Received: 10 October 2018 Revised: 9 November 2018 Accepted: 6 December 2018

DOI: 10.1002/tal.1587

RESEARCH ARTICLE

Experimentally and analytically study on eccentrically braced

frame with vertical shear links
Mohamad G. Vetr1 | Ali Ghamari2

1
International Institute of Earthquake
Engineering and Seismology, Tehran, Iran Summary
2
Department of Civil Engineering, Darreh The excellent hysteretic behavior of eccentrically braced frames (EBFs) using shear
Shahr Branch, Islamic Azad Universality,
links have made these systems an effective alternative for both moment resisting frame
Darreh Shahr, Iran
Correspondence
as well as concentrically braced structures. Generally, horizontal shear links are located
Ali Ghamari, Department of Civil Engineering, either at the middle or at the ends of the beam. However, large deformation of the
Darreh Shahr Branch, Islamic Azad
Universality, Darreh Shahr, Iran.
shear links in those beam must be accepted under severe earthquakes. In those
Email: [email protected] instance in which the beam have to remain elastic, vertical shear links should be
designed underneath the beam thereby transferring the region of plastic deformation
to location where they are tolerable and post‐earthquake repair or replacement of
damaged parts is easier. In this paper, seismic behavior of vertical‐EBF (V‐EBF) have
been considered experimentally and analytically. The required relation to design of
the system have been proposed. In addition, to predict nonlinear behavior of V‐EBF
systems, an analytical model including kinematic–isotropic strain hardening for shear,
only kinematic strain hardening for moment, has been presented. Results showed a
good convergence of the presented analytical model (with an upper bound for yielding
surface to the V‐EBF) with experimental results.

KEY W ORDS

cyclic loading, ductility, dynamic response, experimental, shear yielding, strain hardening

1 | I N T RO D U CT I O N

Eccentrically braced frames (EBFs) have been used as a seismic load resisting system, primarily in buildings. The system, which relies on the
yielding of a horizontal link beam between eccentric braces, has been shown to provide ductility and energy dissipation under seismic loading,
and its behavior in various configurations has been investigated. In other hand, the EBF systems are metallic passive energy dampers under seismic
loading.
Despite high seismic energy dissipation, horizontal‐EBF (H‐EBF) has substantial disadvantages. It is no simple replace or repair of link beam in
the H‐EBF after a severe earthquake. This problem is the main disadvantage of this system. In addition, in industrial structures such as power
plants, deep beams are sometimes used to carry the loads of heavy and highly sensitive equipment. In these cases, to assure the efficiency and
accuracy of the equipment, load‐bearing members such as beams and columns should remain in the elastic range as much as possible. Therefore,
the use of horizontal link beams is not convenient. To overcome the problems in the EBF system, a new system called the vertical‐EBF (V‐EBF)
system was proposed.[1]
In this system, reversed‐V braces are attached to the story beam through a shear panel. Intensive inelastic deformations are localized in the
shear panel, and the internal energy is dissipated by this member. Figure 1 shows plastic hinge formation in the EBFs and V‐EBFs. Because all
inelastic deformation is localized in the shear panel, there is no damage to the main members of the frame, and repairs after a strong earthquake
are easier than in H‐EBF systems because this member can be replaced. Using the vertical links for seismic rehabilitation of the existing buildings is
possible with minor changes in the main structure.[3]

Struct Design Tall Spec Build. 2019;e1587. wileyonlinelibrary.com/journal/tal © 2019 John Wiley & Sons, Ltd. 1 of 21
https://doi.org/10.1002/tal.1587
2 of 21 VETR AND GHAMARI

FIGURE 1 Hinge formation in horizontal and vertical shear links[2]

The shear panel system is one of the simplest and cheapest passive energy dampers. The seismic performance improvement of bridges and
towers with shear panel systems was investigated.[4,5]
Because of the shear panel effect on the behavior of V‐EBF braces as a lateral load resisting system, improvement of shear panel performance
is very important. Several studies have been conducted on this system to prevent the buckling of braces.[6–8] The results showed that with proper
design, out‐of‐buckling do not occur in the vertical links of V‐EBF systems.
In the present study, the nonlinear behavior of the V‐EBF system was evaluated experimentally and theoretically. Because of researchers
reports (desirable performance of link in shear yielding), the short link without out‐of‐plane buckling are evaluated. Therefore, mathematical
models are formulated to provide elastic and inelastic behavior predictions. According to the experimental results, the require suggestion are
detailed and addressed for designing of vertical shear link. Moreover, the nonlinear behavior of V‐EBF system under earthquake acceleration is
investigated in various configurations.
The link is designed to act as a ductile fuse by yielding and dissipating energy while preventing buckling of the brace members.

2 | DESIGN OF SHEAR LINK

2.1 | Horizontal shear link

The link is designed to act as a fuse by yielding and dissipating energy while preventing buckling of the brace members. The inelastic response of a
link is strongly influenced by the link length and the Mp/Vp ratio of the link cross‐section. Using plastic analysis, the plastic shear strength, Vp, and
plastic moment strength, Mp, can be written as

F yw Fyw
Vp ¼ pffiffiffi :tw ðd − 2tf ÞVp ¼ pffiffiffi :Aw :ðd − 2tf Þ; (1)
3 3

F yw tw d2 tw :d2
Mp ¼ F yf tf ðb − twÞ:ðd − tf Þ þ Mp ¼ Fyf :tf :ðb − 2tw Þðd − tf Þ þ Fyw : ; (2)
4 2

where Fyw and Fyf are the web and flange yield strength, respectively; tw is the web thickness; d is the overall beam depth; tf is the flange thick-
ness; tw is the web thickness; and b is the flange width.
Popov and co‐workers et al., 1983–1992 have shown that horizontal beam shear link possess excellent ductile, energy‐dissipating behavior,
provided the link length is designed to allow shear yielding without buckling of the web. In fact, high isotropic strain hardening effects and com-
bined bending and shear could be observed, resulting in increased plastic‐moment and shear capacities. In the ultimate state the shear and bending
VETR AND GHAMARI 3 of 21

moment capacities reached values of about 1.5Vp and 1.2Mp, respectively. In order to impose shear yielding before bending yielding of the hor-
izontal link, the link length e needs to be limited to[9–11]

2 × 1:2Mp Mp
e≤ ¼ 1:6 : (3)
1:5Vp Vp

V p :e
In other word, classification is based on the normalized link length, ρ, defined as ρ ¼ .
Mp
Capacity design approach is followed in an attempt to limit the inelastic activity to the shear links only while all other frame members are
designed to behave elastically.
Mp
According to AISC Seismic Provision, the links with e ≤ 1:6 or ρ ≤ 1.6 are shear links that yield predominantly in shear and have a maximum
Vp
link rotation under the design seismic loading of 0.08 rad. The drift angle of the frame, θ, can then be written in terms of the link rotation angle, ɣ, as
(see Figure 1):
e
θp ¼ γp : (4)
L

There are differences between the horizontal and vertical shear links. So the equations about the calculation of the link rotation should be modified
(see Figure 2).
In addition, when the link with length e deforms inelastically and resists the applied base shear, the framing outside the link is designed to
remain elastic. The elements outside the link are designed to resist the forces generated by the fully yielded and strain hardened link. For short
links (ρ ≤ 1.6), the generated forces can be calculated as[9]link shear = 1.25 Ry Vp, link end moment at beam = Ry Vp, link end moment at
brace = [1.25 Ry Vp. e − Ry Mp] ≥ 0.75 Ry Mp,where Ry is the ratio of the expected yield strength to the minimum specified yield strength Fy pre-
scribed in AISC Seismic Provisions. This ratio is used to account for possible material overstrength. The shear strength is further increased by 1.25
to account for strain hardening. Based on results of 16 links made from A992 steel, the average strain hardening ratio was formed to be 1.28, with
a variation[12] ranging from 1.17 to 1.44. Other studies[10,13–20] have generally recommended a link overstrength factor of 1.5.
In order, the elements outside the link are designed to resist the forces generated by the fully yielded and strain hardened link.
In the next parts, the equation to design of the vertical shear link and structural elements outside the shear link are derived from experimental
test and numerical results. Therefore, the generated forces factor is modified according to strain hardening results of vertical shear link. In order to
design the V‐EBF, an analytical model is necessary.

2.2 | Vertical shear link

In order, it attempts to design vertical shear links according to horizontal links using some modification factors. It will be presented that in the
vertical links, the ultimate state of the shear and bending moment capacities reached values different from 1.5Vp and 1.2Mp for horizontal shear
link.
Therefore, the link length limitation and generated factor for designing other element outside the vertical link should be modified.
Besides, an analytical model for vertical link is necessary to predict the V‐EBF behavior. In order to compare the experimental results with
numerically derived values, it is necessary to develop an analytical model to determine the cyclic response of the V‐EBF. This analytical model
must be able to account the nonlinear cyclic response of the plastificated element.

FIGURE 2 Rotation of horizontal eccentrically braced frame

4 of 21 VETR AND GHAMARI

3 | A N A L Y T I C A L M O D E L OF I N E L A S T I C S H E A R L I N K EL E M E N T

3.1 | General

The most of the inelastic behavior of an EBF is confined to the shear link element. Because the active links (horizontal or vertical) are subjected to
large shear forces and bending moment, any appropriate formulation should include shear and bending effects at the elastic and inelastic states.
An accurate shear link model must be able to consider shear or moment plastification with strain hardening at any time of the loading. In the anal-
ysis of links performance, elastic and inelastic deformations of both the shear and flexural behaviors have to be taken into consideration.
Few researchers attempted to develop link models for the dynamic inelastic analysis of EBFs. They modeled the link as two‐dimensional linear
beam‐column element with inelastic hinges at each ends with six nonlinear rotational and translational springs at each ends. Three rotational and
translational multilinear springs were used to represent the flexural and shear inelastic behavior of the plastic hinge at the link ends represented by
the multilinear function.[20,21]

3.2 | Cyclic inelastic shear link model

The link element is modeled as two dimensional beam‐column element with inelastic hinges at each ends. Besides, it has yield surface where
the yield surface of the subhinges are arranged in a consecutive manner. All axial deformations are confined to the elastic beam between the
hinges.
Before initial yielding, the total stiffness of the complete element is that of the elastic beam, while both hinges are considered to be rigid.
Under the action of the shear and flexural forces, a reduction on element stiffness occurs that causes the hinges to yield. With the increase in
deformation, the hinges reach strain hardening condition, according to multilinear action‐deformation relationships. Strain hardening causes a
translation of the yield surface (Figure 3).
It is used the tangent stiffness between the actions and deformations of yielded subhinges according to Ricles[19] that used the Mroz the-
[22]
ory. The tangent stiffness then combined with the elastic beam stiffness to produce the tangent stiffness of the complete element for the
unloading case, the total stiffness which is dependent only on the elastic stiffness of the beam element. In this case, the rigid subhinges do not
participate in the unloading behavior. Linear link beam element has six nonlinear rotational and translational springs at each ends. Three rotational
and translational multilinear springs were used to represent the flexural and shear inelastic behavior of the plastic hinge at the link ends repre-
sented by the multilinear function shown in Figure 4. This multilinear function based on Ricles and Popov[19] and Ramadan and Ghobarah[21] is
modified for vertical shear link because of its different stiffness at two ends and strain hardening rule.

FIGURE 3 The strain hardening behavior of hinges

VETR AND GHAMARI 5 of 21

FIGURE 4 Multilinear simulation of shear link

3.3 | Strain hardening

Several researchers[17,18,20] observed that in tests of shear links, both isotropic and kinematic hardenings occur in such links that yield predomi-
nantly in shear. Based on experimental evidence, Ricles and Popov[20] concluded that the subhinges should follow an anisotropic hardening rule.
In this rule, shear yielding follows a modified isotropic hardening rule whereas moment yielding obeys only a kinematic hardening rule. After initial
yielding occurs, the behavior of a subhinge assumed to obey the Mroz[20,21] strain hardening assumptions for yield in metals. On this basis, the
yield function of an active link element is written as

ф½S; α; HðεÞ ¼ 1; (5)

where α is the vector of translation direction, and H(ε) is the isotropic expansion of the yield surface as a function of (ε), in which ε is a scalar
parameter monotonically increasing in the course of cyclic plastic shear flow (Figure 5).
Considering the anisotropic rule that assumes isotropic and kinematic hardening in shear yielding, and only kinematic hardening for moment
yielding, the yield function for the two cases become

ф½Mz − αMz  ¼ 1; (6)

ф½V y − αVy  ¼ HðεÞ: (7)

Experimental evidence suggests that there exists an upper bound function that shows and eventually stops the increase in the shear force devel-
oped by the shear hinges. A function that represents the expansion of the yield surface due to shear yielding to account for the upper bound of
shear is proposed by the Ricles & Popov[20] as following:

h
HðεÞ ¼ 0:5 ΔV max − ðΔV max − 2V y0 Þ:eð−βεÞ ; (8)

where Vy0 and ΔVmax are initial shear yield strength and maximum shear yield strength after complete hardening (Figure 6). ε is sum of positive
plastic shear strain, and β is numerical constant to be determined from a best fit of the model results with experimental measurements.

FIGURE 5 Definition of the isotropic hardening parameters

6 of 21 VETR AND GHAMARI

FIGURE 6 The section used in vertical eccentrically braced frame test and view of testing model

The use of Equation (7) implies that the yield function along a horizontal facet, where isotropic shear hardening develops, is equal to

V − αV
ф½S; α; HðεÞ ¼ ¼ 1:0 (9)
HðεÞ

V = the shear force.

When

∂ф½S; α; HðεÞ > 0: (10)

When the flexural yielding is along the vertical facet, the hardening function becomes

M − αM
ф½S; α; HðεÞ ¼ ¼ 1:0; (11)
My

where αV and αM are the translation of the yield surface along the shear and flexure axis, respectively.
The kinematic hardening in a subhinge is controlled by the parameters αV and αM that make the component of the translation vector. In the
shear link region, no interaction between bending and shear yield surface has been seen,[16,17,20,21] and a rectangular yield surface can be adopted
for the subhinges.

3.4 | Determination of plastic stiffness

The plastic stiffness matrix Kspi for each subhinge is

 
KpMi 0
Kspi ¼ : (12)
0 KpVi

The selection of the shear stiffness KpVi of subhinge i is determined by adding the elastic beam shear flexibility to that of the yielded subhinge
flexibility. It can be shown that when the Subhinge 1 yields, Vy1 ≤ V < Vy2, then the flexibility matrix is equal to

1 1 1
¼ þ : (13)
KV2 KV1 KpV1
VETR AND GHAMARI 7 of 21

Thus,

KV1 :KV2
KpV1 ¼ (14)
KV1 − KV2

When the subhinge i yields, Vyi ≤ V < Vyi + 1, the flexibility matrix becomes

1 1 1
¼ þ ∑ij¼1 ; (15)
KViþ1 KV1 KpV1

and

KVi :KViþ1
KpVi ¼ : (16)
KVi − KViþ1

In general case, the plastic stiffness coefficient Kpsi for subhinge i with the action S (shear force or moment) is defined as Sγi ≤ S < Sγi + 1 and

KSi :KSiþ1
KpSi ¼ : (17)
KSi − KSiþ1

4 | M E T H O D O F S T U DY

4.1 | Experimental study

4.1.1 | Design of test specimen

The test specimen is based on a three‐story, three‐bay frame system. The loads were recognized for dead loads of the floor‐slab system 6.50 kN/
m2, the surrounding front system to 2.5 kN/m2, and live load to 2.0 kN/m2. The lateral seismic load was calculated by applied ground acceleration
equal to 0.24 g. This condition leads to the selection of a beam stub cross‐section with a web providing the required plastic shear capacity. In order
to prevent torsional buckling, a wide flange section with inherent high lateral bending stiffness about the weak axis is selected as section for the
link beam stub. To avoid single story due to an early yielding of single link beam stub, the link length of the link beams have to be varied up to an
overall shear force distribution, which allows approximately a common yielding of all link beam stubs at the same load level of lateral forces. The
detail of the three‐story specimen is shown in Figure 6. Due to our laboratory facilities, axial load in the vertical shear link is not measured
although it may affect the shear link behavior. Because axial load reduce capacity of shear link, it is suggested that no gravity load apply to shear
link.

4.1.2 | Overall stability of the test specimen

The diagonal chevron bracing are controlled in and out of plan buckling by the member end condition. Essentially, the out of plane lateral stiffness
at the lower end of the vertical shear link effects the in plane position of the diagonal bracing. Based on the analytical study on lateral stability
condition, the same method were used to lateral bracing of the three‐story test frame. Such framing can be supplemented by vertical side bars
preventing the out of plane motion of the bottom portion of the vertical shear links.

4.1.3 | Test setup and instrumentation

The test was displacement‐controlled, using the horizontal displacement of each story level as the control parameters. The actuator loads and the
associated resulting frame floor displacement were continuously recorded. Horizontal floor displacement and local displacement in the shear link
region were measured by Linear Variable Displacement Transducers. Strain gages have been used at the selected location in the diagonal braces to
compute the brace axial force and then real shear force in the vertical shear link elements.

4.1.4 | Cyclic loading

The quasistatic loading protocol used here was developed based on the guidelines presented in ATC‐24. The applied load history was displace-
ment controlled at third floor with force distribution of 1, 2/3, and 1/3 for third, second, and first floor, respectively.
The two double‐acting actuators on each floor level introduce cyclic horizontal load with displacement control. Based on preliminary analysis,
it was estimated that the yielding and collapse of the vertical shear links occur when the third floor horizontal displacement are about ±10 and
±100 mm, respectively. The test starts with a displacement of ±4 mm in the first step and growth ±2 mm for every later step. Every step was
8 of 21 VETR AND GHAMARI

repeated three times up to the total displacement of ±14 mm. After this point, the displacement was increased to ±20 mm and later to ±30, ±40,
±50, ±75, ±100, ….

4.1.5 | Mechanical properties of material

The steel used for the beams and column was ST‐37 having yield strength = 240 MPa and elasticity modulus = 210 Gpa. Also, the minimum
required concrete compressive strength 25 MPa was used for concrete elements.

4.2 | Dynamic analysis

4.2.1 | Case study

In order to assess the dynamic seismic behavior of the V‐EBF, an eight‐story building is considered. The building was chosen to be representative
of medium‐rise office building (Figure 7). By neglecting the shear resistance of the columns, each story shear is resisted by the shear capacity of
the vertical shear link. The length of the vertical link has been chosen to ensure that the web is yielded in shear. Considering the actual shear
capacity of the vertical shear links as being 2Vp, other frame members except of the vertical links were designed with this conception. As noted
earlier, the structure was designed using the design provisions from the relevant AISC for steel construction. As a result of this relationship, the V‐
EBF designed have enlisted in Table 1.

FIGURE 7 Plan and side of overall system. V‐EBF: vertical eccentrically braced frame

TABLE 1 Structural member of numerical models

Floor level Column Brace section Floor beam Vertical link e (mm)

8 HEB‐140 2UPN‐140 HEB‐180 HEA‐160 180

7 HEB‐140 2UPN‐200 HEB‐180 HEA‐220 250
6 HEB‐140 2UPN‐220 HEB‐260 HEA‐260 330
5 HEB‐200 2UPN‐220 HEB‐260 HEA‐280 330
4 HEB‐200 2UPN‐280 HEB‐280 HEA‐300 330
3 HEB‐240 2UPN‐280 HEB‐280 HEA‐320 330
2 HEB‐240 2UPN‐280 HEB‐280 HEA‐320 330
1 HEB‐320 2UPN‐320 HEB‐280 HEA‐340 330
VETR AND GHAMARI 9 of 21

4.2.2 | Earthquake time history acceleration

In order to investigate the V‐EBF behavior, three different earthquake accelerations are performed to this study. The two of these earthquakes are
the strongest earthquake event from Iran, and the third is the Elcentro. All the three accelerograms are from the strong direction of the earth-
quakes. In order to evaluate and compare of these earthquakes effects to the response of the V‐EBF, all the three earthquakes were normalized.
Therefore, the discussion will focus on the response of the V‐EBF design to the three scaled records. The time history of these earthquakes are
shown in Figure 8.

4.3 | Numerical modeling

The DRAIN‐2DX program with developed shear link element has been used to compare the analytical model with experimental results. The devel-
oped element can consider the inelastic shear response with strain hardening effect together (Table 2).
The shear link element has the same mathematical idealization and assumption as described before. It can be modeled by combining the
hinges and a normal beam‐column element. All inelastic actions of the element were concentrated in the hinges. For the other members, a non-
linear beam‐column element was used. The diagonal braces were assumed to be connected to the floor beam with a pinned connection. The entire
frame (V‐EBF), load cell, and its relation were modeled. A quasistatic cyclic inelastic analysis with the controlled displacement was performed. The
connecting regional of all floor beam to link (panel zone) and braces to link were assumed to be rigid. The yield stress and uniaxial strain hardening
stiffness (kvi, kmi) used in this comparison were obtained from the test results in Table 3.

5 | D I S C U S S I O N A N D RE S U LT S

In this step, the results are discussed in three parts: experimental results (during the test and after that), comparison of the proposed analytical
with the test results, and numerical modeling of the building under earthquake records.

5.1 | Experimental results

5.1.1 | Performance of specimen during testing

Testing began with several cycles of inelastic displacement‐controlled loading. Between ±15 and 20 mm at the third floor, all three links yielded as
shown by cracking and flaking of the white wash. First, the third story shear link and later, the first and second story shear link were yielded,
respectively. After the shear yielding of the first vertical link, the stiffness of the link and global stiffness of the V‐EBF was decreased as inelastic

FIGURE 8 The earthquake ground acceleration time history

TABLE 2 Force‐deformation properties of link elements of the three‐story V‐EBF

Isotropic hardening
Member β ΔVmax Subhing My (kN/mm) Vy (kN) kmi kvi

HEA‐180 4.0 2.2 Vy0 1 91,940 167 0.030 0.100

2 103,890 251 0.015 0.030
3 110,330 334 0.002 0.007
HEA‐240 4.0 2.2 Vy0 1 251,661 332 0.030 0.100
2 284,382 498 0.015 0.030
3 301,990 664 0.002 0.007
HEA‐280 4.0 2.2 Vy0 1 328,821 366 0.030 0.100
2 371,570 549 0.015 0.030

Note. V‐EBF: vertical eccentrically braced frame.

10 of 21 VETR AND GHAMARI

TABLE 3 Force‐deformation properties of link elements of the three‐story V‐EBF

Isotropic hardening
Vy
Member β ΔVmax Subhing My (kN/mm) (kN) kmi kvi
HEA‐180 4.0 2.2 Vy0 1 91,940 167 0.030 0.100
2 103,890 251 0.015 0.030
3 110,330 334 0.002 0.007
HEA‐240 4.0 2.2 Vy0 1 251,661 332 0.030 0.100
2 284,382 498 0.015 0.030
3 301,990 664 0.002 0.007
HEA‐280 4.0 2.2 Vy0 1 328,821 366 0.030 0.100
2 371,570 549 0.015 0.030
3 394,582 732 0.002 0.007

Note. V‐EBF: vertical eccentrically braced frame.

deformation continued. In all three stories, strain hardening developed in the vertical links during plastic deformation following shear yielding.
Consequently, the shear force continued to increase after initial yielding.
During the ±20‐mm displacement cycles, brace bolt slippage was noted, as seen from the story response (Figure 9). The linear elastic portion
of early cycles grew slightly larger with each of these cycles. This growth is indicative of isotropic hardening of the shear link element during early
yielding. From hysteretic curves, it can be seen that the effect of the bolt slippage was more visible in the later cycles. A view of this figure indi-
cates that the loops are extremely pinched and show a serious deterioration. The test was stopped at a displacement of ±100 mm. As a result, the
energy dissipation mechanism due to brace‐bolt slippage was very poor, so the brace connections were welded. After all brace connections were
welded, the test was again started with ±6 mm and later with ±74 and +100 mm, and repeated two times for every cycle. As seen from this figure,
the brace bolt slippage and its pinching effect have been improved very much, and the hysteretic cycles became symmetric.
At every cycle, the strength of the frame continued to increase due to strain hardening during these cycles. The observation from figures
(Figure 9) indicates that the early cycles were strongly influenced by isotropic hardening whereas later cycles were influenced by kinematic hard-
ening. After the first cycle of ±47 mm at the third floor displacement, web bucking of the third floor shear occurred. However, this bucking did not
have a detrimental influence upon the test results, because the cyclic diagonal tension field formed. The web bucking also became more severe
during the second cycle of ±74 mm at the third floor displacement. After first cycle of ±100 mm at the third floor displacements, visible web tear-
ing plus flange buckling at the vertical links in the third story were formed.
On the next half cycle, web tearing developed in the third floor shear links at the upper panel, along the web stiffener. The tearing of the web
stiffener and the flange welding began to appear as second cycle of +100 mm was reversed, and it grew worse as the deflection decreased to
±50 mm. At this cycle in the first story, the web of the shear links began to buckle. The plastic shear deformation progressed more rapidly in other
shear links (in the first and second story shear link) after the third‐story link started to tear. After the link capacity was deteriorated rapidly, the test
was terminated.

5.1.2 | Lateral and lateral‐torsional buckling

The stability comparison between the two different V‐EBFS (V‐EBF with and without lateral bracing at bottom ends) indicated that no lateral or
lateral torsional bucking had occurred. Therefore, it is unnecessary to use lateral bracing at bottom ends of the vertical shear links, where it has
been connected to the diagonal brace elements. Based on this finding, it can be concluded that no lateral bracing and rotational restraint at the
lower end of the vertical shear link is needed for designing V‐EBFS. The lateral bracing at middle of the floor beams or at the upper end of the
vertical shear link elements are sufficient.

FIGURE 9 Base shear‐third displacement for (a) brace bolted connection (b) welded connection
VETR AND GHAMARI 11 of 21

5.1.3 | Diagonal bracing

At beginning of the loading, all the main diagonal bracing connections were bolted, but after the deflection greater than ±20 mm began the brace
slippage. Because of the brace slippage, the horizontal stiffness of the V‐EBF was reduced, and energy dissipation curves were pinched (Figure 10).
This stiffness loss could be corrected by welding the brace connection to prevent slippage. However, the brace connections were conservatively
designed. Thus, the pinching effect due to brace slippage in nonlarge displacement was small compared with the great dissipation provided by the
shear link element. It was evidenced that under large cyclic inelastic deflection, the response of the brace welded connection was better than the
brace bolted connection.

5.1.4 | Inelastic behavior of shear link due to web buckling

The inelastic behavior of a shear link can be categorized into either its pre‐web bucking state or post‐web buckling state. In the pre‐web bucking
state, the hysteretic of shear links is reproducible ductile and stable, and both isotropic and kinematic hardening accompanied the yielded web. In
the post‐web bucking range, the strength and stable deformation capacity of shear link may degrade either gradually (over a small number of
cycles) or suddenly, depending on its topography. Heavily stiffened shear links to omit the web bucking are prone to fail in a brittle or tearing man-
ner along the perimeter of the link panel. In shear links with unstiffened webs, the tearing occurred near the center of the buckled region. The
material tearing failure of the shear link panel is generally caused by stress concentration that developed at the anchorage points of the diagonal
tension field. The post‐web buckling behavior of a stiffened shear link is excessive and impossible to predict. Because the stale response of an EBF
to severe earthquake depends primarily on the ductile, stable of its shear links, the post‐web bucking region is not a desirable collapse limit state
for shear link response. However, the web bucking could be happened after the web yielding and a large amount of strain hardening. It becomes
particularly severe during ±47 to ±100 mm, but because of the expected cyclic diagonal tension field, formed and reformed under cyclic loading
without causing any pinching of the hysteretic curves or deterioration in strength at the same cycle but causes slowly the stiffness degradation.
The diagonal tension field also restraightened the web on the next half cycle after the web had buckled but imposed a failure mechanism where a
tear progressed around the perimeter of the third floor bucked panel.

FIGURE 10 Experimental hysteresis curves of the vertical eccentrically braced frame

12 of 21 VETR AND GHAMARI

5.1.5 | Ultimate and initial strength

The curves of experimental shear force versus shear strain and interstory drift indexes for all three stories are shown in Figure 10. As mentioned
before, because of string hardening, the shear link could resist the loading greater than the nominal shear yielding.
According to these figures, the ultimate strength of shear link reaches more than 2.2 times of nominal shear yielding. Therefore, this effect
must be accounted for in the design of the other structural member outside the shear link. Therefore, this effect must be accounted to the design
of the other members in V‐EBF. In all the stories, load corresponding to initial yielding shear yielding was occurred below the force at the Vp the-
oretical values of shear yielding (5%, 6%, and 7% for first, second, and third, respectively). So the generated factors are proposed as follows:
In the ultimate state, the shear and bending moment capacities reached up to 2.2Vp and 1.2Mp, respectively. In order to impose shear yielding
before bending yielding of the vertical link (by accepting the Vp = 0.9Vy in studies of Ramadan and Ghobarah[21]), the link length e needs to be
limited to

2 × 1:2Mp 2 × 1:2Mp Mp
e≤ ¼ ¼ 1:2 ; (18)
2:2 × V p 2:2 × ð0:9 × V y Þ Vy

and generated factor should be used as 1.75 Ry Vy.

5.1.6 | Energy absorption

The result of the V‐EBF test can be summarized by saying that the test frame performed very well. It was able to withstand several lateral deflec-
tions and dissipated a large amount of imposed energy without any significant pinching of the hysteretic loops. Figure 10 indicates how the inelas-
tic deflections of the V‐EBF are attained. This plot subdivides the deflection into the three components. It clearly indicates that virtually, the
inelastic deflection of the V‐EBF was caused by the inelastic deflection of the vertical shear link element.
Most of the story shear force could be transferred and resisted through the vertical shear link element, and only a few percent of the story
shear were resisted by the columns (see Figure 11).
Also, as mentioned before that, the damage is confined on shear links at V‐EBF systems. Referring to Figure 12 confirms it, and this figure
shows that the dissipated energy was distributed quite uniformly among the shear link elements of the test frame (V‐EBF). This uniform energy
dissipation is a very desirable feature, as indicates that the shear link elements were deforming plastically, and no element was excessively strained
as evidence by observations and investigations of all the other elements.

FIGURE 11 The comparison between story shear and vertical shear link force

FIGURE 12 Energy dissipation of the tested vertical eccentrically braced frame

VETR AND GHAMARI 13 of 21

5.1.7 | Two end moments of shear links

The moment versus rotation at two ends of the vertical shear links obtained specimen tested is shown in Figure 13. Results indicate that the ver-
tical links have a stable hysteresis loops. But moments at two ends show a different performance due to its end condition stiffness. At the shear
yielding of the link web, the ratio of the moment of the bottom and upper end is shown in Table 2. In the upper link region, located at the floor
beam, high bending moment can be activated, whereas at the lower region, adjacent to the chevron‐braced connection, the bending moments are
typically smaller depending on the rotational end stiffness of the braces. The end stiffness effects have to be considered in determining the length
of the vertical link. With elastic link end moment and the conception consideration reflected in Equation (19), the following design condition for
link length could be specified:

Mp
e ≤ 0:35ðK þ 1Þ ; (19)
Vp

where the K factor is ratio of the moment at floor‐beam divide to moment at end bottom. To accommodate large shear distribution in the
web and prevent weld failure at the interface between link and floor beam, it is proposed to limit the moment at the upper end of the vertical
link effectively to an elastic design moment of 0.5Mp. If the relation be utilized for designing of V‐EBF, no failure will occur in the weld.

5.2 | Corrective of analytical model

It must be recognized that load cell force distribution among three stories was similar to that of namely 1, 2/3, and 1/3 for the third, second, and
first floor, respectively. A preliminary study of the experimental result indicates that the amount of axial force developed in the vertical link was
small as predicted before the test because of axial symmetric form of the V‐EBF. Therefore, the axial force effect can be ignored in the link capac-
ity design. The envelope of the shear force resisted by the shear link during the loading is shown in Figure 14, in conjunction with the strength of
each story of the V‐EBF.
The multilinear values governing the shear force‐deformation and bending moment‐rotation relationships shown in Figure 4 were calibrated
using the test results. Compared with the horizontal shear, the numerical formulation of the shear and moment stiffness calibration factors for
vertical shear link have been formulated using basically the following characteristic values.

a. For shear (V values) and moment (M values):

Vy1 ¼ 1:00 Vy My1 ¼ 1:00 My

Vy2 ¼ 1:50 Vy My2 ¼ 1:03 My

Vy3 ¼ 2:00 Vy My3 ¼ 1:06 My

b. The values of K2, K3, K4 for shear and moment are given relative to K1 as follows:

KV2 ¼ 0:100 KV1 KM2 ¼ 0:030 KM1

KV3 ¼ 0:030 KV1 KM3 ¼ 0:015 KM1

FIGURE 13 Two end moment behavior of shear links

14 of 21 VETR AND GHAMARI

FIGURE 14 The predicted response by analytical

KV4 ¼ 0:007 KV1 KM4 ¼ 0:002 KM1

where

3EI GAw
c. KM1 ¼ ; KV1 ¼
e e

5.2.1 | Strength and stiffness

Comparison between the experimental and predicted model (Figure 14) indicates that the lateral elastic stiffness of the test frame is less than that
predicted analytically (a good acceptable different). This fact can be attributed in part to flexibility of the test foundation (assumed to be rigid in
the mathematical model), which has been explained in the earlier section. Moreover, comparison of global response shows that the different
between the test and analytical model are lower than 30% for strength and lateral stiffness.

5.2.2 | Hysteretic response

Test observations showed that the first plastification arises in the third story shear links and later in the first and second story shear links
approximately at the same time. It is the same that was predicted through the numerical computation by the defined vertical shear link element.
As can be seen in the shear link response, the proposed model could cover the test result approximately. The comparison is not at all exact, but it is
VETR AND GHAMARI 15 of 21

good for inelastic analysis purposes. Based on the dissipated energy curves (the area under the shear force–shear strain curves) comparison, it can
be found that they are fitted more than 95%. for this test influence of second‐order effects (P‐Δ effect) is not observed significantly.
The comparison studies have shown that the so‐developed shear link element in DRAIN‐2DX program can accurately predict the cyclic inelas-
tic response of the V‐EBF provided that reasonable estimates of the mechanical properties of the structural members are made.

5.3 | Dynamic response

The action‐deformation relationships for the links of the various analyses are given in Figure 15. Based on the analytical results, the regions in
which yielding occurred in whole analyses were found similarly. These relationships were used to establish the subhinge stiffness. Both action‐
deformation relationship and isotropic hardening parameters were based on the experimental data that discussed before.
Considering the influence of the P‐Delta effect on the seismic response of the frame, the results of earthquake (Elcentro) were compared with
the results of an addition analysis, which did not include a geometric stiffness (P‐Delta effect). In all the analyses, nonproportional damping were
imposed.
The maximum developed‐shear force and rotation in vertical links were shown in Figures 16–18. The greater amount of the links strain hard-
ening developed in the lower floors. This effect can be seen in Figure 16, by the increase in the link shear force beyond the yield strength (Vp).
Figure 18 indicates that the first floor developed a maximum shear more than 2.0Vp during all the three earthquakes. These figures indicate
that greater link deformation occurred during the response of the Artificial earthquake. It was found that the plastification takes place only in the
vertical shear link but not at the same time for the three analyses (also see Figure 19).
In all figures, the meaning of the EPP is Elastic Perfectly Plastic. As shown in Figures 20 and 21, the influence of second‐order effects (P‐Delta)
for V‐EBF systems are significant. In addition, the EPP model developed more of a one‐side lateral floor displacement time history, leading to
larger magnitudes of lateral displacement than the analyses with strain hardening. The one‐sided displacement is an indication of permanent link
deformation resulting from the link yielding.

FIGURE 15 The maximum links shear forces

FIGURE 16 The links deformation envelops

16 of 21 VETR AND GHAMARI

FIGURE 17 (a) Maximum story shear with strain hardening EPA = 0.52 g. (b) Normalized maximum shear link shear force

FIGURE 18 Link hysteresis response for earthquakes of (a) Tabas, (b) Elcentro, and (c) Artificial

The maximum story drifts during all the three analyses (with strain hardening and P‐Delta effect) are shown in Figure 21b. As seen from this
figure, the maximum story drifts during the Elcentro occur in the upper floors and for the Tabas and artificial earthquakes occur in the lower floors.
A comparison between the strain hardening model and EPP model indicates that the EPP model developed greater story drift (Figure 22a).
However, in both models, the developed maximum story drift is smaller than code limit of 2%.
As can be seen from Figure 22, the story drifts for the two models (with P‐Delta effect and without it) have the same magnitude,
approximately.
VETR AND GHAMARI 17 of 21

FIGURE 19 Floor displacement with strain hardening and P‐Delta effect, EPA = 0.52 g

FIGURE 20 Floor displacement with EPA = 0.52 g and (a) strain hardening (b) P‐Delta effect

Figures 22 and 23 indicate the maximum story shear envelopes with strain hardening and without strain hardening or EPP, respectively. The
limited resistance of the model without strain hardening is evident in the base shear envelops are shown in Figure 22, where it is compared with
the results of the model with strain hardening effect caused larger story shears, with a greater increase occurring in the lower floors. By including
shear strain hardening, the base shear increase up to 90%.
18 of 21 VETR AND GHAMARI

FIGURE 21 (a) Maximum story drift ratio with strain hardening, EPA = 0.52 g. (b) Comparison of the maximum story drift ratio between Elastic
Perfectly Plastic (EPP) and strain hardening

FIGURE 22 (a) Maximum story shear with strain hardening, EPA = 0.52 g. (b) Comparison of the maximum story shear between Elastic Perfectly
Plastic (EPP) and strain hardening

FIGURE 23 (a) Maximum axial column force envelops with strain hardening, EPA = 0.52 g. (b) Maximum axial brace force envelops with strain
hardening, EPA = 0.52

The brace and columns of the models were checked after the completion of each analysis and found to have developed no instability prob-
lems. The maximum axial force and moment developed in the columns and braces were shown for all the three earthquakes in Figures 24 and 25.
The increase in the axial columns and brace's forces are because of vertical link strain hardening.
The energy time histories in Figure 26 indicate that cyclic response to the scaled Elcentro earthquake record resulted in greater amount of
energy dissipation compared with the response to the Tabas and Artificial records. This occurred in spite of the fact that the artificial earthquake
record caused larger shear forces in links and base shear to develop. An elevation of the hysteretic and damping energy dissipated by the V‐EBFs
under the Elcentro indicates that dissipated energy through the hysteretic is more than 80% of the total dissipated energy and through the
damping is up to 15% of the total dissipated energy. An energy dissipation comparison between the two analyses with and without P‐Delta effect
indicates apparently, the energy dissipated by the both analyses was not significantly influenced by P‐Delta effect.
VETR AND GHAMARI 19 of 21

FIGURE 24 Comparison of the maximum axial column force between Elastic Perfectly Plastic (EPP) and strain hardening

FIGURE 25 Energy time history of the earthquake, EPA = 0.52 g

20 of 21 VETR AND GHAMARI

FIGURE 26 Comparison of the energy time history between (a) strain hardening and Elastic Perfectly Plastic (EPP), without P‐Delta effect and
(b) between strain hardening and EPP, with P‐Delta effect

6 | C O N CL U S I O N S

From the study of theoretical and experimental results, the following major conclusions can be derived:

• All the approaches studied in this paper (experimentally and numerically studies, cyclic static and dynamic analyses) showed that the main
inelastic behavior can be confined in vertical shear links. It is able to withstand lateral deflections and dissipated a large amount of imposed
energy without any significant pinching of the hysteretic loops.

• It can be concluded that no lateral bracing and rotational restraint at the lower end of the vertical shear link is needed for designing V‐EBFs.
The lateral bracing at middle of the floor beams or at the upper end of the vertical shear link elements are sufficient.
• Ultimate strength of shear link reaches more than 2.2 times of nominal shear yielding. Therefore, this effect must be accounted to design of
the other members outside the shear links. It is proposed to use 1.75 Ry Vy for this effect.

• To accommodate large shear distribution in the web and prevent weld failure at the interface between link and floor beam, it is proposed to
limit the moment at the upper end of the vertical link effectively to an elastic design moment of 0.5Mp.
• By using the proposed relation of vertical links, in the dynamic analysis, most of the story shear force could be transferred and resisted
through the vertical shear link element. Also, uniform energy dissipation was a very desirable feature, as it indicated that the shear link ele-
ments were deforming plastically, and no element was excessively strained as evidence by observations and investigations of all the other
elements.
• Comparison between the experimental and proposed analytical model results indicated a good convergence in lateral elastic stiffness, ultimate
strength, and energy dissipation. Based on the dissipated energy curves (the area under the shear force–shear strain curves) comparison, it can
be found that they are fitted more than 95%. It was also found that for this test building influence of second‐order effects (P‐Delta effect) to
be insignificant.
• It was found that the plastification take place only in the vertical shear link but not at the same time for the three time history analyses. In
addition, influence of second‐order effects (P‐Delta) for V‐EBF systems are significant, and EPP model developed more of a one‐side lateral
floor displacement time history, leading to larger magnitudes of lateral displacement than the analyses with strain hardening.
• An elevation of the hysteretic and damping energy dissipated by the V‐EBFs indicated that dissipated energy through the hysteretic is more
than 80% of the total dissipated energy and through the damping is up to 15% of the total dissipated energy. An energy dissipation compar-
ison between the two analyses with and without P‐Delta effect indicates apparently, the energy dissipated by both analyses was not signif-
icantly influenced by P‐Delta effect.

ORCID
Ali Ghamari https://orcid.org/0000-0003-4204-1743

RE FE R ENC E S
[1] M. A. Shayanfar, A. R. Rezaeian, A. Zanganeh, Struct. Des. Tall and Special Build. 2012, 23, 1.
[2] A. S. Kazemzadeh, C. Topkaya, J. Constr. Steel Res. 2017, 128, 53.
[3] A. Ghobarah, H. Abou Elfath, Eng. Struct. 2001, 23, 745.
VETR AND GHAMARI 21 of 21

[4] M. Lian, S. M, Struct. Des. Tall Spec. Build. 2018, 24, e1455.
[5] J. Berman, M. Bruneau, J. Earthq. Eng. Struct. Dynam. 2002, 34(4–5), 497.
[6] A. Karalis, K. Stylianidisa, Earthq. Struct. 2013, 5(2), 143.
[7] M. D'Aniello, G. Corte, F. Mazzolani, (2006), Seismic upgrading of RC buildings by steel eccentric braces: Experimental results vs. numerical modeling,
Conference Steel Structures in Seismic Area, STESSA Japan.
[8] R. Montuori, E. Nastri, V. Piluso, J. Constr. Steel Res. 2014, 92, 122.
[9] C. Shih‐Ho, C. Subhash, (2005), Performance‐base seismic design of EBF using target drift and yield mechanism as performance criteria, Research
Report. UMCEE 05‐05. Department of Civil and Environmental Engineering. University of Michigan.
[10] E. Popov, M. Engelhardt, J. Constr. Steel Res. 1984, 10, 321.
[11] AISC, Seismic Provisions for Structural Steel Buildings, Chicago 2016.
[12] G. Arce, (2002), Impact of higher strength steels on local buckling and over‐strength of links in eccentrically braced frames. Master's Thesis, Depart-
ment of Civil Engineering, University of Texas at Austin, Austin, TX.
[13] E. Popov, V. Bertero, J. Eng. Mech. Div. 1980, 106(1) (1980)), 75.
[14] J. Malley, E. Popov, (1983), Design consideration for shear link in eccentrically braced frames. Report No. UCB/EERC‐83/24, Berkeley, California,
1983.
[15] C. Roeder, E. Popov, J. Struct. Div. 1987, 104(3), 391.
[16] K. Hjelmstad, E. Popov, J. Struct. Eng. 1983, 109(10), 2387.
[17] J. Malley, E. Popov, J. Struct. Eng. 1984, 110(9), 2275.
[18] K. Kasai, E. Popov (1986), Study of seismically resistant eccentrically braced steel frame systems. Report No. UCB/EERC‐86/01. Earthquake Engineer-
ing Research Center.
[19] J. Ricles, E. Popov, (1987), Dynamic analysis of seismically resistant eccentrically braced frames. Report No. UCB/EERC‐87/07 Berkeley (CA): Earth-
quake engineering research center, University of California.
[20] J. Ricles, E. Popov, Experiments on eccentrically braced frames with composite floors. Report No. 87/06, Earthquake Engineering Research Center, Uni-
versity of California, Berkeley 1987.
[21] T. Ramadan, A. Ghobarah, J. Struct. Eng. ASCE 1995, 121(11), 1574.
[22] Z. Morz, Acta Mech. 1969, 7(2‐3), 199.

AUTHOR BIOGRAPHI ES
Mohamad Ghasem Vetr received his BS degree in Civil Engineering from Gillan University, MSc from Tarbiat Modares University in Iran, and
PhD from Darmstud University in Germany. He is currently a full faculty member at International Institute of Earthquake Engineering and Seis-
mology. His research interests includes seismic behavior of structures especially steel structures.

Ali Ghamari is a scientific member at ACI‐Iran chapter. He received his BS degree in Civil engineering from Ahvaz Uvi, MSc of structural engi-
neering from Yazd Uni, and PhD from Iran University of Science and Technology. He is currently a faculty member at Islamic Azad University
(IAU). In addition, he is a visiting lecturer at Aria University of Science and Sustainability and a counselor of Ilam‐Science and Technology Park.
In 2014, he got Best Researcher Awards, Islamic Azad University. His research interests includes seismic behavior of structures.

How to cite this article: Vetr MG, Ghamari A. Experimentally and analytically study on eccentrically braced frame with vertical shear links.
Struct Design Tall Spec Build. 2019;e1587. https://doi.org/10.1002/tal.1587