You are on page 1of 14

Journal of Constructional Steel Research 146 (2018) 155–168

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Quantification of the seismic performance factors for steel


diagrid structures
Saman Sadeghi, Fayaz R. Rofooei ⁎
Civil Engineering Department, Sharif University of Technology, Tehran, PO Box 11155-4313, Iran

a r t i c l e i n f o a b s t r a c t

Article history: There are a number of studies on seismic performance of a diagrid structure indicating that it is an effective choice
Received 10 November 2017 for constructing tall buildings. This research investigates the seismic behavior of diagrid structures and quantifies
Received in revised form 24 February 2018 its seismic performance factors including the response modification coefficient (R-factor), the over-strength fac-
Accepted 23 March 2018
tor (Ω0) and the displacement amplification factor (Cd) based on the FEMA P695 methodology. In that regard, a
group of 3-D steel diagrid archetype models with different number of floors and various diagonal angles are de-
Keywords:
Diagrid steel structures
signed using different R-factors. Then, in the first step, the over-strength factors of these models are determined
FEMA P695 by performing nonlinear static analyses. Then, utilizing the incremental dynamic analysis (IDA), the median col-
Seismic performance factors lapse capacity and collapse margin ratio (CMR) of these models are calculated and their Cd factors are estimated
Incremental dynamic analysis using the computed R-factors. The obtained results indicate that the R-factor for steel diagrid systems depend on
Nonlinear static analysis the diagonal angles. For diagrids with angles of 45°, 63.4° and 71.5°, amounts of R-factor were determined as 1.5,
2, and 3, respectively. Moreover, the pin and rigid types of end-connection of diagrid perimeter beams was found
to have no effects on stiffness of the diagrid models. However, the IDA results indicate that the diagrids with pin-
ended beams tolerate larger collapse displacements especially in shorter models. Furthermore, replacing the
rigid-ended beams by pin-ended beams improves the seismic performance of diagrids, particularly for models
with larger diagonal angles (63.4° and 71.5°). The OpenSees program was used for modeling and numerical
analyses.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction members as the lateral load carrying system is an old concept in steel
structures. But, diagonals were supposed to be highly blocking and
Recently, utilizing a steel lattice grid system named diagrid has be- were commonly used within the building interior cores by architects
come popular in modern constructions, especially for tall buildings [3]. Although the first use of diagrid is attributed to Vladimir Shukhov
and geometrically complicated structures. In this type of structure, tri- for designing Shukhov's Radio Tower (constructed in 1922), the first
angulated placement of structural members is employed to maintain case of diagrid building structure is IBM building in Pittsburgh which
the stability and the integrity of the system [1]. Diagrid structures are was built in 1963. After 1960s and more specifically, in the last few de-
classified among the tubular systems and are consist of a perimeter cades, diagrids have become a hallmark of the modern tall buildings and
space truss which resist the shear and flexural loads through the axial are being built in different configurations. Fig. 1 shows some famous
reactions of the diagonal members. Axial behavior of the diagrid mem- diagrid structures around the world.
bers makes it a stiff system that easily satisfies the displacement re- Most of the studies in recent years are concentrated on the elastic
quirements, especially in the case of tall buildings. The original form of characteristics of diagrids under wind loading, and just a few studies
diagrid structure has no vertical gravity column, since diagonal mem- have investigated their seismic behavior. Some researchers have pro-
bers resist both gravity and lateral forces owing to triangulated config- posed simple methodologies for determining preliminary member
uration [2]. So, in contrast to the most structural systems considered sizes of diagrids without conducting any finite element analyses
in the design and construction of tall buildings, the diagrid systems gen- (FEM), based on the elastic shear and flexural rigidities, diagonal angles,
erally do not need a structural core to resist lateral loads. However, and height limitations [2–4]. Considering both shear and flexural behav-
eliminating the lateral load carrying cores can result in some height lim- ior of diagrids, Moon et al. [2] concluded that the optimal angle of
itations for these structural system. Generally, application of diagonal diagrid members ranges between 65° to 70°. Also, they realized that
the inner braced core has a slight effect on overall stiffness of the diagrid
⁎ Corresponding author. structures; so they can be designed just for the gravity loads. Moon [5]
E-mail address: Rofooei@sharif.edu (F.R. Rofooei). showed that diagrids need less material than other tall building

https://doi.org/10.1016/j.jcsr.2018.03.018
0143-974X/© 2018 Elsevier Ltd. All rights reserved.
156 S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168

Fig. 1. The diagrid buildings: (a) Swiss Re Building in London, (b) Hearst Tower in New York, (c) CCTV in Beijing, (d) Cyclone Tower in Asan, (e) Lotte Super Tower in Seoul [4].

systems. Also, for diagrids taller than 60 stories, decreasing the diagonal In spite of the fact that quantifying the seismic performance factors
angles along the height would result in a more efficient structural sys- (SPFs) of structures requires a comprehensive study and recorded
tem. The shear lag phenomenon can be observed in diagrids like other data about their performance under past earthquakes, FEMA P695 pre-
tall buildings, but Leonard [6] indicates that shear lag is diminutive in sents a new methodology for analytically determining SPFs of structural
diagrids, and this braced system is less pervious to shear lag problem. systems by the help of existing high speed computers in simulating col-
Most of the constructed diagrids are standing in the low seismic re- lapse behavior of structures. According to this methodology, perfor-
gions without experiencing any intense ground motions. This might be mance of a seismic resisting system is acceptable if the system could
the primary reason of unwillingness of researchers in studying the seis- satisfy the “life safety” performance objectives of the building codes.
mic behavior of diagrids. Nevertheless, in a recent study, Baker et al. [7] Three main steps for quantifying SPFs of a system are (1) defining arche-
determined the response modification factor (R-factor) and over- type models, (2) conducting nonlinear analyses and (3) evaluating their
strength ratio (Ω0) of diagrids to be equal to 3.64 and 1.5, respectively. seismic performance. The archetype models should represent the com-
They obtained these results by conducting an iterative analytical pro- mon system configurations, construction material characteristics, en-
cess according to FEMA P695 on an eight-story diagrid with diagonal ergy dissipation mechanisms, and intended range of application. The
angles of 63.4°. Another research [8] indicates that the diagrid structures structural archetypes are distributed into “performance groups” that
exhibit a complete brittle behavior relative to the other tubular moment are characterized based on building height, fundamental period, fram-
frames and collapses in drift ratios b1% due to formation of plastic ing configurations, magnitude of gravity loads, and member and con-
hinges at lower stories. Other characteristics and performance objec- nection detailing requirements. Performance of a seismic resisting
tives such as the node behavior of diagrids (intersection of floor system is evaluated with the aid of statistical basis provided by perfor-
beams and the diagonals), application of concrete-filled steel tube mance groups.
(CFST) for diagonals, topology optimization, geometrical patterns, and
progressive collapse of diagrids have been investigated by various re-
searchers [5,9–12]. This paper focuses on quantifying the seismic per- 2.1. Nonlinear analyses
formance factors of diagrids according to the FEMA P695
methodology. In that regard, 12 diagrid structures with 8 and 12 num- To simulate the collapse performance of a system, nonlinear static
ber of stories and three different diagonal angles of 45°, 63.4° and and dynamic analyses shall be conducted using nonlinear models. At
71.5° are considered as the index archetype models which are prepared first, by pushover analysis, the over-strength (Ω0) and period based
using the OpenSees software [13]. With regard to the main horizontal ductility (μT) of the archetype models are evaluated. Then, nonlinear dy-
beam connections with the diagrid system, 6 models are considered to namic analyses are performed to characterize the collapse properties of
have pin-ended beams, and the connections in the remaining models ^CT) and collapse
the seismic resisting system. Median collapse capacity (S
are assumed to be of rigid-ended beam type. In the last part of this margin ratio (CMR) are two key parameters of each model which are
study, the seismic performance of diagrids with pin-ended beams is obtained from the results of dynamic analyses. The CMR is defined as
compared with models with rigid-ended beams. ^CT to the spectral intensity of the maximum considered
the ratio of S
earthquake (MCE) at the fundamental period of the structure (SMT).
2. Overview of the FEMA P695 methodology

Over the past several decades, the design codes as well as the effi- 2.2. Performance evaluation
cient structural systems have been greatly improved leading to an in-
creasing number of the code-approved lateral load resisting systems. Performance evaluation process for each model consists of three
So far, a considerable number of these structural systems have been main steps as: (1) adjusting CMR of the model by spectral shape fac-
listed in the internationally well-known building codes that have been tor (SSF); (2) determining acceptable collapse margin ratios based on
assigned with different seismic design coefficients. Clearly, some of total uncertainty of the model; and (3) checking performance of the
these structural systems have never been exposed to intense earth- model by comparing the adjusted collapse margin ratios (ACMR)
quakes, and their capability to fulfill the seismic design requirements with the acceptable values of collapse margin ratio. If ACMR becomes
is not well-known [14]. FEMA P695 provides a procedure for quantify- greater than acceptable CMR, seismic performance of the model is
ing the inelastic response characteristics and the performance of struc- suitable, and the employed R in designing the model is valid. If not,
tures designed according to any of these provisions. Also, this procedure the seismic performance of the model is not acceptable and whole
can be employed to evaluate the new seismic-force-resisting systems process should be repeated by using a different R or modifying the de-
proposed for inclusion in the building codes. sign requirements.
S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168 157

Fig. 2. General configurations of the archetype models: (a) plan, (b) 8-story model and (c) 12-story model.

3. Development of structural archetypes constant in this study. Additionally, all the archetype models of the
study have periods greater than the transition period of the design spec-
Although the main configuration of the diagrid structures is a combi- trum (the period Ts between the constant acceleration and constant ve-
nation of simple triangles, there is a high level of variety in the shape of locity regions of the design spectrum) and are classified as the long
the constructed diagrid buildings. Nonetheless, there are only a few period structures. In all the archetype models, the story height, the
studies available on investigating the seismic performance of diagrids, bay length, and the number of bays are kept constant as 4.5 m, 9 m
and it is expected that many yet unknown parameters will affect their and 4 m, respectively. Also, the dead and live loads are assumed to be
seismic behavior. In this research, a simple configuration of diagrids is 650 kg/m2 and 400 kg/m2, correspondingly. In order to reduce the effect
modeled by concentrating on the parameters that generally have most of core rigidity on the lateral load resistance of models, perimeter
influence on the performance of diagrids. The following parameters diagrid frames are connected to the central gravity columns by pin-
are considered in the process of defining the diagrid archetypes: (1) ended beams; therefore, the perimeter diagrid frames resist all the lat-
two 8 and 12 story diagrid structural system (the height effect); (2) di- eral loads.
agonal angles of 45°, 63.4° and 71.5°, and (3) pin and rigid diagonal- Archetype models are designed according to the ASCE/SEI 7-10 [15]
floor beam connection types. Fig. 2 shows the main plan and configura- specifications using R-factor equal to 3.6 as proposed by Baker et al. [7].
tion of archetype models, and Fig. 3 illustrates the facade of 12-story ar- For the purpose of controlling displacement requirements, Cd is as-
chetype models. Some other specifications such as seismic design sumed to be equal to the R-factor as suggested by FEMA P695. Equiva-
category (SDC), period domain, and gravity load level can affect the be- lent static method and modal response spectrum analyses are utilized
havior of a lateral seismic resisting system, but they are assumed to be to design the models. It is assumed that the models are located in a

Fig. 3. The facade of 12-story models with various angles: (a) 45°, (b) 63.4°, and (c) 71.5°.
158 S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168

Table 1

10, 11, 12
The spectral parameters for high seismic regions.

1, 2, 3
4, 5, 6
Story

7,8,9
SS S1 Fa Fv SMS SM1 SDS SD1

1.5 0.6 1 1.5 1.5 0.9 1 0.6

Tube 320 × 320 × 22.2


Tube 320 × 320 × 20
Tube 260 × 260 × 20
Tube 260 × 260 × 16
12-Story-71.5°
region with soil type D and SDC of type D. FEMA P695 recommends the
design and evaluation of structural models for all intensity levels of con-
sidered SDC, but generally the maximum level of intensity governs the
whole design process. Consequently, the archetype models are designed
and evaluated for SDC Dmax. Table 1 represents spectral parameters
used to design the models. Risk category of all models is considered to

11, 12
Story

9, 10
be I or II, based on FEMA P695 requirements. Table 3 presents the sum-

3, 4
5, 6
7, 8
1,2
mary of the performance groups of the considered models. Because the
utilized seismic provisions do not address diagrid systems, special con-

Tube 320 × 320 × 22.2


Tube 300 × 300 × 22.2
centrically braced frames (SCBF) specifications are used as the supple-

Tube 280 × 280 × 20


Tube 260 × 260 × 20
Tube 240 × 240 × 16
Tube 220 × 220 × 16
mentary requirements in design process. Tubular sections are used for

12-Story-63.4°
all of the diagonal members, horizontal beams of perimeter diagrid,
and central core columns. Central gravity load bearing beams, which
connect perimeter diagrid to central columns, have I-shaped sections.
The steel material used for all members is Grade 50 with fy = 344.7
MPa. Table 2 shows the designed sections for the diagrid models with
rigid-ended beam connections.

11, 12
Story

9, 10
1, 2
3, 4
5, 6
7, 8
4. Nonlinear analyses

Tube 320 × 320 × 28


Tube 300 × 300 × 25
Tube 300 × 300 × 20
Tube 260 × 260 × 20
Tube 220 × 220 × 20
Tube 200 × 200 × 16
4.1. Nonlinear modeling

12-Story-45°
To simulate the collapse behavior of the archetype models in dy-
namic and static analyses, the nonlinear characteristics of different
members of the designed models are assigned to them. Collapse behav-
ior of diagrid structures primarily depend on the nonlinear behavior of
their diagonal members. In this study, the model proposed by Uriz
[16] is utilized to simulate the nonlinear behavior of the diagonal mem-
1, 2, 3
4, 5, 6
Story

7, 8
bers. This distributed-plasticity nonlinear model relies on the co-rota-
tional theory to represent the global buckling and large deformations
of the diagonals. In the modeling process, the diagonal members are di-
Tube 280 × 280 × 20
Tube 260 × 260 × 20
Tube 260 × 260 × 16

vided to two equal elements that are connected to each other by apply-
ing a small imperfection at the connection point. The nonlinear force-
8-Story-71.5°

based beam-column element used herein benefits fiber technique and


accounts for the interaction of the axial force and bending moment
along the diagonal member. This nonlinear beam-column element has
the following two main shortcomings: (1) the plane sections remain
The considered sections for diagrid structures with rigid ended beam connections.

plane, and (2) the shear deformation is ignored [16].


Story

In order to accurately capture the nonlinear response of diagonal el-


1, 2
3, 4
5, 6
7, 8

ements, an initial imperfection of 0.005L (L is total length of the mem-


ber) is assigned to the middle of the member. The hysteretic steel
Tube 260 × 260 × 22.2
Tube 260 × 260 × 17.5

material Menegotto-Pinto with isotropic hardening is assigned to each


Tube 280 × 280 × 20

Tube 220 × 220 × 16

fiber of the element. Main parameters of this material are defined as:
the yield stress (Fy = 344.7 Mpa), the Young's modulus (E = 199.9
8-Story-63.4°

Gpa), strain hardening ratio (b = 0.003), three parameters (R0 = 18.5,


cR1 = 0.925, cR2 = 0.15) to control the transition from elastic to plastic
branches, four parameters (a1 = 0.0005, a2 = 0.01, a3 = 0.0005, a4 =
0.01) to account for cyclic isotropic hardening. Each box section of the
diagonal members is divided into 3 fibers along the thickness and 6 fi-
Story

bers along the edges. Gauss-Lobatto numerical integration with 5 inte-


1, 2
3, 4
5, 6
7, 8

gration points along the element are considered. Using this nonlinear
fiber-based element, it is possible to simulate the buckling and overall
Tube 300 × 300 × 22.2
Tube 260 × 260 × 22.2
Tube 220 × 220 × 22.2

post buckling behavior of an axially loaded member.


Tube 200 × 200 × 16

To consider the failure of the member under cyclic and monotonic


loading, the low cycle fatigue (LCF) behavior proposed by Uriz [16]
8-Story-45°

were assigned to the material of the element. These properties have


been suggested by Uriz [16] and rely on numerous tests performed by
Table 2

him, and Young and Mahin [17]. Although, the LCF material is not capa-
ble of taking into account the local buckling of a steel brace, but the
S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168 159

Table 3
The performance groups of diagrid models.
F x ∝mx ∅1;x ð2Þ
Group Grouping criteria Number of
no. Archetypes
Basic Design load level Period Beam where x indicates the floor level, F is the horizontal force acting on dif-
configuration domain connection
Gravity Seismic ferent floors, and m is the story mass. Fig. 5 shows the FEMA P695 sug-
type
gested method to determine the over-strength (Ω) and the ductility (μT)
PG-1 θ = 45° Low SDC Long Pin 2
coefficients as defined by Eq. (3) and Eq. (4):
Dmax Rigid 2
°
PG-2 θ = 63.4 Low SDC Long Pin 2
Dmax Rigid 2
Ω ¼ V max =V ð3Þ
PG-3 θ = 71.5° Low SDC Long Pin 2
Dmax Rigid 2 μ T ¼ δu =δy;eff ð4Þ

The effective yield displacement is defined as:


parameters of the material have been calibrated in a way to consider the
local buckling indirectly. Fig. 4 depicts the nonlinear beam-column ele- V max h g i
δy;eff ¼ C 0 ð maxðT; T 1 ÞÞ2 ð5Þ
ment, material behavior of the fibers and the overall force-deformation W 4π 2
behavior of the member under the cyclic loading. In the last cycle, hys-
teresis loop of the element begins to deteriorate and finally failure oc- in which W is the total seismic weight of the structure, T1 is fundamen-
curs due to fatigue phenomenon. tal mode obtained from the modal analysis, and T is the fundamental
In order to eliminate the lateral load resistance capacity of gravity I- mode of the structure determined according to the section 12-8-2-1 of
shaped beams that connect joints of the perimeter diagrid to the central ASCE/SEI 7-05 [15]. In case of diagrid systems, it can be calculated as a
columns, these members were modeled in OpenSees program as pin- function of total height (H) of the structures using the empirical relation
ended elements with the command “element truss”. Floor system was defined by Eq. (6):
not modeled, but to simulate the effects of rigid floor, all nodes of each
T ¼ 1:4  0:0488H0:75 ð6Þ
level were constrained by the command “rigidDiaphragm”. As already
mentioned, to investigate the effect of horizontal beam connections' ri-
gidity of the perimeter diagrid on the performance of the whole system, The parameter C0 that relates the roof displacement to the funda-
the results of the cases with pin-ended and rigid-ended beams are com- mental mode of the structures can be approximated as:
pared. The pin-ended beams were modeled with the simple truss ele- PN
mx ∅1;x
ments, while the rigid-ended beams were modeled using nonlinear C 0 ¼ ∅1;r P1N 2
ð7Þ
Beam-Column elements with the same nonlinear characteristics as the 1 mx ∅1;x
diagonal elements. Central columns were modeled as elastic members
with the command “element elastic BeamColumn”, in order to prevent where ∅1, x (∅1, r) is the ordinate of the fundamental mode at the story x
them from having plastic behavior due to large displacements. The P- (where r represents the roof level), and N is total number of stories. Fig.
Delta coordinate transformation command was used to include the P 6 shows the actual and idealized pushover curves for models with pin-
− Δ second-order effects in modeling the central gravity columns. ended beams. Also, Table 4 summarizes the results of the pushover anal-
yses for all archetype models.
4.2. Nonlinear static analysis (pushover analysis) According to FEMA P695, if the maximum roof displacement in a
pushover curve becomes less than the ultimate displacement, the col-
Following the FEMA P695 guidelines, Ω0 and μT of the archetype lapse drift (defined in Section 4.3) is considered as the ultimate dis-
models are evaluated using pushover analysis. For that purpose, first placement. As it is evident from Fig. 7, the over-strength factors for
the gravity load combination given by Eq. (1), in which D and L repre- the models with rigid-ended beams are generally greater than that of
sent the median values of dead and live loads respectively, is applied the models with pin-ended beams. Moreover, any increase in the diag-
to the models. Then, these models are pushed laterally in a regular onal angles will reduce the over-strength factor. For 12-story models,
push over analysis using a lateral load pattern defined by Eq. (2): the computed over-strength factors for cases with pin-ended and
rigid-ended beams are close. Also, for diagonal angles of 45°, the end-
connection type of beams almost has no effect on the determined
1:05D þ 0:25L ð1Þ over-strength factors in all models.

Table 4
The results of pushover analysis.

Index archetype Seismic weight (MN) Vmax (MN) Vd (MN) δy, eff (m) δu (m) Ω μT

Performance Group no. 1 (θ = 45°)


8-Story pin_ended beams 75.1 51.87 12.4 0.11 0.22 4.18 2.0
8-Story rigid_ended beams 75.1 52.35 12.4 0.11 0.21 4.22 1.91
12-Story pin_ended beams 112.63 49.03 13.7 0.2 0.36 3.57 1.8
12-Story rigid_ended beams 112.63 48.63 13.7 0.2 0.38 3.55 1.9
Performance Group no. 2 (θ=63.4°)
8-Story pin_ended beams 75.38 21.32 12.51 0.06 0.1 1.7 1.67
8-Story rigid_ended beams 75.38 37.36 12.51 0.13 0.47 2.98 3.61
12-Story pin_ended beams 113.1 37.45 13.85 0.18 0.68 2.7 3.78
12-Story rigid_ended beams 113.1 39.51 13.85 0.20 0.81 2.85 4.05
Performance group no. 3 (θ = 71.5°)
8-Story pin_ended beams 75.48 19.19 12.53 0.17 0.9 1.53 5.29
8-Story rigid_ended beams 75.48 28.43 12.53 0.18 0.72 2.27 4.0
12-Story pin_ended beams 113.16 30.97 13.86 0.16 0.4 2.23 2.5
12-Story rigid_ended beams 113.16 34.33 13.86 0.18 0.53 2.47 2.94
160 S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168

Table 5
The results of the first performance evaluation process using R = 3.6.

Index archetype Framing (gravity loads) SDC Static SMT ^CT


S CMR μT SSF ACMR Accept Pass/fail
Ω [g] [g] ACMR

Performance group no. 1 (θ = 45°), R = 3.6


8-Story pin_ended beams P Dmax 4.18 0.89 0.86 0.96 2.0 1.18 1.36 1.54 Fail
8-Story rigid_ended beams P Dmax 4.22 0.89 0.75 0.84 1.91 1.17 1.18 1.52 Fail
12-Story pin_ended beams P Dmax 3.57 0.66 0.5 0.76 1.8 1.21 1.1 1.52 Fail
12-Story rigid_ended beams Dmax 3.55 0.66 0.56 0.86 1.9 1.21 1.25 1.52 Fail
Mean of performance Group 3.88 0.85 1.9 1.18 1.21 1.9 Fail
Performance group no. 2 (θ=63.4°), R = 3.6
8-Story pin_ended beams P Dmax 1.7 0.89 0.72 0.80 1.67 1.15 1.1 1.51 Fail
8-Story rigid_ended beams P Dmax 2.98 0.89 0.67 0.75 3.61 1.29 1.16 1.62 Fail
12-Story pin_ended beams P Dmax 2.7 0.66 0.71 1.08 3.78 1.35 1.75 1.62 Pass
12-Story rigid_ended beams P Dmax 2.85 0.66 0.66 1.0 4.05 1.37 1.64 1.62 Pass
Mean of performance group 2.56 0.91 3.28 1.29 1.41 2.09 Fail
Performance group no. 3 (θ = 71.5°), R = 3.6
8-Story pin_ended beams P Dmax 1.53 0.89 0.98 1.09 5.29 1.36 1.78 1.62 Pass
8-Story rigid_ended beams P Dmax 2.27 0.89 0.86 0.96 4.0 1.31 1.51 1.62 Fail
12-Story pin_ended beams P Dmax 2.23 0.66 0.67 1.02 2.5 1.25 1.53 1.61 Fail
12-Story rigid_ended beams P Dmax 2.47 0.66 0.66 1.0 2.94 1.29 1.55 1.62 Fail
Mean of performance Group 2.12 1.02 3.68 1.30 1.59 2.09 Fail

Fig. 8 illustrates the effect of changing the ductility factor versus di- that represents the damage measure (DM) is considered to be the max-
agonal angles. It indicates that by increasing diagonal angles, ductility imum story drift, like the common IDA curves. However, the vertical
ratio increases, except for angles larger than 63.4° in 12-story models. axis of the IDA curves representing the intensity measure (IM) is as-
sumed to be the “median spectral intensity (ST)” of the scaled record
4.3. Nonlinear dynamic analyses set, rather than spectral intensity of each individual record. This ap-
proach can be simply defined by the following algorithm:
To estimate the “median collapse capacity” and to evaluate the seis-
1- Normalize each pair of records with respect to its PGVpeer.
mic performance factors of the models, nonlinear time history analyses
are conducted using 22 pairs of the far-field earthquake records sug- 2- Obtain the 5%-damped acceleration response spectrum of each indi-
gested by FEMA P695 and registered on soil type D with their spectral vidual normalized record (44 curves) and determine their median
acceleration shown in Fig. 9. response spectrum curve (Fig. 10).
For incremental dynamic analysis, the prepared 3-D diagrid struc- 3- Read the median spectral acceleration at the fundamental period of
tural models are subjected to bi-directional horizontal excitations with structure (ST1) from the median response spectrum curve.
varying intensities. Each pair of records are normalized with relative 4- Choose some proper scale factors (SFs) for scaling the normalized re-
to their geometric mean of peak ground velocities (PGVpeer) in order cords and perform the time history analyses subject to the scaled re-
to remove the unwarranted variability between the records due to in- cords and extract the maximum drift ratios from the results of each
herent differences in event magnitude, distance to source, source type analysis as its damage measure.
and site conditions, without eliminating record-to-record variability 5- Calculate the median spectral acceleration of the scaled records at
[14]. Then, the normalized pair of records is scaled upward to some spe- the fundamental period (ST) by multiplying the SFs by ST1and locate
cific intensity and the time history analyses are performed using these the obtained values on the vertical axis of the IDA curves as intensity
scaled records. This method has somewhat different concept relative measures.
to the incremental dynamic analysis (IDA) proposed by Vamvatsikos 6- Construct the IDA curves using DMs and IMs obtained from the steps
and Cornell [18]. In this approach the horizontal axis of the IDA curves above.

Table 6
The Seismic performance evaluation of different archetype models using various R-factors.

Index archetype Framing (gravity loads) SDC Static SMT ^CT


S CMR μT SSF ACMR Accept Pass/fail
Ω [g] [g] ACMR

Performance group no. 1 (θ=45°), R = 1.5


8-Story pin_ended beams P Dmax 3.04 0.89 1.22 1.37 1.85 1.17 1.92 1.52 Pass
8-Story rigid_ended beams P Dmax 3.1 0.89 1.11 1.25 1.83 1.16 1.74 1.52 Pass
12-Story pin_ended beams P Dmax 2.81 0.66 0.89 1.35 2.8 1.29 2.09 1.6 Pass
12-Story rigid_ended beams P Dmax 2.8 0.66 0.92 1.39 3.0 1.3 2.17 1.6 Pass
Mean of performance group 2.94 1.24 2.37 1.23 1.98 1.97 Pass
Performance group no. 2 (θ=63.4°), R = 2.5
8-Story pin_ended beams P Dmax 2.0 0.89 1.33 1.49 1.85 1.17 2.1 1.52 Pass
8-Story rigid_ended beams P Dmax 2.7 0.89 1.26 1.41 3.0 1.25 2.12 1.6 Pass
12-Story pin_ended beams P Dmax 3.0 0.66 0.94 1.43 2.6 1.27 2.18 1.59 Pass
12-Story rigid_ended beams P Dmax 3.0 0.66 0.87 1.32 2.71 1.28 2.03 1.59 Pass
Mean of performance Group 2.67 1.41 2.54 1.24 2.11 2.0 Pass
Performance group no. 3 (θ=71.5°), R=3
8-Story pin_ended beams P Dmax 2.32 0.89 1.22 1.37 5.55 1.37 2.25 1.62 Pass
8-Story rigid_ended beams P Dmax 2.68 0.89 1.07 1.2 4.5 1.33 1.92 1.62 Pass
12-Story pin_ended beams P Dmax 2.4 0.66 0.89 1.35 3.5 1.33 2.15 1.62 Pass
12-Story rigid_ended beams P Dmax 2.68 0.66 0.84 1.27 4.1 1.38 2.1 1.62 Pass
Mean of performance group 2.52 1.22 4.16 1.35 2.1 2.09 Pass
S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168 161

Fig. 4. The nonlinear beam-column element: (a) initial imperfection, (b) behavior of the material of the fibers, (c) cyclic loading history, and (d) the force-displacement behavior of the
member.

The median collapse capacity (S ^CT ) is a specific value of ST for which becomes large enough to offset the shear resistance of the structure,
the application of one-half of the scaled records would lead to the col- which makes it to lose its gravity load resistance [20]. Numerical insta-
lapse of the structural models. Therefore, in performing step 4 of the bility due to non-convergence of the time-integration process features
above algorithm, if 22 analyses under the scaled record pairs cause the sidesway collapse of nonlinear diagrid models in the OpenSees. Also,
collapse of the models, the evaluation process can be stopped. This is a the collapse point is observable in IDA curves, when a minor increase
different approach to estimate the median collapse capacity in compar- in IM causes large increment in DM.
ison to the common statistical methods where median collapse capacity In performance evaluation process, the collapse point is considered
is the median of 44 collapse points; Hsiao et al. [19] indicate that gener- as occurrence of simulated or non-simulated collapse criteria. Simulated
ally, the statistical method will result in greater values of S^CT . collapse modes were defined in Section 4.1, but the non-simulated col-
According to FEMA P695, collapse is defined as the instability of lapse criteria of diagrid systems are not well known due to lack of re-
whole seismic resisting system and not the local failure of elements search on seismic behavior of diagrids. Some experiments [21]
and components. In case of braced frames subjected to horizontal indicate that certain non-simulated collapse modes for diagrids, such
ground motions, sidesway collapse is a governing mechanism that as tensile fracture of members, are concentrated in the connection
leads to lateral dynamic instability and collapse of the system. The zone and are similar to the net section fracture of braces in braced
sidesway collapse occurs when the displacements of any specific story frames. Therefore, similar to the braced frames, a drift ratio of 5% can
be considered as a limit state for checking this event in the diagrid struc-
tures [22]. Moreover, the results of experiments performed by Lehman
and Roeder [23] on HSS tube braces with angles of 45° proves that at
drift ratio of 5%, failure occurs due to brace fracture, weld fracture, or
plate edge buckling. Therefore, using a drift ratio of 5% seems to be an
appropriate limit state to check the occurrence of non-simulated col-
lapse modes of diagrids with angles of 45°. Drift limit states rise as the
diagonal angles of braces increase. Thus the 5% limit state for 45° braces
grows when the diagonal angles become steeper. For steel moment
frames 10% drift ratio is considered to control the collapse prevention
[18]. So, for 63.4° and 71.5° diagonal angles, the drift ratio limits range
between 5 and 10%. In this study, it is assumed that 7 and 9% drift ratios
can be used to check the non-simulated collapse modes for angles of
63.4° and 71.5° respectively, but it should be noticed that the credibility
of this assumption needs to be investigated experimentally.
As already mentioned, to evaluate the seismic performance of
models, S ^CT is compared with the maximum considered earthquake
spectrum, which in this study is considered as the spectrum of the
SDCDmax. The SMT for 8 and 12-story models are obtained to be 0.66 g
and 0.9 g, respectively. These are computed using the response spec-
trum curve presented in Fig. 10 and based on the fundamental periods
Fig. 5. An idealized pushover curve. of the models, which are determined to be 1 (s) for 8-story and 1.36
162 S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168

Actual Curve Actual Curve


Idealized Curve Idealized Curve

(a) (b)

Actual Curve
Idealized Curve

Actual Curve
Idealized Curve

(c) (d)

Actual Curve
Idealized Curve

Actual Curve
Idealized Curve

(e) (f)

Fig. 6. Pushover curves of diagrid models with pin-end beams: (a) 8-story-45°, (b) 12-story-45°, (c) 8-story-63.4°, (d) 12-story-63.4°, (e) 8-story-71.5°, (f) 12-story-71.5°.

(s) for 12-story models employing Eq. (6). Fig. 11 presents the IDA assessment of this model. The collapse margin ratio (CMR) for the 12-
curves for the 12-story, 63.4° model. Differences between the curves story-63.4° model is determined as:
are just due to the record-to-record variability. As it can be seen from
this figure, some records do not cause structural collapse, and many of
the records cause collapse near drift ratio of 10% due to the simulated ^SCT 0:75
CMR ¼ ¼ ¼ 1:05 ð8Þ
collapse modes. So, non-simulated collapse modes govern performance SMT 0:66
S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168 163

Fig. 7. The over-strength factor as a function of diagonal angles: (a) 8-story models, (b) 12-story models.

The spectral shape of the design spectrum of ASCE/SEI 7-10 [15] is and for systems with low ductility such as diagrid buildings, values of
smooth and constant over the short period region. On the contrary, βRTR are determined as follows [25]:
spectral curve of the ground motions generally have erratic behavior
and rapidly varies by variation of period. This means that, if earthquake βRTR ¼ ð0:1 þ 0:1μ T Þb0:4 ð10Þ
intensities are defined based on spectral accelerations at the fundamen-
tal period of a structure with adequate ductility and period elongation The DR uncertainty (βDR) represents the robustness and complete-
due to inelastic deformations, the ground motions cause less damage, ness of the design requirements and how this regulations provide safety
and structure receives less input energy than expected based on the de- margin against non-simulated collapse modes. In general, the design
sign spectrum [24]. To take into account this phenomenon in the seis- codes do not yet address the diagrid systems, but the members of the
mic performance evaluation of structures, CMR values of models are diagrid structures are designed according to robust relations, so the
increased by the spectral shape factor (SSF), as defined in Eq. 9, and ad- DR uncertainty of diagrids is considered to be fair (βDR = 0.35). The
justed collapse margin ratio (ACMR) will be used in the performance TD uncertainty (βTD) is related to the comprehensiveness and robust-
evaluation process instead of CMR. ness of the test data used to define the behavior of the components of
a system and depends on the quality of test results [14]. Although
there is no complete testing program on the seismic behavior of diagrids
ACMRi ¼ SSFi  CMRi ð9Þ in particular, a large number of experiments have been conducted on
the performance of axially-loaded elements, which are the primary
members in a diagrid system. Therefore, TD uncertainty of diagrids is
where index i represents the ith archetype model. SSFi depends on the
supposed to be superior (βTD = 0.1). The MDL uncertainty indicates
fundamental period of the structure, period based ductility (μT), and
how well the archetype models represent all the response characteris-
seismic design category (SDC). Additionally, according to FEMA P695
tics and design parameters and how well collapse performance of the
requirements, the obtained ACMR values are increased by 1.2, in order
models are evaluated from simulated and non-simulated collapse be-
to consider effects of three dimensional modeling.
havior of components. In this study, parameters of nonlinear models
In the FEMA P695 methodology, there are four sources of uncer-
have been well calibrated on the basis of numerous experimental re-
tainty that affect performance evaluation process: (1) record-to-record
sults, so the MDL uncertainty is considered to be good (βMDL = 0.2).
uncertainty (RTR); (2) design requirements uncertainty (DR); (3) test
Total uncertainty (βTOT) is a combination of the RTR, DR, TD, and MDL
data uncertainty (TD); and (4) nonlinear modeling uncertainty (MDL).
uncertainties. In seismic performance evaluation process, acceptable
Acceptable values of CMR are defined based on the total uncertainty de-
values of CMR are a function of βTOT. Also, the fragility curve of a diagrid
rived from these sources of uncertainty. The RTR variability which is an
model is constructed using the βTOT and S ^CT through the following rela-
outcome of frequency content variability and dynamic and hazard prop-
erties of the far-field record set, is clearly evident from the IDA curves. tion:
Values of the RTR uncertainty (βRTR) ranges over 0.35 to 0.45 for various
structural systems. For systems that have little or no period elongation, SCT ¼ ^SCT  λTOT ð11Þ

Fig. 8. The ductility ratios as functions of diagonal angles: (a) 8-story models, (b) 12-story models.
164 S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168

1.2

0.8
- - - Median+1stdv
Median

ST [g]
0.6

0.4

0.2

0
0 0.05 0.1 0.15 0.2 0.25
Maximum Story Drift Ratio

Fig. 11. The IDA curves for the 12-story model with the diagonal angle of 63.4°.

5. Performance evaluation

Fig. 9. The response spectrum of the normalized far-field records. 5.1. Computation of response modification factor, R-factor

Acceptable seismic performance of each performance group is


where SCT is a random variable with the median of S ^CT and the standard achieved when two following criteria are satisfied simultaneously:
deviation of βTOT. The λTOT is a lognormal random variable with the me-
• Mean ACMR of the performance group (ACMR) exceeds acceptable
dian of unity and standard deviation of βTOT. λTOT is assumed to be equal
collapse margin ratio corresponding to 10% collapse probability
to the product of four random variables as the following:
(ACMR10%):

λTOT ¼ λRTR λDR λTD λMDL ð12Þ


ACMR≥ACMR10% ð14Þ
where λRTR, λDR, λTD, and λMDL are lognormally distributed random var-
iables with the median of unity and standard deviation of βRTR, βDR, βTD
and βMDL, respectively. These four random variables are assumed to be • ACMR of each index archetype model within the performance group
statistically independent, so the total collapse uncertainty is given by: exceeds acceptable collapse margin ratio corresponding to 20% col-
lapse probability (ACMR20%):
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
βTOT ¼ βRTR 2 þ βDR 2 þ βTD 2 þ βMDL 2 ð13Þ
ACMRi ≥ACMR20% ð15Þ
Acceptable values of CMR are based on the total collapse uncertainty
shows summarized results of seismic performance evaluation process of
and acceptable collapse probabilities under the MCE ground motions.
diagrid system. Last column of the table indicates that majority of the ar-
The FEMA P695 methodology limits the acceptable collapse probability
chetype models have smaller ACMR than acceptable values of CMR. As a
to 10% for each performance group and 20% for each index archetype
result, acceptable performance was not achieved by diagrid models lo-
within a performance group. Acceptable values of ACMR corresponding
cated in high seismic regions and designed using R-factor equal to 3.6.
to the βTOT and 10% and 20% collapse probabilities extracted from FEMA
To satisfy the seismic performance objectives of diagrids, designing ar-
P695 are shown in Table 5.
chetypes and performance evaluation process are reiterated using
smaller R-factors. Although this is a time consuming process, evaluating
the acceptable R-factor generally does not require conducting a full IDA,
because S ^CT can be determined where just 22 of the 44 analyses cause
collapse. Table 6 represents results of iteration process and acceptable
R-factors for each performance group. As it is evident from the table,
diagrids with angles of 45° can be designed using R = 1.5, and diagrids
with diagonal angles of 63.4° and 71.5° should be designed considering
R = 2.5 and R = 3.0, respectively.

5.2. Evaluating displacement amplification factor, Cd

The parameter Cd can be approximated based on the acceptable R-


factor as:

R
Cd ¼ ð16Þ
BI

where BI, given by Table 18.6-1 of ASCE/SEI 7-10, is a function of effec-


tive damping (βI) and fundamental period (T). For systems with effec-
Fig. 10. Response spectrum curves for the MCE ground motions. tive damping ratio b2%, BI is equal to 0.8. Here, similar to other studies
S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168 165

Pin-end beams Pin-end beams


Rigid-end beams Rigid-end beams

Pin-end beams
Rigid-end beams

(a) (b) (c)

Pin-end beams Pin-end beams


Rigid-end beams Rigid-end beams

Pin-end beams
Rigid-end beams

(d) (e) (f)

Fig. 12. Pushover curves of pin and rigid-end beam models: (a) 8-story-45°, (b) 12-story-45°, (c) 8-story-63.4°, (d) 12-story-63.4°, (e) 8-story-71.5° and (f) 12-story-71.5°.

[8], 2% damping ratio was assigned to the diagrid models in dynamic with rigid beam connections is larger than the stiffness of the model
analyses, so Cd factors for diagrids with angles of 45°, 63.4°, and 71.5° with pin-ended beam connections. This disagreement is due to the dif-
are equal to 1.87, 3.12, and 3.75, respectively. ference between the first mode amplitudes of these two models; in fact,
the diagrid modules in these two cases are not complete in the top two
5.3. Evaluating the overstrength factor, Ω0 stories. So, they comprise one perfect 6-story module and one imperfect
two-story module, which forces the 8-story-71.5° case with rigid con-
To assess the Ω0, first the average value of archetype overstrength, Ω, nections to act like a moment resisting frame in the upper two stories.
is determined for each performance group. Then, system overstrength, But in the model with pin-ended beams, diagonal members of 7th and
Ω0, is defined as the largest value of Ω of all performance groups. With 8th floors are connected to each other by beams that just transfer
regard to Table 6, values of Ω for diagrids with diagonal angles of 45°, axial loads. This difference greatly affects the amplitudes of the first
63.4°, and 71.5° are 2.94, 2.67 and 2.52, respectively. Thus, the mode of the models. Thus, the first mode amplitudes of the top two
overstrength of diagrid system, Ω0, is determined as 3.0, which is a prac- stories in the model with pin-ended beams will be larger than that of
tical limit of Ω0 for design of lateral load resisting systems recom- model with rigid connections.
mended by ASCE 7-05. For diagrids with diagonal angles of 45°, replacing rigid-ended
An important result of quantification process for diagrid systems is beams by pin-ended beams does not change the elastic stiffness or the
that the seismic performance factors of diagrids are highly dependent post yielding behavior of archetype models. It can be seen that the ef-
on the diagonal angles. While the R and the Cd factors have an increasing fects of changing beam connection type increase as the angles of diago-
trend with increased diagonal angles, Ω shows a decreasing pattern nals increase. In fact, when diagonal angles come closer to 90°, diagrid
with respect to the increasing diagonal angles. structural system tends to behave more like a moment resisting frame.
Fig. 13 represents statistical distributions of collapse drifts and inten-
6. Comparison seismic performance of the diagrids with rigid-ended sities obtained from the IDA analyses. For each model, a lognormal cu-
beams to diagrids with pin-ended beams mulative distribution function has also been constructed based on the
mean and standard deviation of IDA collapse data. Collapse intensities
As mentioned before, two types of beam-diagonal member connec- have been defined as the geometric means of the 5%-damped spectral
tions are used in the construction of different diagrid structures. So, con- accelerations of each pair of records at the fundamental period of the
sidering a number of models, the effect of connection type (pin-ended structure (Sa(T,5%)):
or rigid-ended beam-diagonal connection) on the seismic performance pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
of the diagrid structures is investigated using the results of the pushover Sa ðT; 5%Þ ¼ Sa1 ðT; 5%Þ  Sa2 ðT; 5%Þ ð17Þ
and IDA analyses conducted in previous sections. Fig. 12. presents the
pushover curves of diagrid models with both pin-ended and rigid- where Sa1(T, 5%) and Sa2(T, 5%) are the 5%-damped spectral accelera-
ended beam connections. As clearly shown in this figure, beam connec- tions of the considered pair of records at the fundamental period of
tion type generally affects the post yielding behavior of the models, and the structure. This definition has the advantage of considering the char-
the elastic stiffness of the structures does not change by changing the acteristics of the pair records in describing the collapse intensities,
beam connection type. The only exception is the pushover curve of rather than specifying the intensities by properties of an individual
the 8-story-71.5° model, in which the elastic stiffness of the model record.
166 S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168

1.2 1.2

Cumulative Frequency
Cumulative Frequency
1 1

0.8 0.8

0.6 0.6

0.4 0.4 Data


Data
0.2 Lognormal_Fitted Curve 0.2 Lognormal_Fitted Curve

0 0
0 0.02 0.04 0.06 0.08 0.1 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
(a)
Simulated Collapse Drift Ratio Sa (T , 5%) at Collapse
1.2 1.2
Cumulative Frequency

Cumulative Frequency
1 1

0.8 0.8

0.6 0.6

0.4 Data 0.4


Data
Lognormal_Fitted Curve
0.2 0.2 Lognormal_Fitted Curve
0 0
0 0.02 0.04 0.06 0.08 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Simulated Collapse Drift Ratio (b) Sa (T , 5%) at Collapse

1.2 1.2
Cumulative Frequency

Cumulative Frequency
1 1

0.8 0.8

0.6 0.6

0.4 0.4
Data Data
0.2 Lognormal_Fitted Curve 0.2 Lognormal_Fitted Curve

0 0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.2 0.4 0.6 0.8 1 1.2 1.4
Simulated Collapse Drift Ratio (c) Sa (T , 5%) at Collapse

1.2 1.2
Cumulative Frequency
Cumulative Frequency

1 1

0.8 0.8

0.6 0.6

0.4 0.4
Data Data
0.2 Lognormal_fitted Curve 0.2 Lognormal_Fitted Curve
0 0
0 0.05 0.1 0.15 0.2 0.25 0.2 0.4 0.6 0.8 1 1.2 1.4
Simulated Collapse Drift Ratio (d) Sa (T , 5%) at Collapse

1.2 1.2
Cumulative Frequency

Cumulative Frequency

1 1

0.8 0.8

0.6 0.6

0.4 0.4
Data Data
0.2 Lognormal_Fitted Curve 0.2
Lognormal_Fitted Curve
0 0
0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Simulated Collapse Drift Ratio Sa (T , 5%) at Collapse
(e)

1.2 1.2
Cumulative Frequency
Cumulative Frequency

1 1

0.8 0.8

0.6 0.6

0.4 Data 0.4


Data
0.2 Lognormal_Fitted Curve 0.2 Lognormal_Fitted Curve

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.3 0.5 0.7 0.9 1.1 1.3
Simulated Collapse Drift Ratio (f) Sa (T , 5%) at Collapse

Fig. 13. Cumulative distributions of collapse drifts and spectral intensities of diagrid models with rigid-end beams: (a) 8-story-45°, (b) 12-story-45°, (c) 8-story-63.4°, (d) 12-story-45°, (e)
8-story-71.5°, (f) 12-story-71.5°.
S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168 167

2 2.5
PI for Collapse Drifts of 12 Story Models PI for Collapse Sa of 12 Story Models
PI for Collapse Drifts

PI for Collapse Drifts of 8 Story Models 2 PI for Collapse Sa of 8 Story Models

PI for Collapse Sa
1.5
1.5
1
1
0.5
0.5

0 0
45 63.4 71.5 45 63.4 71.5
Angle of Diagonals (degrees) Angle of Diagonals (degrees)
(a) ( b)

Fig. 14. Performance index of diagrid models: (a) PI for collapse drifts and (b) PI for collapse spectral intensities.

To better assess the effects of beam connection type on the behavior collapse increases by increasing angle. Thus, the diagrids with angles
of the diagrid systems, a performance index (PI) is defined as the ratio of of 71.5° perform better than 63.4° and 45° models in seismic regions.
the median collapse drift ratio (DR) or spectral acceleration, Sa(T, 5%), of ○ Seismic performance of diagrid systems highly depends upon the di-
the diagrid with pin-ended beams to the median collapse drift ratio or agonal angles. Suggested R-factors for diagrids with angles of 71.5°,
spectral acceleration of the diagrid system with rigid-ended beams: 63.4°, and 45° in regions of high seismicity are determined to be 3,
2.5 and 1.5, respectively.
MedianðDR or Sa of pin−ended beam modelÞ ○ Due to the axial performance of diagrids, end-connection type of pe-
PI ¼ ð18Þ rimeter beams has no effect on stiffness of the diagrid models, but
MedianðDR or Sa of rigid−ended beam modelÞ
post yield behavior of models can be affected by end-connection
type of beams.
Fig. 14 represents the PI values corresponding to the considered IM
○ IDA results indicate that the diagrids with pin-ended beams tolerate
and DM parameters. The PIs for nearly all models of Fig. 14 (a) are
larger displacements before collapse in comparison to the rigid-
greater than one, meaning that the diagrids with pin-ended beams, in
ended beam models. This is more visible in shorter models (8-
average, experience greater displacements before collapse than the
story models) than taller ones (12-story models).
diagrids with rigid-ended beams do. This observation is predominant
○ Generally, diagrid models with pin-ended beams can also resist
especially for 8-story models with diagonal angles N45°. According to
more intense ground motions with respect to the models with
Fig. 14 (b), PI for 8-story models is greater than one and increases
rigid-ended beams.
with rising diagonal angles, so 8-story diagrids with pin-ended beams
○ Replacing rigid-ended beams by pin-ended beams will improve the
can resist more intense ground motions than 8-story diagrids with
seismic performance of diagrid models, especially for diagrid angles
rigid-ended beams. The same is true for all 12-story models except the
of 63.4° and 71.5°.
12-story-45°, which has a PI of collapse intensities less than one. Com-
pared with 8-story models, PI values of collapse drifts or collapse IM
for 12-story models do not considerably change by variation of diagonal
References
angles. These results indicate that for diagrids with diagonal angles of
63.4° and 71.5°, replacing the rigid-ended beams by the pin-ended [1] T.M. Boake, Diagrid Structures, Systems, Connections, Details, Basel: Worldsteel As-
beams will improve their seismic performance in terms of increasing sociation, 2014.
[2] K.S. Moon, J.J. Connor, J.E. Fernandez, Diagrid structural systems for tall buildings:
their collapse displacements. Moreover, they can resist higher intensi-
characteristics and methodology for preliminary design, Struct. Des. Tall Spec.
ties of earthquake ground motions. Build. 16 (2) (2007) 205–230.
[3] G.M. Montuori, E. Mele, G. Brandonisio, A. De Luca, Geometrical patterns for diagrid
buildings: exploring alternative design strategies from the structural point of view,
7. Conclusion
Eng. Struct. 71 (2014) 112–127.
[4] R. Brun, “Shock tubes and shock tunnels: design and experiments,” Chem. React.
The diagrid structures are becoming a sign of modern construction Shock Waves Detonations Handb, Shock Waves 1 (2001) 1–26.
of tall and geometrically complicated buildings. Structural efficiency, in- [5] K.S. Moon, Sustainable structural engineering strategies for tall, Buildings 17 (5)
(2008).
herent stiffness, and aesthetic quality are primary reasons of current [6] J. Leonard, J.J. Connor, Investigation of Shear Lag Effect in High-rise Buildings with
tendency to use this lattice grid system. However, there are only a lim- Diagrid System, Massachusetts Institute of Technology, 2007.
ited number of studies available about the seismic performance of the [7] W. Baker, C. Besjak, M. Sakisian, P. Lee, C.S. Doo, Proposed Methodology to Deter-
mine Seismic Performance Factors for Steel Diagrid Framed Systems, Counc. Tall
diagrids. The main findings of this study can be summarized as follows: Build. Urban Habitat, 2010.
[8] J. Kim, Y.-H. Lee, Seismic performance evaluation of diagrid system buildings, Struct.
○ The brittle behavior of diagrid models is noticeable in the pushover Des. Tall Spec. Build. 21 (12) (2012) 867–878.
curves with small ductility. Moreover, increasing the diagonal angles [9] D. Lee, S. Shin, Advanced high strength steel tube diagrid using TRIZ and nonlinear
will increase the related ductility ratios. pushover analysis, J. Constr. Steel Res. 96 (2014) 151–158.
[10] J. Teng, W.L. Guo, B.S. Rong, Z.H. Li, Z.J. Dong, Research on seismic performance ob-
○ The over-strength factor of diagrid models reduces as the diagonal jectives of high-rise diagrid tube structures, Adv. Mater. Res. 163–167 (2011)
angles increase. More specifically, the 45° models have largest 1100–1106.
over-strength factor as well as smallest ductility ratio, which [11] D.K. Lee, U. Starossek, S.M. Shin, Optimized topology extraction of steel-framed
DiaGrid structure for tall buildings, Int. J. Steel Struct. 10 (2) (2010) 157–164.
makes them behave more brittle than other considered models.
[12] N. Mashhadiali, A. Kheyroddin, Progressive collapse assessment of new hexagrid
○ The seismic performance of diagrids designed with R = 3.6 is not ac- structural system for tall buildings, Struct. Des. Tall Spec. Build. 23 (12) (2014)
ceptable in high seismic regions, i.e., for SDC Dmax. 947–961.
○ The results of seismic evaluation of diagrids designed with R = 3.6 [13] S. Mazzoni, F. McKenna, M.H. Scott, G.L. Fenves, Open System for Earthquake Engi-
neering Simulation (OpenSees)(Berkeley, California) 2006.
indicate that generally, increasing the diagonal slope enhances the [14] F.E.M.A. P695, Quantification of Building Seismic Performance Factors, Federal
seismic performance of the diagrids, because safety margin against Emergency Management Agency, Washington, D.C, 2009, June.
168 S. Sadeghi, F.R. Rofooei / Journal of Constructional Steel Research 146 (2018) 155–168

[15] ASCE 7-10, Minimum Design Loads for Buildings and Other Structures, American So- [21] Y.J. Kim, M.H. Kim, I.Y. Jung, Y.K. Ju, S.D. Kim, Experimental investigation of the cyclic
ciety of Civil Engineers, Reston, Virginia, 2010. behavior of nodes in diagrid structures, Eng. Struct. 33 (7) (2011) 2134–2144.
[16] P. Uriz, Towards Earthquake Resistant Design of Concentrically Braced Steel Struc- [22] B.V. Fell, A.M. Kanvinde, G.G. Dierlein, Large Scale Testing and Simulation of Earth-
tures, University of California, Berkeley, 2005. quake Induced Low Cycle Fatigue in Bracing Members Subjected to Cyclic Inelastic
[17] F. Yang, S.A. Mahin, Limiting Net Section Failure in Slotted HSS Braces (In Prepara- Buckling, 2010.
tion), Structural Steel Education Council, Moraga, C.A., 2005 [23] D. Lehman, C. Roeder, Improved seismic design of concentrically braced frames and
[18] D. Vamvatsikos, C.A. Cornell, The incremental dynamic analysis and its application Gusset plate connections, Struct. Congr. (2008) 1–10.
to performance-based earthquake engineering, Eur. Conf. Earthq. Eng. (2002) 10. [24] J. Baker, C.A. Cornell, Spectral shape, epsilon and record selection, Earthq. Eng.
[19] P.-C. Hsiao, D.E. Lehman, C.W. Roeder, Evaluation of the response modification coef- Struct. Dyn. 34 (10) (2006) 1193–3417.
ficient and collapse potential of special concentrically braced frames, Earthq. Eng. [25] L.F. Ibarra, M. Recardo, H. Krawinkler, Hysteretic models that incorporate strength
Struct. Dyn. 42 (2013) 1547–1564. and stiffness deterioration, Int. J. Earthq. Eng. Struct. Dyn. 34 (2005) 1489–1511.
[20] E. Karamanci, D.G. Lignos, Computational approach for collapse assessment of con-
centrically braced frames in seismic regions, J. Struct. Eng. 140 (8) (2014),
A4014019. .

You might also like