Applied Soft Computing 110 (2021) 107646

Contents lists available at ScienceDirect

Applied Soft Computing

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

Optimum geometrical pattern and design of real-size diagrid

structures using accelerated fuzzy-genetic algorithm with bilinear
membership function
∗
Payam Ashtari, Roghaye Karami, Salar Farahmand-Tabar
Department of Civil Engineering Eng., Faculty of Engineering, University of Zanjan, Zanjan, Iran

graphical abstract

article info a b s t r a c t

Article history: Diagrids are the efficient systems of tube structures for tall buildings. One of the design considerations
Received 12 October 2020 for these structures is the geometrical pattern of the system. In this paper, a new method of fuzzy-
Received in revised form 9 February 2021 genetic algorithm based on bilinear membership functions is proposed with an improved crossover
Accepted 19 June 2021
operator and penalty function. The method is applied on tall buildings with a diagrid system to
Available online 24 June 2021
find the optimum geometrical patterns and the overall structural weight. Various three-dimensional
Keywords: diagrid structures with 24, 36, 42, 56, and 60 stories and different slenderness ratios are analyzed
Structural optimization under gravity and wind load. Then the effects of variation in the number of bays (4, 6, and 8) are
Tall buildings investigated and compared with each other. The results show that by increasing the dimension of the
Diagrid system structure, the structural weight is reduced up to 33% in some cases. However, the obtained angle of the
Optimum geometrical patterns diagrid members (range of 63 to 79 degrees) is increased by increasing the number of stories and the
Genetic algorithm height of the structure. The optimum weight and geometrical pattern of the models is obtained and a
Fuzzy concept
formulation is extracted from the results regarding the optimum angle of a diagrid system. Considering
GA, results show the merit of the accelerated fuzzy-genetic algorithm regarding the convergence and
the avoidance of being trapped in local minimum.
© 2021 Elsevier B.V. All rights reserved.

1. Introduction
∗ Corresponding author.
E-mail address: farahmandsalar@znu.ac.ir (S. Farahmand-Tabar). The diagrid system is an efficient structural system to con-
struct free-form structures, having an effective lateral load resis-
https://doi.org/10.1016/j.asoc.2021.107646 tance. As a new structural system with complicated nodes, only
1568-4946/© 2021 Elsevier B.V. All rights reserved.
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

limited experiments with the construction and manufacturing of Lee et al. [35]. The material layouts (diagrid angle and topology)
the system have been carried out. Thus, the constructional cost of diagrid frames using computer-oriented topology optimization
of these systems is still high in comparison with the alternative was investigated by Lee et al. in which eigenfrequency for re-
systems. Structural diagrid system has been broadly utilized in sisting dynamic responses was maximized and compliance for
recent years owing to their structural efficiency and aesthetic static responses were minimized [36]. A smart optimization task
potential. In fact, these structures are known as the last muta- to measure the variation effects of optimal geometric configu-
tion in tube structures. Diagrid structures are advantageous in rations like topology and shape of the diagrid was provided by
reducing the shear deformations by the axial resistance of the Lee et al. [37]. A simple stiffness-based size optimization and
inclined members. Diagrids are designed using members with design of diagrid tall buildings was investigated by Gerasimidis
tension and compression resistance; Therefore, most diagrids are et al. [38]. Size optimization of diagrid patterns for tall build-
made of steel. The structures with inclined members save 20% in ings was carried out using a single-objective genetic algorithm
material usage in comparison with the Moment Resisting Frames by Tomei et al. [39]. Kim and Kang used the multi-objective
(MRF). This system can be used for structures up to 100 stories. GA to control and design the diagrid structure under wind and
One of the oldest examples of these structural systems is the 13- earthquake [40].
story building of IBM in Petersburg which was built in 1963. Other Although GA has several advantages in optimizing the prob-
examples of diagrid structures are Hearst tower in New York, lems, it has several disadvantages such as premature convergence
Lotte tower in Korea, and Swiss Re building in London. on local optimum that leads researchers to propose and use the
Developments of the structural systems of tall buildings, es- improved or hybrid version of the GA. In this regard, combining
pecially diagrid systems, have been investigated by different re- fuzzy rules to the GA can improve its performance and cover its
searchers [1–7]. The case studies and review article were also disadvantages.
presented for the diagrid system in tall buildings [8,9]. There In this research, the size and topology of diagrid structures
are several research studies related to the behavior of the di- were optimized using the accelerated fuzzy-genetic algorithm
agrid systems such as lateral stiffness of high-rise diagrid tube with bilinear membership function and modified cross over re-
structures [10], concrete diagrid connections [11,12], shear lag ef- lated to different-length chromosomes. The angles of diagrid
fect [13], stress concentration [14], progressive collapse [15], and units and the structural cross-sections (beams and inclined dia-
performance-based plastic design of the diagrid structures [16]. grid members) are the variables of the optimization problem with
The cyclic behavior of diagrid nodes was investigated by Kim the objective function of the structural weight. Demand to ca-
et al. [17]. Also, the seismic performance of structures with di- pacity ratios (D/C) of the members, top displacement, inter-story
agrid systems have been studied [18–24]. drifts, and the constructability and compatibility of the size of two
For preliminary design of the diagrid tube structures, the columns in two adjacent stories are considered as the constraints
diagonal arrangement is a prerequisite. Several efforts have been of the problem. The bilinear membership function is proposed
done on this subject such as gradually varying angles of straight and utilized for the objective function and constraints that help
diagrid tube structures that were studied by Zhang et al. using a accelerating the convergence of the genetic algorithm. A method
graphic approach [25]. The arrangements of curved and varying- is proposed for uniform crossover operation of chromosomes
angle straight diagrid structures were investigated by Zhao and with different length. Furthermore, a penalty method is proposed
Zhang [26]. Angelucci and Mollaioli used non-uniform pattern and used in the optimization problem. Various structural models
configuration generated from topological assessments to change with different slenderness ratios and different stories (24, 36, 42,
the pattern configuration [27]. Geometrical patterns were studied 56, and 60) and bays (4, 6, and 8) are considered. The models
by Montuori et al. as an alternative design strategy for diagrid are optimized under gravity and wind load to reach the optimum
buildings. Eight patterns for the diagrid system (with and without angle of the diagrids and the optimal structural weight. According
secondary bracings) were proposed and assessed under gravity to results, the accelerated Fuzzy-Genetic algorithm outperforms
and wind load [28]. the GA regarding the convergence and the overall optimum
Finding the best pattern or arrangement of diagrids requires solution.
the study on the effects of the variations in diagrids’ angle.
Therefore, to find the best arrangement which means the better 2. Structural optimization problem
performance of the system and the structure, using the optimiza-
2.1. Objective function and constraints
tion process would be meaningful. Optimum structural design is
aimed to find the best plan to make the full use of mechanical
The process of optimization in this study involves the im-
characteristics and to use the best structural state. In diagrid
proved Fuzzy-Genetic algorithm with 30 population and the max-
structures, the optimization improves the structural capacity and
imum iteration of 350 to design the real-size tall building with
reduces the material usage by considering the design method
a diagrid system. The objective function and the parameters are
and parameters such as the pattern and the angle of the diagrid
defined based on the geometry and properties of the structure.
members. The optimization of tall buildings with diagrid systems
The optimum diagrid pattern (angle) and the structural weight
can be studied from several points of view: Shape, topology and
optimization are carried out using the metaheuristic of Fuzzy-
size optimization.
Genetic with the modified fuzzification, crossover, and penalty
Genetic algorithm and its modified and hybrid forms such as
function. The objective of this study is to find a vector of integer
fuzzy-genetic, hybrid fuzzy-genetic have been used for optimiza-
values (Eq. (1)) including the number of the listed sections for the
tion of structures and tall buildings [29–33]. GA-based optimiza-
elements, and the geometrical pattern of the diagrid members. To
tion of tall diagrid buildings considering the effect of geometric
determine the range of angles, the number of modules that can
modifications were carried out by Mirniazmandan et al. [34].
be placed at the height of the structure is considered so that the
Rhino software with its plug-in Grasshopper and Karamba were
range of angles in examples is between 27 to 88 degrees. It should
used for modeling and structural analysis respectively with the
be noted that the joints in the diagrid system should be placed on
objectives of minimum total structural weight and the horizontal
the floors. As a result, the range of angles is discrete and it is a
displacement of the top floor. The optimal range for diagrid
function of the dimension of the structure.
angle was achieved between 53◦ and 70◦ . The topology opti-
mization of the steel diagrids for tall buildings was studied by I Total = [I1 , I2 , I3 , . . . , In ] (1)
2
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

The objective function for weight minimization of the structure Table 1

Wind Load Parameters (ASCE 10-16).
is as follows:
Wind dir. angle 0
Nt
Windward 0.8
Pressure Coeff., Cp
∑
W = γi Ai Li (2) Leeward 0.5
Wind speed (m/s) 47
i=1
Exposure type B
where, Li , Ai , and γi is the length of the member, cross-section Importance factor 1
area of sections, and unit weight of the steel section assigned Topo. factor, Kzt 1
Gust factor, G 0.85
to group i of members respectively. Nt is the total number of Directionality factor 0.85
members. Members with similar characteristics are classified in
specific groups to reduce the computational cost. The strength,
constructability, and serviceability constraint are considered for
the current problem. Members assigned to flexure and compres- 2.2. Numerical model: Real-size diagrid structure
sion should satisfy Eqs. (3) and (4) based on AISC-LRFD design
In this study, three-dimensional building models with 24, 36,
method. The structural sections violating the strength limit-states
42, 56, and 60 stories and dimensions of 15, 21, and 27 m have
of this constraint, are penalized. been optimized. Also, to investigate the effects of the number
of bays, models with 4, 6, and 8 bays with 21 m dimension
[ ] [ ] [ ]
Pr 8 Mrx Mry Pr
+ + ≤ 1.0 for ≥ 0.2 were considered. The considered examples are symmetric in the
ϕc Pn 9 ϕb Mnx ϕb Mny ϕc Pn
plan with the typical story height of 3.5 m. Steel material with
(3) elasticity modulus of 2.3e08 kN/m2 , yield stress of 2.8e05, and
weight per volume of 76.98 kN/m3 is considered. Dead and live
[ ] [ ]
Pr Mrx Mry Pr
+ + ≤ 1.0 for < 0.2 load of 7 and 4 kN/m2 are assigned to the structure, respectively.
2ϕc Pn ϕb Mnx ϕb Mny ϕc Pn
For each design example, a list of AISC standard steel box sections
(4)
is considered. A group is defined to beams and inclined members
where, Pr and Mr are the required axial and flexural strength, of each diagrid unit. Fig. 1 illustrates the three-dimensional view
ϕc Pn and ϕb Mn are the axial and flexural design strength which of a typical model. To make a meaningful comparison of results,
are determined according to AISC-LRFD. Subscripts x and y are the results of the structure are investigated by the parameter
related to major and minor axes of bending, respectively. of height to width ratio (H/B) known as slenderness ratio. The
Generally, the occupants’ comfort in tall buildings strongly considered design load combination is based on ASCE-LRFD and
depends on serviceability limit-states that govern the design pro- is defined as:
cess. In consequence, the design method of tall buildings is cat- Q = 1.2D + 1.0L ∓ 1.0W (8)
egorized as displacement based. Thus, the constraint of drift is
Based on the analytical method of ACSE7-16, the wind load is
written as:
calculated and the building is assumed to be enclosed, and basic
Drift δf wind speed is considered as 47 m/s (∼ =105 mph). According to
Cf = −1≤0 (5)
δ limit
f
Table 1, by considering the following factors, the wind load is
achieved: Contribution of the Topographic condition (Kzt ), wind
where, δf is horizontal drift of the f th story (f = 1, 2, 3, . . . , n). directionality factor (Kd ), wind directionality factor (Ke ), velocity
The drift of tall buildings under wind load is limited to 0.002 pressure exposure coefficient (Kz ), Gust effect factor (G), closure
of the height of each story. According to Huang et al. [41], most classifications, velocity pressure (qz ), basic wind speed (V), and
of the codes also restrain the overall drift of the building, while pressure coefficients (Cp ). The velocity pressure at the height of z
it is not obvious whether the limit-states are related to inter- is calculated as follows:
story drifts or total drift. Thus, a new constraint is proposed. The qz = 0.613Kz Kzt Kd Ke V 2 (N/m2 ); V in m/s (9)
constraint of total drift is defined to obtain a faster convergence
as follows: The velocity pressure at mean roof height (qh ) is computed like
qz using Kz at mean roof height h. The pressure of the wind load
Drift_tot ∆i for lateral load-resisting system regarding enclosed or partially
Cind = −1≤0 (6)
∆limit
i enclosed buildings of all height is determined as follows:
Drift_tot
where, Cind is achieved for each individual in the population, P = qz GCp−windward − qh GCp−leeward (10)
∆i and ∆limit
i are the overall drift and its allowable value which
is equal to 1/500 of the total height of the building. 2.3. Verification of the analysis and the optimization
As a last constraint, the constructability constraint is defined
to reach the practical standards of tall buildings construction that In this section, the analysis of the model and performance
prevents construction of members with larger sections, above the of the proposed algorithm is validated by previous studies by
smaller ones in two adjacent stories or diagrid units. Without Montuori et al. [28] and Moon et al. [1]. The properties of the
this constraint, it is possible for the algorithm to choose and put used model structures are presented in Table 2. According to
a bigger section on the smaller section of the lower story or a Table 3, the analysis is verified by comparing lateral displacement
diagrid unit. This is not feasible in the construction. Therefore, it of the 90-story diagrid structures studied by Montuori et al. [28].
is defined as follows: Displacement curve illustrated in Fig. 2 indicates the reliability
top.
of the analysis. Then, the diagrid angle and the corresponding
i,Const di weight obtained by the optimization process using Accelerated
CIel = .
−1≤0 (7)
dbot Fuzzy-Genetic algorithm is verified considering the formulations
i
presented by Moon et al. [1] for the 42-story diagrid struc-
where, dibot . and di are the height of the column sections located
top
tures. Again, results show the reliability of the analysis and the
on the bottom and top floors, respectively. optimization process.
3
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

Fig. 1. Typical model of tall building with diagrid system (each triangular diagrid element is considered as two bays).

Table 2 Table 3
Properties of the verification models. Verification of the analysis.
Properties Montuori et al. [28] Moon et al. [1] Models Reference Present Diff (%)
study
E (kN/m2 ) 2.1e08 2.3e08
Fy (kN/m2 ) 275 000 360 000 Proposed formula Sap2000
Dimensions (m) 53×53 36×36 1 [28] Lateral disp. (m) 0.65 0.68 0.046
No. stories 90 42
Story height (m) 3.9 4 2 [1] Lateral disp. (m) 0.3 0.29 0.33 3.71
Wind speed (m/s) 50 49.1 Weight (ton) 1620 1850 1652.3 0.02
Diagrid angle (◦ ) 61.1 63.4
Dead & live loads (kN/m2 ) 7 & 4 3 & 2.4

function and constraints. Then, considering the utilized examples,

a penalty function and crossover operator for different-length
chromosomes are proposed.

3.1. Proposed membership function

A phenomenon can become either true or false based on

classic (or crisp) logic which can be mapped to a real number
of 0 or 1 as false or true, respectively. The conditions for struc-
tural optimum design will be complicated while employing the
crisp logic because large amounts of uncertainty and imprecision
interfere in the problem. For example, in crisp logic, if σ u is the
upper bound, σ < σ u becomes acceptable and σ < σ u + ∆σ be-
comes unacceptable. However, the transition between inclusion
and exclusion is considered in fuzzy logic using a membership
Fig. 2. Lateral displacement of the model and reference [28] under wind load. function, µY , which map each member of the set X to a fuzzy set
Y as follows:
µY (x) ∈ [0, 1]
{
(11)
3. Proposed optimization method: Accelerated Fuzzy-Genetic x ∈ X
Algorithm
In structural optimization, the membership function in crisp and
fuzzy logic for stress constraint is illustrated in Fig. 3 as an
The Genetic Algorithm (GA) and its fuzzified version (Fuzzy- example. For element i in crisp logic, the value of membership
GA) has been widely utilized in the optimization applications [42– is equal to 1 if σi < σ u and if the constraint is not satisfied, it
61]. In this study, several modifications in three parts of the is equal to zero. But for fuzzy logic, the membership value (α )
Fuzzy-GA are proposed to improve its performance. Initially, is varied from 0 to 1 so that α = 1 means the constraint is
the bilinear membership function is applied to the objective satisfied and α = 0 is related to the violation of the constraint
4
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

Fig. 3. The comparison of crisp and fuzzy logic in membership function (stress
constraint).

Fig. 5. Simple and modified bilinear membership functions of the objective

function and constraints.

Fig. 4. Fuzzy domain membership function.

from desired tolerance. Therefore, 0 < α < 1, means that the

constraint is satisfied to the respective amount.
Fuzzification is applied to both inequality constraints (con-
stricts without being strictly satisfied) and the objective function
in fuzzy optimization with the membership function is expressed
as follows
{m }
⋂
µD (x) = µF (x) ∩ µgi (x) (12)
i=1

where, µF (x) is the membership function for the objective func-

tion and µgi (x) is the membership function for the ith inequality
constraint. By using a max–min procedure, xopt is achieved from
the region as:
µD xopt = max µD (x)
( )
(13)
{ }
µD (x) = min µF (x) , min µgi (x) (14)
i=1,...,m
Fig. 6. A typical diagrid unit.
According to Fig. 4, a fuzzy domain in the membership func-
tion is achieved by intersecting the membership functions of
constraints and the objective function. This max–min procedure ⎧ ⎫
nod ]2 ⎬
λ
[
can be solved by maximizing an overall satisfaction parameter 1
⎨∑
(λ). Thus, the optimization problem can be stated as:
+ γj − 1 + ωj
2⎩ µδ j
maximize λ subject to
⎭
j=1
{ nm ] }
2
λ
[
λ ≤ µF (x) (15) 1 ∑
+ γk − 1 + ωk (18)
2 µgk
λ ≤ µgi (x) ; i = 1, . . . , m (16) k=1

0≤λ≤1 (17) where, Ls is a scale factor; α and β are Lagrangian penalty func-
tion and constraint parameters for the membership function of
By the combination of the fuzzy-logic with the augmented the objective function. γ and ω are similar to α and β for the
Lagrangian GA, a formulation is proposed as an optimization membership function of the constraints. So, it is aimed to maxi-
method. mize λ considering the constraints of (15) and (16).
To minimize Generally, stress and displacement are two types of con-
{ [ ]2 }
1 λ straints in structural optimization problems. Considering the
ϕ (x, y) = −Ls λ + α −1+β membership function for each constraint, Eq. (16) can be ex-
2 µF pressed as:
{ noe [ ]2 }
1 ∑ λ λ ≤ µuσi (x) ; i = 1, . . . , noe (Stress constraint)
+ γi − 1 + ωi (19)
2 µaσ i λ ≤ µlσi (x)
i=1

5
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

Fig. 7. Increasing the parent chromosome’s length.

λ ≤ µδj (x) ; j = 1, . . . , nod (Displacement constraint) function for the objective function is illustrated in Fig. 5.a. How-
(20) ever, it may have several problems. The probability of a fast
convergence to a local optimal solution increases if the Fuzzy-GA
λ ≤ µgk (x) ; k = 1, . . . , nm (Constructability constraint) procedure is applied after the process of simple GA to improve the
(21) solution [29]. On the other hand, if F ′′ is equal to the best solution
of GA, we typically have F > F ′′ in the initial iterations (gen-
where, µlσi (x), µuσi (x), are the membership functions for lower
erations) of the fuzzified GA, and the value of membership (µF )
and upper stress limits, and µdj (x) is the displacement’s mem-
becomes zero for majority of chromosomes, and it is not possible
bership function. noe, nod, and nm are the number of elements,
to compare them with each other. Thus, a bilinear membership
constrained degrees of freedom, and diagrid units, respectively.
function (Fig. 5.b) can be utilized for increasing the rate of the
The last two terms in Eq. (18) can be removed if λ is equal to
convergence. In Fig. 5.b, Fu ≤ F and If membership function No. 2
min(µF (x) , mini=1,...,noe µaσi , minj=1,...,nod µδj , mink=1,...,nm µgk ) in- is considered, then the slope of the line becomes less than line No.
stead of a random value, and the problem is changed to an 1 which means that the variations of the chromosomes’ fitness
unconstrained optimization problem as values are decreased (same number of candidates for the best
maximize ϕ (x) = λ (22) and the worst chromosomes in next generation). Therefore, the
{ probability of trapping the algorithm in a local optimum solution
µuσi ; if σi > 0 will be increased and consequently we encounter the reduced
µσi =
a
(23)
µlσi ; if σi ≤ 0 convergence rate.
In Fig. 5.b, F ′′ can be chosen to the mean value of F ′ and Fu
First, design variables (x) are assigned to each chromosome or as the best value of previous works. The membership function
by GA. Then, membership values for the objective functions and for the constraints can be selected simply using their permissible
constraints are calculated through structural analysis. By using values (ga ). In current study, the assigned membership function to
Eq. (22), the fitness value is obtained for each chromosome. the constraints is shown in Fig. 5.c, in which, ∆g is the magnitude
For determining an answer, it is important to choose a proper of the constraint relaxation, and gu is equal to n × ga that n > 1.
membership function for the objective function. If F ′ and F ′′ are
assumed to be the lower and upper limit of the objective function 3.2. Proposed fitness function and penalty function
for minimization problems, then
The fitness function was proposed in the previous section
1; if F ≤ F ′
{
µF = (24) based on the membership function. For better convergence, an
0; if F > F ′′ appropriate penalty function is required. So, a new penalty func-
A desired membership function can use the membership value tion is proposed in order to achieve faster convergence rate, and
(F ) in the range of F ′ < F < F ′′ . The F ′′ can be estimated by it is added to Eq. (22) which is defined as
design experience or through several iterations of the simple GA. Penalty Function = max (0, 0.1/µF (x) − 1)
To achieve the value of the lower bound (F ′ ), F ′′ can be reduced to noe
0.1
∑ ( )
minimum value or a certain percentage of the objective function + max 0, −1
even if the violated constraints appear. The linear membership µaσ i (X )
i=1

6
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

3.3. Proposed crossover operator

In the present study, the chromosome length is a function of

optimization variable as follows:
nstory
Lc = log 2np + × (log 2pd + log 2pb ) (28)
nst
where, nstory and nst are the number of total stories and the num-
ber of stories covered by a diagrid unit, respectively. pd and pb
are the number of the considered sections for inclined members
and beams, respectively. np is number of the possible cases that a
diagrid unit can covers the stories. A diagrid unit is illustrated in
Fig. 6 in which the angle of the inclined members is related to the
length of the bay and the number of stories covered by an inclined
member. To find the best topology in diagrid structures which
leads to minimum weight, the whole possible angles should be
included in a generation which is carried out by considering the
log 2np gene. Considering this gene, the number of stories covered
by an inclined member will be varied in different chromosomes
and all of the possible cases for the structure will be investigated.
Therefore, the length of the chromosome in a generation will be
different. The coefficient of the second part of Eq. (28) represents
the number of diagrid units. In this study, a group is defined
to beams and the inclined members of each diagrid unit, sepa-
rately. The number of groups can be evaluated by considering the
value of log 2np for each chromosome and the number of stories
covered by a diagrid unit.
It is not possible to do the crossover using the common uni-
form combination since the length of the parent chromosomes
Fig. 8. The flowchart of Accelerated Fuzzy-Genetic Algorithm. are not identical. Also, by combining the first several blocks
of two parent chromosomes (number of units and groups) the
length of the child chromosomes will not be the same as parents.
nod
0.1 To solve this problem a method is proposed. First, the genes
∑ ( )
+ max 0, −1 related to the number of diagrid units in both selected parents
µδj (X )
j=1 are combined and then mutation is applied; after decoding, the
mn
∑ (
0.1
) number of diagrid units for each child and the corresponding
+ max 0, −1 (25) chromosome length is defined. Then, to generate each child, the
µgk (X ) length of both parents is compared to the children. If the length
k=1
of the child is shorter than both parents, the crossover is applied
Thus, the lower limit of membership value varies from zero to a until the considered length of the child is achieved. However, if
small value (ε = 10−5 ), as shown in Fig. 5.d, in which, Y is the the length of one or both parents becomes shorter, the short-
value of constraint or the objective function. Therefore, the new length parents should be repeated until their lengths become
fitness function is expressed as identical to children’s.
After the number of diagrid units in each chromosome is
Maximize ϕ (x) = −Ls λ − c × Penalty (26) achieved, there are genes related to beams and inclined members
(Fig. 7.a). If the genes of beams and inclined members in each unit
where, Ls and c are the scale and penalty coefficient, and the value is considered as a single block, the process of increasing length
of each chromosome is continued until the length of parents
of the λ is calculated as follows:
and the children’s become identical. In this method of length
( )
increment and considering the length variation of the parent and
λ = min µF , min µafi , min µδj , min µgk
( ) ( ) ( )
(27) child chromosomes, a repetition parameter is applied to each
i=1,...,noe j=1,...,nod k=1,...,mn
block of the parent chromosome.
The difference between the worst and the best chromosomes’ The length increment is done block by block starting from the
fitness is increased by adding the penalty function in comparison first one. Initially, each block is repeated once and the repetition
process is terminated until the lengths become identical. If the
with those of Eq. (22) in which the chromosomes’ fitness are
parent length is shorter after one cycle of the whole blocks’
achieved only by considering the lower value of their member- repetition, the second cycle of block increment is started. Instead
ships. Thus, the algorithm will be accelerated and converged of the mentioned method, if the total length of the parent chro-
to the global optimum because the best chromosomes provide mosome is repeated all together from its end, the block related
more candidates in the further generations. This means, if each to the last unit has a weak section and the block related to
chromosome’s fitness is achieved from Eq. (26) (instead of Eq. first units (placed after the block of the last unit during the
process of length increasement) has the strong section. While
(22)), due to the penalty function, the chromosomes’ fitness with
crossover, these blocks are placed in the block related to the
λ ≥ 0.1 becomes much greater in comparison with the case of middle units in which the medium sections are needed. Thus, the
λ < 0.1 (fast convergence due to great number of candidates in convergence becomes slower. But in the proposed method, each
the mating pool). block is repeated immediately for the required number which is
7
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

defined by the length differences. The convergence is accelerated angle and the overall weight of the structure. The algorithm was
since similar sections are required in the adjacent units. Parent implemented on the considered models with different stories
chromosomes with different length and the process of increasing and slenderness ratios (H/B). Considering the displacement and
length are illustrated in Fig. 7.b and .c. Demand to capacity (D/C) constraints the optimum solutions
The pseudocode and flowchart of the proposed Fuzzy-Genetic were obtained. According to Tables 4 and 5, the optimum cross-
algorithm is presented in Algorithm 1 and Fig. 8, respectively. sections, structural weight, and the optimum angles for diagrid
systems are presented.
4. Results and discussion First, the effect of the slenderness ratio is investigated con-
sidering the results of building with different stories and various
According to the previous sections, the Accelerated Fuzzy- dimensions of plan (Table 4). In the 24-story building, the length
Genetic algorithm was proposed to optimize the diagrid system’s of B1, B2, and B3, as the plan’s dimensions, are 9, 15, and 21 m

Fig. 9. Convergence history of the diagrid structures using Accelerated Fuzzy-GA algorithm.

8
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

Table 4
Results obtained from optimization of the diagrid models with different stories and dimensions (6 bays).
Diagrid unit 24 story 36 story 42 story 56 story 60 story
B = 9 B = 15 B = 21 B = 15 B = 21 B = 27 B = 15 B = 21 B = 27 B = 15 B = 21 B = 27 B = 15 B = 21 B = 27
Da 40×1.5 35×1.5 40×1 45×3 45×2.5 55×2 50×4.5 60×3 65×2.5 80×7.5 75×7 65×5 80×8 80×6.5 75×5
1
Bb 10×1 20×1 25×1 20×1 30×1 35×1 20×1 30×1 35×1 25×1 25×1 35×1.5 20×1 30×1 35×1.5
D 40×1 35×1 35×1 35×3 40×2.5 55×1.5 50×4 60×2 60×2 80×7.5 75×5.5 60×5 80×8 70×6.5 65×5.5
2
B 10×1 20×1 25×1 20×1 30×1 35×1 20×1 30×1 35×1 25×1 25×1 35×1.5 20×1 30×1 35×1.5
D 40×1 30×1 30×1 35×3 35×2 45×1.5 40×4 60×2 45×2.5 70×7 70×6 60×5 80×7.5 70×6.5 65×5.5
3
B 10×1 20×1 25×1 20×1 30×1 35×1 20×1 30×1 35×1 25×1 25×1 35×1.5 20×1 30×1 35×1.5
D 40×1 30×1 25×1 35×3 35×2 45×1.5 40×4 60×2 45×2.5 70×7 55×5 60×5 80×7.5 70×6.5 65×5
4
B 10×1 20×1 25×1 20×1 30×1 35×1 20×1 30×1 35×1 25×1 25×1 35×1.5 20×1 30×1 35×1.5
D 40×1 25×1 25×1 35×2.5 35×1.5 45×1 40×4 35×3 45×1.5 70×7 55×5 55×5 80×5 70×6 65×4
5
B 10×1 20×1 25×1 20×1 30×1 35×1 20×1 30×1 35×1 25×1 25×1 35×1.5 20×1 30×1 35×2
D 40×1 20×1 20×1 30×1.5 30×1 40×1 35×3 35×2.5 40×1.5 60×5 55×4.5 50×5 80×4.5 70×4 60×2
6
B 10×1 20×1 25×1 20×1 30×1 30×1.5 20×1 30×1 35×1 25×1 25×1 35×1.5 20×1 30×1 35×2
D 40×1 35×1.5 35×1.5 40×1 55×2 50×4 45×2.5 80×4.5 70×4
7
B 10×1 20×1 30×1 35×1 25×1 25×1 35×1.5 20×1 30×1
D 20×2 45×2.5 80×4.5 70×3
8
B 10×1 25×1 20×1 30×1
D 20×2 65×4.5 70×2.5
9
B 10×1 20×1 30×1
D 15×2 55×2.5 35×2.5
10
B 10×1 20×1 30×1
D 15×2
11
B 10×1
D 15×2
12
B 10×1
Opt. weight 250.2 233.3 252.9 836.7 720.3 828 1457.3 1276.2 1215.3 5415.6 4021.1 3625.3 6050.5 4828.1 4356.3
Opt. angle 66.8◦ 70.4◦ 63.4◦ 76.6◦ 71.6◦ 66.8◦ 76.6◦ 71.6◦ 66.8◦ 79.9◦ 74.1◦ 72.2◦ 76.6◦ 71.5◦ 75.6◦
Conv. Iter. 112 132 134 159 126 136 194 161 156 207 267 216 203 207 214
a
Diagrids.
b
Beams.

respectively; In other cases, these values become 15, 21, 27 m.

As it is shown, in the 24-story building, the minimum weight
was obtained in the structure with a dimension of 15 m. In other
words, the 24-story building with the slenderness ratio (H/B) of
5.6 and the angle of 70.3 degree is the optimum case. In the 36-
story building, the 21 m dimension, or the slenderness ratio of 6
with the angle of 76.1 degree is the optimum one. Likewise, in 42-
,56-, and 60-story buildings with 27 m dimension or slenderness
ratios of 5.44, 7.529, 7.78 with degrees of 66.8, 72.2, and 75.6,
respectively are the optimum cases.
After considering the effects of slenderness ratio, it is kept con-
stant and the effects of variation in the number of bays are inves-
tigated. According to Fig. 9, the convergence history of the struc-
tures is illustrated based on the classifications of Tables 4 and 5;
It means that the convergence of the models with same number
of stories/bays but different slenderness ratios are plotted beside
each other to make a comparison.
To show the efficiency of the proposed Fuzzy-Genetic algo-
rithm, the convergence history of all structures (24-story to 60-
story models) are presented in Fig. 10 comparing with con-
vergence histories achieved by simple Genetic Algorithm. Re-
markable variations are seen in all diagrams including better
convergence rate and final optimum solution. The premature
convergence and being trapped in local minimum is obvious in
the results of simple GA.
Table 6 contains the results of the optimum design of the
considered diagrid structures, such as forces, displacements and
demand to capacity ratios of sections. It is obvious that by in-
creasing the height of the structure the base shear and base
moment is increased. The values of the inter-story drifts, as
shown in Fig. 11, indicate that the displacements are determinant
in the design process in most cases that is common in the design
of tall buildings.
As it can be seen from Figs. 12 and 13, in 42, 56 and 60-
story structures, the weight of the structure decreases with an
increase of plan’s dimension, despite the increase in the length
9
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

Table 5
Results obtained from optimization of the diagrid models with different stories and bays (B = 21 m).
4-bay 6-bay 8-bay
Var. 36 St. 42 St. 56 St. 36 St. 42 St. 56 St. 36 St. 42 St. 56 St.
Da 55×3 60×5 80×7.5 45×2.5 60×3 75×7 50×2 55×3 60×6
1
Bb 40×1 40×1 45×2 30×1 30×1 25×1 20×1 25×1 20×1
D 50×2.5 60×4.5 70×7 40×2.5 60×2 75×5.5 50×1.5 50×2.5 60×6
2
B 40×1 40×1 45×2 30×1 30×1 25×1 20×1 25×1 20×1
D 50×2 60×3 70×6 35×2 60×2 70×6 45×1.5 40×3 60×4.5
3
B 40×1 40×1 45×2 30×1 30×1 25×1 20×1.5 25×1 20×1
D 50×2 55×2.5 70×4 35×2 60×2 55×5 45×1.5 35×2.5 60×4.5
4
B 40×1 40×1 45×2 30×1 30×1 25×1 20×1.5 25×1 20×1
D 50×2 50×2.5 35×1.5 35×3 55×5 40×1 25×1 60×4.5
5
B 40×1 35×1.5 30×1 30×1 25×1 20×1.5 20×1.5 20×1
D 45×1.5 50×2.5 30×1 35×2.5 55×4.5 25×1 25×1 55×4.5
6
B 35×1.5 35×1.5 30×1 30×1 25×1 20×1.5 20×1.5 20×1
D 50×1.5 35×1.5 50×4 25×1 50×3
7
B 35×1.5 30×1 25×1 20×1.5 20×1
D 50×1.5
8
B 20×1
Opt. weight (ton) 800.1 1369.8 3844.6 720.3 1276.2 4021.1 823.9 1341.3 4318.8
Opt. angle 63.4◦ 63.4◦ 77.9◦ 71.6◦ 71.6◦ 74.1◦ 76◦ 76◦ 77.9◦
Conv. Iter. 129 92 221 126 161 267 183 165 183

a
Diagrids.
b
Beams.

Table 6
Results comparison of optimized diagrid structures.
Model No. Bays B (m) Base shear (kN) Base Moment (kN-m) Max. disp.(m) Max. inter-story drift Avg. D/C ratio
Beams Diagrids
9 2032.9 1079.4 0.12864 0.001991 0.73705 0.322599
24 St. 15 2980.1 1043.1 0.12084 0.001798 0.81461 0.546555
21 4136.6 988.48 0.10127 0.001678 0.84546 0.674499
15 4955.4 2816.6 0.21237 0.001969 0.66222 0.376115
36 St. 21 6891.8 2075.0 0.19028 0.001799 0.71155 0.616477
27 8817.7 2580.2 0.17571 0.001428 0.85527 0.669635
15 6084.8 4241.2 0.23530 0.001985 0.73770 0.270468
42 St. 6 21 8455.2 4918.9 0.22499 0.001901 0.72314 0.537357
27 10 812 4627.1 0.21266 0.001779 0.85224 0.700793
15 8718.3 19 119 0.37677 0.001996 0.81643 0.330027
56 St. 21 12 242 12 518 0.35076 0.001980 0.84752 0.370215
27 15 526 13 009 0.32658 0.001835 0.89014 0.526037
15 9806.2 20195.7 0.32443 0.001997 0.29287 0.316140
60 St. 21 13 242 13687.3 0.27780 0.001862 0.81950 0.370785
27 16 526 14209.3 0.21618 0.001472 0.87670 0.558409
36 St. 21 6227.4 2925.6 0.13027 0.001517 0.85179 0.779690
42 St. 4 21 7635.4 3770.0 0.20081 0.001846 0.91823 0.619629
56 St. 21 10 594 15 380 0.30441 0.0019964 0.86456 0.644735
36 St. 21 7224 4871.3 0.19408 0.001906 0.75693 0.512648
42 St. 8 21 8865.1 7103.4 0.23698 0.001977 0.76567 0.368877
56 St. 21 12 838 11593.0 0.32973 0.001998 0.78134 0.398218

of the beams and diagonal elements. This reduction is equal to

16% in 42-story structure, 33.1% in 56-story structure and 28% in
60-story structure; It can be stated that in structures with bigger
dimension of plan (lower slenderness ratio), due to the increase in
the gravity-induced torque arm, the resisting moment increases
and the drift due to the overturning moment decreases. As a
result, the need for large cross-sectional elements is decreased,
and the weight of the structure is reduced, subsequently. In 24
and 36-story structures, by changing dimensions, there is not a
remarkable change in the weight of the structure.
According to Fig. 14, in 6 and 8-bay structures with 36, 42
and 56 floors, as well as in 4-bay structures with 36 and 42
floors, the optimal angle does not change much with increasing
the height of the structure. But in a 4-bay structure, the optimal
angle increases with increasing the number of floors from 42
to 56. It can be said that in a 56-story structure, because the
Fig. 10. The comparison of the optimum results of GA and Accelerated height of the structure is more than the 36-story and 42-story
Fuzzy-GA. structures, with increasing lateral force, the overturning moment
10
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

Fig. 11. The inter-story drifts of the optimum models.

Fig. 12. Effects of dimension variation on the weight of 6-bay structure. Fig. 13. Height effect on the structural weight.

also increases. Because the number of elements is less in a 4- results are summarized and compared in Table 7. The optimum
bay structure, these diagonal elements will have a wider angle shapes of the diagrid models are illustrated in Fig. 16.
Finally, a formula can be extracted from the obtained results
to resist the overturning moment.
for optimum geometrical pattern (diagrid angle) of the diagrid
According to Fig. 15, the process of weight increment with in- structures. According to Fig. 17, the formulation is proposed
creasing the height is almost equal in different bays. The optimum using linear regression on the available data. This formula can
11
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

Table 7
Results summary: Optimum weight and optimum diagrid angles.
Story No. nbay = 4 nbay = 6 nbay = 8
B = 21 B = 15 B = 21 B = 21 B = 21
n θopt (◦ ) Wopt (ton) θopt (◦ ) Wopt (ton) θopt (◦ ) Wopt (ton) θopt (◦ ) Wopt (ton) θopt (◦ ) Wopt (ton)
36 63.4 800.10 76.6 836.7 71.6 720.3 66.8 828.0 76.0 823.9
42 63.4 1369.8 76.6 1457.3 71.6 1276.2 66.8 1215.3 76.0 1341.3
56 77.9 3488.6 79.9 4515.6 74.1 4021.1 72.2 3625.3 77.9 4318.8
60 – – 76.6 6050.5 71.6 4828.1 75.6 4356.3 – –

Fig. 14. Optimum diagrid angles for the structure with different bays and stories.

Fig. 17. Formula of the optimum angle of diagrids using linear regression.

be utilized in the preliminary design of diagrid systems in tall

buildings.

5. Conclusions

In the present study, the Genetic Algorithm with fuzzy concept

was used to optimize the weight of the diagrid structure and the
angles of its elements. A penalized fitness function was utilized
considering the proposed bilinear membership function which
were also applied to constraints. Also, a method was proposed
Fig. 15. Optimum weight for the structure with different bays and stories. for uniform crossover of chromosomes with a length diversity.
The diagrid structures were analyzed and optimized under the
gravity and wind load. The remarkable results are concluded as
follows:

Fig. 16. The optimum pattern of the models (optimum diagrid angles).

12
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

Through the improvements applied on the fuzzy-genetic al- [11] W. Zhou, J. Zhang, Z. Cao, Experiment and analysis on X-Shaped reinforced
gorithm using the bilinear membership function and improved concrete joint in diagrid structures, ACI Struct. J. 110 (2) (2013) 171–181,
http://dx.doi.org/10.14359/51684398.
crossover and penalty function, the obtained convergence im-
[12] W. Zhou, Z. Cao, J. Zhang, Experiment and analysis on reinforced concrete
proved remarkably. spatial connection in diagrid tube, Struct. Des. Tall Spec. Build. 25 (4)
By increasing the dimensions of the plan, the optimum weight (2015) 179–192, http://dx.doi.org/10.1002/tal.1234.
of the diagrid structure is decreased. Also, by increasing the [13] Q. Shi, F. Zhang, Simplified calculation of shear lag effect for high-rise
height of the structure, the effect of plan’s dimension on the diagrid tube structures, J. Build. Eng. 22 (2019) 486–495, http://dx.doi.org/
10.1016/j.jobe.2019.01.009.
weight reduction is increased. In other words, the percent of the [14] S.H. Lee, S.J. Lee, J.H. Kim, S.M. Choi, Mitigation of stress concentration in
weight reduction due to increasing the dimensions of the plan a diagrid structural system using circular steel tubes, Int. J Steel Struct. 15
become higher as the height of the structure increases in a way (3) (2015) 703–717, http://dx.doi.org/10.1007/s13296-015-9016-4.
that in the 56-story building with increment in the dimension, [15] J. Kim, J. Kong, Progressive collapse behavior of rotor-type diagrid build-
ings, Struct. Des. Tall Spec. Build. 22 (16) (2012) 1199–1214, http://dx.doi.
33.1% of the weight is decreased; However, this weight reduction
org/10.1002/tal.762.
is achieved 1% in the 36-story building. [16] T. Li, T.Y. Yang, G. Tong, Performance-based plastic design and collapse
The angle of the inclined diagrid elements is increased by assessment of diagrid structure fused with shear link, Struct. Des. Tall Spec.
increasing the height and the slenderness ratio of the structure, Build. 28 (6) (2019) e1589, http://dx.doi.org/10.1002/tal.1589.
varied between 63 to 79 degrees. [17] Y.J. Kim, I.Y. Jung, Y.K. Ju, S.J. Park, S.D. Kim, Cyclic behavior of diagrid
nodes with H-section braces, J. Struct. Eng. 136 (9) (2010) 1111–1122,
By decreasing the number of bays in the structure, the ele-
http://dx.doi.org/10.1061/(asce)st.1943-541x.0000203.
ments in the architectural view of the structure are decreased, [18] J. Kim, Y.H. Lee, Seismic performance evaluation of diagrid system build-
wide openings are created, and overall structural weight is con- ings, Struct. Des. Tall Spec. Build. 21 (10) (2010) 736–749, http://dx.doi.
sequently reduced. org/10.1002/tal.643.
A formula was extracted from the obtained results for the [19] J. Lee, J. Kong, J. Kim, Seismic performance evaluation of steel diagrid
buildings, Int. J. Steel Struct. 18 (3) (2018) 1035–1047, http://dx.doi.org/
optimum geometrical pattern (angle) of the diagrid structures 10.1007/s13296-018-0044-8.
which can be utilized in the preliminary design of the diagrid [20] S. Sadeghi, F.R. Rofooei, Quantification of the seismic performance factors
system in tall buildings. for steel diagrid structures, J Constr. Steel Res. 146 (2018) 155–168, http:
//dx.doi.org/10.1016/j.jcsr.2018.03.018.
[21] E. Asadi, Y. Li, Y. Heo, Seismic performance assessment and loss estimation
CRediT authorship contribution statement
of steel diagrid structures, J. Struct. Eng. 144 (10) (2018) 4018179, http:
//dx.doi.org/10.1061/(asce)st.1943-541x.0002164.
Payam Ashtari: Conceptualization, Methodology, Validation, [22] E. Asadi, H. Adeli, Seismic performance factors for low- to mid-rise steel
Supervision. Roghaye Karami: Methodology, Software, Valida- diagrid structural systems, Struct. Des. Tall Spec. Build. 27 (15) (2018)
e1505, http://dx.doi.org/10.1002/tal.1505.
tion, Formal analysis. Salar Farahmand-Tabar: Methodology, Val-
[23] M. Heshmati, A.A. Aghakouchak, Quantification of seismic performance
idation, Visualization, Writing - original draft, Writing - review & factors of steel diagrid system, Struct. Des. Tall Spec. Build. 28 (3) (2018)
editing. e1572, http://dx.doi.org/10.1002/tal.1572.
[24] M. Sohrabi-Haghighat, P. Ashtari, Evaluation of seismic performance factors
Declaration of competing interest for high-rise steel structures with diagrid system, KSCE J. Civ. Eng. 23 (11)
(2019) 4718–4726, http://dx.doi.org/10.1007/s12205-019-1546-4.
[25] C. Zhang, F. Zhao, Y. Liu, Diagrid tube structures composed of straight
The authors declare that they have no known competing finan- diagonals with gradually varying angles, Struct. Des. Tall Spec. Build. 21
cial interests or personal relationships that could have appeared (4) (2012) 283–295, http://dx.doi.org/10.1002/tal.596.
to influence the work reported in this paper. [26] F. Zhao, C. Zhang, Diagonal arrangements of diagrid tube structures for
preliminary design, Struct. Des. Tall Spec. Build. 24 (3) (2014) 159–175,
http://dx.doi.org/10.1002/tal.1159.
References [27] G. Angelucci, F. Mollaioli, Diagrid structural systems for tall buildings:
Changing pattern configuration through topological assessments, Struct.
[1] K.S. Moon, J.J. Connor, J.E. Fernandez, Diagrid structural systems for tall Des. Tall Spec. Build. 26 (18) (2017) e1396, http://dx.doi.org/10.1002/tal.
buildings: characteristics and methodology for preliminary design, Struct. 1396.
Des. Tall Spec. Build. 16 (2) (2007) 205–230, http://dx.doi.org/10.1002/tal. [28] G.M. Montuori, E. Mele, G. Brandonisio, A. Luca, Geometrical patterns
311. for diagrid buildings: Exploring alternative design strategies from the
[2] K. Jani, P.V. Patel, Analysis and design of diagrid structural system for high structural point of view, Eng. Struct. 71 (2014) 112–127, http://dx.doi.org/
rise steel buildings, Proc. Eng. 51 (2013) 92–100, http://dx.doi.org/10.1016/ 10.1016/j.engstruct.2014.04.017.
j.proeng.2013.01.015. [29] M. Yassami, P. Ashtari, Using fuzzy genetic algorithm for the weight
[3] K.S. Moon, Diagrid structures for complex-shaped tall buildings, Proc. Eng. optimization of steel frames with semi-rigid connections, Int. J Steel Struct.
14 (2011) 1343–1350, http://dx.doi.org/10.1016/j.proeng.2011.07.169. 15 (1) (2015) 63–73, http://dx.doi.org/10.1007/s13296-014-1105-2.
[4] G.M. Montuori, E. Mele, G. Brandonisio, A.D. Luca, Design criteria for [30] P. Ashtari, F. Barzegar, Accelerating fuzzy genetic algorithm for the op-
diagrid tall buildings: Stiffness versus strength, Struct. Des. Tall Spec. Build. timization of steel structures, Struct. Multidiscip. Optim. 45 (2) (2012)
23 (17) (2013) 1294–1314, http://dx.doi.org/10.1002/tal.1144. 275–285.
[5] D. Lee, S. Shin, Advanced high strength steel tube diagrid using TRIZ [31] M. Yassami, P. Ashtari, Using fuzzy genetic, artificial bee colony (ABC)
and nonlinear pushover analysis, J Constr. Steel Res. 96 (2014) 151–158, and simple genetic algorithm for the stiffness optimization of steel frames
http://dx.doi.org/10.1016/j.jcsr.2014.01.005. with semi-rigid connections, KSCE J. Civ. Eng. 19 (5) (2015) 1366–1374,
[6] E. Asadi, H. Adeli, Diagrid: An innovative sustainable and efficient struc- http://dx.doi.org/10.1007/s12205-014-0517-z.
tural system, Struct. Des. Tall Spec. Build. 26 (8) (2017) e1358, http: [32] F. Nouri, P. Ashtari, Weight and topology optimization of outrigger-braced
//dx.doi.org/10.1002/tal.1358. tall steel structures subjected to the wind loading using GA, Wind Struct.
[7] F. Fu, Bracing, diagrid, 3D space frame, and mega frame structural systems 20 (4) (2015) 489–508, http://dx.doi.org/10.12989/was.2015.20.4.489.
in tall buildings, Des. Anal. Tall Complex Struct. (2018) 137–175, http: [33] S. Farahmand-Tabar, P. Ashtari, Simultaneous size and topology optimiza-
//dx.doi.org/10.1016/b978-0-08-101018-1.00005-8. tion of 3D outrigger-braced tall buildings with inclined belt truss using
[8] E. Mele, M. Toreno, G. Brandonisio, A. Luca, Diagrid structures for tall genetic algorithm, Struct. Des. Tall Spec. Build. 29 (13) (2020) e1776,
buildings: case studies and design considerations, Struct. Des. Tall Spec. http://dx.doi.org/10.1002/tal.1776.
Build. 23 (2) (2012) 124–145, http://dx.doi.org/10.1002/tal.1029. [34] S. Mirniazmandan, M. Alaghmandan, F. Barazande, E. Rahimianzarif, Mutual
[9] C. Liu, Q. Li, Z. Lu, H. Wu, A review of the diagrid structural system effect of geometric modifications and diagrid structure on structural
for tall buildings, Struct. Des. Tall Spec. Build. 27 (4) (2017) e1445, http: optimization of tall buildings, Archit. Sci. Rev. 61 (6) (2018) 371–383,
//dx.doi.org/10.1002/tal.1445. http://dx.doi.org/10.1080/00038628.2018.1477043.
[10] C. Liu, K. Ma, Calculation model of the lateral stiffness of high-rise diagrid [35] D.K. Lee, U. Starossek, S.M. Shin, Optimized topology extraction of steel-
tube structures based on the modular method, Struct. Des. Tall Spec. Build. framed diagrid structure for tall buildings, Int. J Steel Struct. 10 (2) (2010)
26 (4) (2016) e1333, http://dx.doi.org/10.1002/tal.1333. 157–164, http://dx.doi.org/10.1007/bf03215827.

13
P. Ashtari, R. Karami and S. Farahmand-Tabar Applied Soft Computing 110 (2021) 107646

[36] D. Lee, J. Lee, J. Kim, U. Srarossek, Investigation on material layouts of [49] C.L. Yang, R.J. Kuo, C.H. Chien, N.T.P. Quyen, Non-dominated sorting
structural diagrid frames by using topology optimization, KSCE J. Civ. Eng. genetic algorithm using fuzzy membership chromosome for categorical
18 (2) (2014) 549–557, http://dx.doi.org/10.1007/s12205-014-0107-0. data clustering, Appl. Soft Comput. 30 (2015) 113–122, http://dx.doi.org/
[37] Y. Lee, J. Oh, H.H. Abdu, K.J. Young, Finite element analysis of optimized 10.1016/j.asoc.2015.01.031.
brace angle for the diagrid structural system, Int. J. Steel Struct. 16 (4) [50] J.C. Gámez, D. García, A. González, R. Pérez, An approximation to solve
(2016) 1355–1363, http://dx.doi.org/10.1007/s13296-016-0086-8. regression problems with a genetic fuzzy rule ordinal algorithm, Appl. Soft
[38] S. Gerasimidis, P. Pantidis, B. Knickle, K.S. Moon, Diagrid structural system Comput. 78 (2019) 13–28, http://dx.doi.org/10.1016/j.asoc.2019.02.012.
for high-rise buildings: Applications of a simple stiffness-based optimized [51] B. Dennis, S. Muthukrishnan, AGFS: Adaptive genetic fuzzy system for
design, Int. J. High-Rise Build. 5 (4) (2016) 319–326, http://dx.doi.org/10. medical data classification, Appl. Soft Comput. 25 (2014) 242–252, http:
21022/ijhrb.2016.5.4.319. //dx.doi.org/10.1016/j.asoc.2014.09.032.
[39] V. Tomei, M. Imbimbo, E. Mele, Optimization of structural patterns for [52] A. Nasrollahzadeh, G. Karimian, A. Mehrafsa, Implementation of neuro-
tall buildings: The case of diagrid, Eng. Struct. 171 (15) (2018) 280–297, fuzzy system with modified high performance genetic algorithm on
http://dx.doi.org/10.1016/j.engstruct.2018.05.043. embedded systems, Appl. Soft Comput. 60 (2017) 602–612, http://dx.doi.
[40] H.S. Kim, J.W. Kang, MOGA-based structural design method for diagrid org/10.1016/j.asoc.2017.07.007.
structural control system subjected to wind and earthquake loads, Int. J. [53] E.D. Santis, A. Rizzi, A. Sadeghian, Hierarchical genetic optimization of a
Steel Struct. 18 (5) (2018) 1598–1606, http://dx.doi.org/10.1007/s13296- fuzzy logic system for energy flows management in microgrids, Appl. Soft
018-0055-5. Comput. 60 (2017) 135–149, http://dx.doi.org/10.1016/j.asoc.2017.05.059.
[41] M.F. Huang, C.M. Chan, K.C.S. Kwok, Occupant comfort evaluation and [54] A.S. Koshiyama, M.M.B.R. Vellasco, R. Tanscheit, GPFIS-CLASS: A Genetic
wind-induced serviceability design optimization of tall buildings, Wind Fuzzy System based on Genetic Programming For classification problems,
Struct. 14 (6) (2011) 559–582, http://dx.doi.org/10.12989/was.2011.14.6. Appl. Soft Comput. 37 (2015) 561–571, http://dx.doi.org/10.1016/j.asoc.
559. 2015.08.055.
[42] A. Mariajayaprakash, T. Senthilvelan, R. Gnanadass, Optimization of process [55] M. Huang, Y. Ma, J. Wan, X. Chen, A sensor-software based on a ge-
parameters through fuzzy logic and genetic algorithm – A case study in a netic algorithm-based neural fuzzy system for modeling and simulating
process industry, Appl. Soft Comput. 30 (2015) 94–103, http://dx.doi.org/ a wastewater treatment process, Appl. Soft Comput. 27 (2015) 1–10,
10.1016/j.asoc.2015.01.042. http://dx.doi.org/10.1016/j.asoc.2014.10.034.
[43] N.T.P. Quyen, R.J. Kuo, Partition-and-merge based fuzzy genetic clustering [56] J.R. Srivastava, T.S.B. Sudarshan, A genetic fuzzy system based optimized
algorithm for categorical data, Appl. Soft Comput. 75 (2019) 254–264, zone based energy efficient routing protocol for mobile sensor networks
http://dx.doi.org/10.1016/j.asoc.2018.11.028. (OZEEP), Appl. Soft Comput. 37 (2015) 863–886, http://dx.doi.org/10.1016/
[44] J.M. Adánez, B.M. Al-Hadithi, A. Jiménez, Multidimensional membership j.asoc.2015.09.025.
functions in T–S fuzzy models for modelling and identification of nonlinear [57] M.H. Alavidoost, M. Tarimoradi, M.H. Fazel Zarandi, Fuzzy adaptive genetic
multivariable systems using genetic algorithms, Appl. Soft Comput. 75 algorithm for multi-objective assembly line balancing problems, Appl. Soft
(2019) 607–615, http://dx.doi.org/10.1016/j.asoc.2018.11.034. Comput. 34 (2015) 655–977, http://dx.doi.org/10.1016/j.asoc.2015.06.001.
[45] A. Ghodousian, A. Babalhavaeji, An efficient genetic algorithm for solving [58] D. Strnad, N. Guid, A fuzzy-genetic decision support system for project
nonlinear optimization problems defined with fuzzy relational equations team formation, Appl. Soft Comput. 10 (4) (2010) 1178–1187, http://dx.
and max–Lukasiewicz composition, Appl. Soft Comput. 69 (2018) 475–492, doi.org/10.1016/j.asoc.2009.08.032.
http://dx.doi.org/10.1016/j.asoc.2018.04.029. [59] R. Hassanzadeh, E. Khorram, I. Mahdavi, N. Mahdavi-Amiri, A genetic
[46] E.K. Aydogan, I. Karaoglan, P.M. Pardalos, hGA: Hybrid genetic algorithm algorithm for optimization problems with fuzzy relation constraints using
in fuzzy rule-based classification systems for high-dimensional problems, max-product composition, Appl. Soft Comput. 11 (1) (2011) 551–560,
Appl. Soft Comput. 12 (2) (2012) 800–806, http://dx.doi.org/10.1016/j.asoc. http://dx.doi.org/10.1016/j.asoc.2009.12.014.
2011.10.010. [60] B.M.M. Neta, G.H.D. Araújo, F.G. Guimarães, R.C. Mesquita, P.Y. Ekel, A
[47] Z. Yang, J. Liu, Learning of fuzzy cognitive maps using a niching-based fuzzy genetic algorithm for automatic orthogonal graph drawing, Appl. Soft
multi-modal multi-agent genetic algorithm, Appl. Soft Comput. 74 (2019) Comput. 12 (4) (2012) 1379–1389, http://dx.doi.org/10.1016/j.asoc.2011.11.
356–367, http://dx.doi.org/10.1016/j.asoc.2018.10.038. 023.
[48] C.H. Chou, S.C. Hsieh, C.J. Qiu, Hybrid genetic algorithm and fuzzy clus- [61] M. Begum, G.K.I. Mann, R.G. Gosine, Integrated fuzzy logic and genetic
tering for bankruptcy prediction, Appl. Soft Comput. 56 (2017) 298–316, algorithmic approach for simultaneous localization and mapping of mobile
http://dx.doi.org/10.1016/j.asoc.2017.03.014. robots, Appl. Soft Comput. 8 (1) (2008) 150–165, http://dx.doi.org/10.1016/
j.asoc.2006.11.010.

14