Cross-sections for columns and beams are em-
sity for the members of the population. ployed from 30×30  (cm2) as the smallest section
to  70×70  (cm2) as the largest section.  To determine
3. Test models and results the efficiency of the algorithm  the  stress  ratios for
structural elements are captured. Since in this model
In this paper, three different sample models were
two groups for beams and two groups for columns are
studied to evaluate the performance of the algorithm.
defined to be representative for twelve members, the
The topologies of the models are similar to those
average maximum stress in these groups of beams and
studied in the previous works of the literature (Kaveh,
columns are shown in Figures  5 and 6.  These stress
Sabzi 2011, 2012). These models of three-span struc-
ratios are shown for the optimal solutions of the 1st,
tures with four, eight and twelve floors are considered
20th, 50th, and 100th populations as representative of
as representative structures for low, medium and high
all members.
rise buildings. For all models a story height of 3.3 m
and a span of 7.5 m are considered. Uniform live and 12.4 kN B2 B2 B2
dead loads are applied to beams and lateral loads are
C1 C2 C2 C1
distributed between beam-column joints. According to 9.3 kN B2 B2 B2
the ACI regulations three load combinations are con-

4ґ3.3 m = 13.2 m
sidered to design structural elements as follows: C1 C2 C2 C1
6.2 kN B1 B1 B1
U = 1.2D + 1.6L;
C1 C2 C2 C1
U = 1.2D + 1.0L ± 1.0E; 3.1 kN B1 B1 B1
U = 0.9D + 1.0E. (10)
C1 C2 C2 C1
Uniform dead and live loads of DL = 22.3 KN/m
and LL  = 10.7 KN/m are applied, respectively. The 7.5 m 7.5 m 7.5 m

compressive strength of concrete  is assumed to be

Fig. 2. Topology of the first model
fc′ = 280 kg/ cm2 and the yield strength of steelis con-
sidered as fy  = 4200 kg/cm2. The termination crite- First population
1.6
ria for the algorithm can be the convergence or the
1.4
number of iterations. Based on some evaluations im- 1.2
Displacement (cm)

plemented for this algorithm the number of iterations 1

is applied as the termination criteria. To find the best 0.8

population number and the best number of iterations 0.6

0.4
different populations with different iterations are ex-
0.2
amined and ultimately a population of 100 members 0
3000 4000 5000 6000 7000 8000
and number of repetitions of 100 obtained to be ef-
Cost (USD)
fective.
Fig. 3. The population diversity in the first model
3.1. First model
Final population
Topology and loadings for this model, which is a rep- 2

resentative of low rise buildings, is illustrated in Fig- NSGA-II

Displacement (cm)

1.5
Fitted curve of NSGA-II
ure 2. Columns and beams are grouped into two differ-
ent types. The costs for unit volume (m3) of concrete of 1

30 USD, unit weight (kg) of steel of 0.50 USD and unit 0.5

area (m2) of formwork of 4.42 USD are considered for
0
cost estimation. 2000 3000 4000 5000 6000 7000 8000 9000
Cost (USD)
In Figures  3 and 4, the population diversity in
the first iteration and the Pareto-front are shown. To Fig. 4. Pareto-front obtained in the first model
 Engineering Structures and Technologies, 2016, 8(4): 157–164 161

In Figures 7 and 8 the population number of the words, the population diversity in the first front was
first front and congestion distance ​​are illustrated. As improved through the iterations. As shown in Figure 7,
it is clear in these figures, when the population size in after about 32 iterations all of the population members
the first front was increased during the iterations the are selected to be a member of the first front. In other
algorithm decreased the congestion distance. In other words, the algorithm distributes the solutions through
the objective space in an efficient way so that no solu-
Maximum stress ratio of beams
1
tion is dominated by another one. It should be noted
pop(1)
0.9 that the congestion distance (C.D.) in Figure 8 is a fac-
0.8 pop(20)
0.7
tor to define density estimation of the solution for any
pop(50)
specific Pareto front. It is defined as the distance of
Stress ratio

0.6
pop(100)
0.5
two neighbouring solutions on either side of a solution
0.4
0.3 along each objective axis.
0.2
0.1
0
3.2. Second model
type 1 type 2
As with the first example, a span of 7.5 m and a height
Fig. 5. Stress ratio of the first model beams groups of 3.3 m are considered for this example also. Beams
are grouped into three types and columns are grouped
1
pop(1)
into four types as illustrated in Figure 9. Gravity loads
0.9
0.8 pop(20) include live loads and dead loads, same as the previ-
0.7
pop(50) ous example. Grouping and loadings assumed for this
Stress ratio

0.6
0.5 pop(100) sample are shown in Figure 9.
0.4 By examining the initial population in the first
0.3
0.2 iteration and the obtained Pareto front from the
0.1 100th iteration (Fig.  10), robustness of the algorithm
0
type 1 type 2 for multi-objective optimization is clear. Considering
Fig. 6. Stress ratio of the first model columns groups this figure and results for the population diversity, the

100 28 kN B3 B3 B3
90
80
Number of population

C3 C4 C4 C3
70
24.5 kN B3 B3 B3
60
50
40 C3 C4 C4 C3
30 21 kN B2 B2 B2
20
10 C3 C4 C4 C3
0 17.5 kN B2 B2 B2
0 10 20 30 40 50 60 70 80 90 100
Iteration
8ґ3.3 m = 13.2 m

C3 C4 C4 C3
Fig. 7. The population number in the first front 14 kN B2 B2 B2
for the first model
C1 C2 C2 C1
Max C.D. 10.5 kN B1 B1 B1
0.6
0.5 C1 C2 C2 C1
0.4 7 kN B1 B1 B1
C.D.

0.3
C1 C2 C2 C1
0.2 3.5 kN B1 B1 B1
0.1
C1 C2 C2 C1
0
0 10 20 30 40 50 60 70 80 90 100
Iteration 7.5 m 7.5 m 7.5 m
Fig. 8. Maximum values ​​of the congestion distance
in the first model (C.D.: congestion distance) Fig. 9. Topology of the second model
 162 M. Babaei, M. Mollayi. Multi-objective optimization of reinforced concrete frames using NSGA-II algorithm

cost in the initial population varies from 7000 USD to 100

90
37000 USD, while in the final population it varies from 80

Number of population
7000 USD to 18000 USD. The maximum roof displace- 70
60
ment in the first population is about 30 cm, while it is 50
obtained 6 cm in the final population, which shows a 40
30
sharp decrease. 20
In this model, to compare the stress ratio of 10
beams and columns, four optimal solutions of the 1st, 0
0 10 20 30 40 50 60 70 80 90 100
10th, 30th, and 100th from the Pareto-front are selected, Iteration

as shown in Figures 11 and 12. As expected the maxi- Fig. 13. The population number in the first front
of the second model
mum amount of stress ratio is obtained for the first
solution (with the lowest cost), while the minimum 1.2
Max C.D.

stress ratio is obtained for the 100th solution (with the 1

lowest displacement). 0.8

C.D.
0.6
Final population
6 0.4

5 0.2
Displacement (cm)

0
4 0 10 20 30 40 50 60 70 80 90 100
Iteration
3
Fig. 14. Maximum values ​​of the congestion distance
2
in the second model
1

0 For this model, population number in the first

7000 9000 11000 13000 15000 17000 19000
Cost (USD)
front and the congestion distance are calculated and
presented in Figures 13 and 14. It is observed that af-
Fig. 10. Pareto-front of the second model
ter 34 iterations all members of the population were in
Maximum stress ratio for beams the first front. In other words, the algorithm distributes
1.00
pop(1) the solutions efficiently through the objective space.
0.90
0.80
pop(10) The maximum congestion distance has declined in two
pop(30)
0.70
pop(100) phases, which reflects the strength of the algorithm in
Stress ratio

0.60
0.50 sorting members in a front.
0.40
0.30
0.20
3.3. Third model
0.10
0.00
This example is considered to represent a tall structure.
type 1 type 2 type 3 Similar s​pan lengths and story height are considered
Fig. 11. The maximum stress ratio of beams as assumed for the previous examples. The only differ-
for the second model ences are the number of stories and the lateral loads,
as shown in Figure 15. Trade-off between two objec-
Maximum stress ratio for beams
1
tive functions, cost and maximum roof displacement,
pop(1)
0.9 is demonstrated in Figure 16. Figures 17 and 18 illus-
pop(10)
0.8 pop(30) trate the stress ration of beams and columns for four
0.7 pop(100) optimal solutions selected from the final Pareto-front.
0.6
Stress ratio

0.5
These four optimal solutions are representatives of the
0.4 scenarios in Pareto-front showing from the upper to
0.3 the lower limit of the objective functions. The popula-
0.2 tion number in the first front and the congestion dis-
0.1
tance are calculated and presented in Figures 19 and
0
type 1 type 2 type 3 type 4 20. It is observed that after 34 iterations all members
Fig. 12. The maximum stress ratio of columns of the population were in the first front and similar to
for the second model the previous examples the algorithm works efficiently.
 Engineering Structures and Technologies, 2016, 8(4): 157–164 163

36 kN B3 B3 B3 Maximum stress ratio for beams

1
C5 C6 C6 C5 0.9
33 kN B3 B3 B3 0.8
0.7

Stress ratio
C5 C6 C6 C5 0.6
30 kN B3 B3 B3 0.5
0.4
C5 C6 C6 C5
0.3
27 kN B3 B3 B3 0.2
0.1
0
C5 C6 C6 C5
type 1 type 2 type 3 type 4 type 5 type 6
24 kN B2 B2 B2
Fig. 18. The stress ratio of columns for the third model
C3 C4 C4 C3
21 kN B2 B2 B2
100
90

12ґ3.3 m = 13.2 m
C3 C4 C4 C3
80

Number of population
18 kN B2 B2 B2
70
60
C3 C4 C4 C3
50
15 kN B2 B2 B2
40
30
C3 C4 C4 C3 20
12 kN B1 B1 B1
10
0
C1 C2 C2 C1 0 10 20 30 40 50 60 70 80 90 100
9 kN B1 B1 B1 Iteration

Fig. 19. The population number in the first front

C1 C2 C2 C1
6 kN
for the third model
B1 B1 B1

Max C.D.
C1 C2 C2 C1
1.2
3 kN B1 B1 B1
1
C1 C2 C2 C1
0.8
C.D.

7.5 m 7.5 m 7.5 m 0.6

Fig. 15. Grouping of beams and columns in the third model 0.4

0.2
Final population
16 0
14 0 10 20 30 40 50 60 70 80 90 100
Iteration
Displacement (cm)

12
10 Fig. 20. Maximum values ​​of the congestion distance
8 in the third model
6
4
2 Conclusions
0
14000 16000 18000 20000 22000 24000 26000 28000 In this paper, a bi-objective algorithm of optimization
Cost (USD)
of RC moment resisting frame has been investigated
Fig. 16. Pareto-front for the third model using the NSGA-II method. Two objective functions,
Maximum stress ratio for beams including the total structural cost and the maximum
0.80 roof displacement, which are in conflict, are defined
pop(1)
0.70
pop(10) and applied for the bi-objective optimization of RC
0.60 pop(30) models.  Discrete variables including dimensions of
0.50 pop(100)
sections of beams and columns were employed as de-
Stress ratio

0.40
sign variables. Three models were considered for rep-
0.30
resentative for four, eight and twelve-story buildings.
0.20
Stress ratio for beams and columns, amount of conges-
0.10
0.00
tion distance in the first front, and changing strategy
type 1 type 2 type 3 for the size of the first front were presented. Trade-offs
Fig. 17. The stress ratio of beams for the third model between cost and displacement objective functions
164 M. Babaei, M. Mollayi. Multi-objective optimization of reinforced concrete frames using NSGA-II algorithm

showed potential optimal solutions, however, as shown Deb,  K.; Agrawal,  S.; Pratap,  A.; Meyarivan, T. 2000. A fast
in the article body, focusing more on displacement op- elitist non-dominated sorting genetic algorithm for multi-
objective optimization: NSGA-II, Lecture Notes in Computer
timization leads to sensitive design for cost function. Science 1917: 849–858.
In other words, the trade-off curve is stretched along https://doi.org/10.1007/3-540-45356-3_83
the cost function more than the displacement func- Gervytė, A.; Jarmolajeva, E. 2013. Analysis and optimization of
tion, indicating that the most cost-effective structure elastic-plastic framing structures under complex constraints,
Engineering Structures and Technologies 5(4): 159–166.
does not have a very large displacement but the most
https://doi.org/10.3846/2029882X.2014.898366
stiffed structure, with the lowest displacement, needs
Ghosh, A; Dehuri, S. 2004. Evolutionary algorithms for multi-
very large sections, which leads to an expensive design. criterion optimization: a survey, International Journal of
Results showed smooth Pareto-front curve for all mod- Computing & Information Sciences 2(1): 38–57.
els, which means that the algorithm is robust enough Hoit, M.; Soeiro, A.; Fagunda, F. 1991. Probabilistic design and
to find dominant solutions with a reasonable diversity. optimization of reinforced concrete frames, Journal of Engi-
neering Optimization 17: 229–235.
Horn, J.; Nafpliotis, N.; Goldberg, D. E. (Eds.). 1994. A niched
References Pareto genetic algorithm for multi-objective optimization,
Adamu, A.; Karihaloo, B. L.; Rozvany, G. I. N. 1994. Minimum in Proceedings of the First IEEE Conference on Evolutionary
cost design of reinforced concrete beams using continuum- Computation, IEEE World Congress on Computational Intel-
type optimality, Structural and Multidisciplinary Optimiza- ligence, 27–29 June 1994, Orlando, Florida, 1: 82–87.
tion 7: 91–102. https://doi.org/10.1007/BF01742512 Kaveh, A.; Sabzi, O. 2011. A Comparative study of two meta-
Andersson, J. 2001. Multiobjective optimization in engineering heuristic algorithms for optimum design of reinforce con-
design: Dissertations No. 675. Linköping University. crete frames, International Journal of Civil Engineering 9(3):
193–206.
Babaei, M. 2015a. Exploring practical optimal topology for rein-
forced concrete moment resisting frame structures, Ameri- Kaveh, A.; Sabzi, O. 2012. Optimal design of reinforced concrete
can Journal of Civil Engineering 3: 102–106. frames using big bang big crunch algorithm, International
https://doi.org/10.11648/j.ajce.20150304.12 Journal of Civil Engineering 10(3): 189–200.
Babaei, M. 2015b. The economical effect of ductility levels on Kim, I.Y.; De Weck, O. 2005. Adaptive weighted-sum method
reinforced concrete frames design, American Journal of Civil for bi-objective optimization: Pareto front generation, Struc-
and Structural Engineering 2: 1–6. tural and Multidisciplinary Optimization 29(2): 149–158.
Babaei, M.; Jabbar, M. 2016. Optimal intermediate steel moment https://doi.org/10.1007/s00158-004-0465-1
resisting frames with different spans and story numbers, Park, H. S.; Kwon,  B.; Shin,  Y.; Kim,  Y.; Hong,  T.; Choi, S.W.
Journal of Engineering and Applied Sciences 10(8): 223–226. 2013. Cost and CO2 emission optimization of steel rein-
Babaei, M.; Omidi, F. 2015. Determining the optimum spans for forced concrete columns in high-rise buildings, Energies 11:
special steel moment resisting frames with special eccentric 5609–5624. https://doi.org/10.3390/en6115609
braces, Research Journal of Applied Sciences 10(9): 474–478. Park, J. H.; Jeon, J. H.; Park, H. S. 2012. Optimization of CO-
Baling, R. J.; Yao, X. 1997. Optimization of Reinforced Con- 2emissions and cost for composite building design with NSGA-
crete Frames, Journal of Structural Engineering 123: 193–202. II, Creative Research Support Project (No. 2011-0018360).
https://doi.org/10.1061/(ASCE)0733-9445(1997)123:2(193) Sanaei,  E.; Babaei, M. 2012. Topology Optimization of struc-
Camp, C. V.; Pezeshk, S.; Hansson, H. 2003. Flexural design of tures using cellular automata with constant strain triangles,
reinforced concrete frames using a genetic algorithm, Jour- International Journal of Civil Engineering 10(3): 179–188.
nal of Structural Engineering 129: 105–115. Sarker, R; Liang, K.-H.; Newton, C. 2002. A new multiobjec-
https://doi.org/10.1061/(ASCE)0733-9445(2003)129:1(105) tive evolutionary algorithm, European Journal of Operational
Carrillo, V. M.; Taboada, H. 2012. A post-Pareto approach Research 140(1): 12–23.
for multi-objective decision making using a non-uniform https://doi.org/10.1016/S0377-2217(01)00190-4
weight generator method, Procedia Computer Science 12: Stanković,  T.; Štorga,  M.; Marjanović, D. 2012. Synthesis of
116–121. https://doi.org/10.1016/j.procs.2012.09.040 truss structure designs by NSGA-II and NodeSort algo-
Deb, K. 2001. Multi-objective optimization using evolutionary rithm, Strojniški vestnik – Journal of Mechanical Engineering
algorithms. John-Wiley & Sons. 58(3):203–212. https://doi.org/10.5545/sv-jme.2011.042

Mehdi BABAEI is an Assistant Professor in the Department of Civil Engineering at the University of Zanjan. He received his PhD
(2012) and Master’s (2003) degrees in Structural Engineering from the Iran University of Science and Technology after obtaining
his Bachelor’s in Civil Engineering in 2000. During his PhD study, he visited the Oxford University as a research student as well
as participated in a short program at the Massachusetts Institute of Technology in 2010. His research has focused on the optimal
design of structures, tall buildings, spatial structures, structural systems and sustainable design.

Masoud MOLLAYI earned his MSc in Structural Engineering from the University of Zanjan in 2015, after receiving his Bachelor
in Civil Engineering. His main interest is to study heuristic algorithms and to develop new algorithms for structural optimization.