You are on page 1of 18

A Horizontal Stiffener Detailing for Shear Links

at the Link-to-Column Connection in


Eccentrically Braced Frames
Shih-Ho Chao, P.E., M.ASCE 1; Ghassan Almasabha 2;
Brandon Price, P.E., S.E. 3; and Chatchai Jiansinlapadamrong, P.E. 4

Abstract: This study investigates the performance and viability of a horizontal stiffener detailing (HSD) to (1) be used at the link-to-column
connection in a D-braced eccentrically braced frame layout where conventional link-to-column connections show poor performance,
(2) minimize the overstrength factor of shear links caused by the hardening of link flanges, and (3) mitigate an undesirable failure mechanism
previously observed in shear links fabricated using ASTM A992 steel. HSD and conventional stiffener detailing (CSD) were tested using
W460 × 60 sections under two loading protocols, including one simulating near-collapse excitation. While the flanges of CSD shear links are
greatly stiffened by vertical stiffeners, webs of HSD shear links exhibit inelastic buckling at large rotation angles. This allows the flanges to
freely deform, thereby reducing the likelihood of brittle fracture at the flange connections, as well as undesirable link overstrength induced by
the flange contribution. This study shows that the horizontal stiffener configuration developed a ductile failure mechanism resulting from
gradually increasing local buckling in the web. The strength degradation of links with horizontal stiffener detailing was much more gradual
due to the ever-increasing web buckling, unlike the sudden brittle fracture at the flange-to-column face of links with CSD. This research
shows that the HSD is a viable and economic alternative for shear links, which has the potential to decrease welding in the shear link while
exhibiting adequate seismic performance. HSD also maintains a low overstrength factor and utilizes a simplistic design approach. Finally, the
bolted extended end-plate connection and its details at the link ends used in the test experiment provided a viable solution for replaceable
links. DOI: 10.1061/JSENDH.STENG-12090. © 2023 American Society of Civil Engineers.
Author keywords: Eccentrically braced frames; Shear link; Steel frames; Seismic design; Seismic analysis; Steel columns; Digital image
correlation.

Introduction viability of EBFs. Active links are classified based on the length
of the link, e, as either shear yielding links (e ≤ 1.6 M p =V p ),
Eccentrically braced frames (EBFs) combine high elastic stiffness intermediate links (1.6 M p =V p ≤ e < 2.6 M p =V p ), or flexural
similar to that of buckling-restrained braced frames with the high yielding links (e ≥ 2.6 M p =V p ) (AISC 2016a). Typically, active
ductility of special moment frames. As the name suggests, EBFs links, which fall into the shear yielding link (short link) categori-
are frames in which the braces do not attach to the beam concentri- zation, are preferred because they provide a frame with an elastic
cally. The eccentric segment of the beam is referred to as the active lateral stiffness similar to buckling-restrained braced frames while
link. During a strong earthquake, the inelastic activity within the maintaining exceptional ductility and energy dissipation capacity
frame is contained within the active link and the first-floor column, (Malley and Popov 1984; Hjelmstad and Popov 1983; Kasai and
limiting the force developed in nondesignated yielding members. Popov 1986; Engelhardt and Popov 1989). Also, unlike a flexural
The design forces for the members outside of the active link are yielding link (long link), the shear strength of a shear yielding link
determined using a capacity-design approach based on the adjusted is independent of the link length, e, which means the frame’s
link shear strength considering the material overstrength and strain- strength does not change if the link length is adjusted due to archi-
hardening of the link (AISC 2016a). As such, the performance of tectural or detailing requirements. In addition, the design moment
the active link is paramount to the safety, reliability, and economic of the beam outside of the link can be reduced by using shear yield-
1
ing links because the end moments for shear yielding links are
Professor, Dept. of Civil Engineering, Univ. of Texas at Arlington,
smaller than for longer links (AISC 2016a).
Arlington, TX 76019 (corresponding author). ORCID: https://orcid.org
/0000-0003-2679-7364. Email: shchao@uta.edu
2
Assistant Professor, Dept. of Civil Engineering, Faculty of Engineer-
High Stress-Triaxiality and Fracture at Weld–Web
ing, The Hashemite Univ., Zarqa 13133, Jordan. ORCID: https://orcid.org
/0000-0002-1414-3913. Email: ghassans@hu.edu.jo
Interface
3
Regional Manager, DN Tanks, 410 E Trinity Blvd., Grand Prairie, A previously unobserved shear link failure mode was revealed by
Dallas, TX 75050. Email: brandon.price@dntanks.com experimental tests conducted on wide-flange shear links reinforced
4
Project Engineer, AG&E, 15280 Addison Rd., Addison, TX 75001. with closely spaced vertical stiffeners (Arce 2002; Galvez 2004;
Email: chad_a2@hotmail.com
Okazaki 2004). These specimens developed a ductile web fracture
Note. This manuscript was submitted on September 21, 2022; ap-
proved on April 4, 2023; published online on June 19, 2023. Discussion before the onset of web-buckling initiating from the ends of the
period open until November 19, 2023; separate discussions must be sub- vertical stiffener welds and propagating horizontally across the
mitted for individual papers. This paper is part of the Journal of Struc- web of the shear link. This failure mode prevented several speci-
tural Engineering, © ASCE, ISSN 0733-9445. mens from obtaining the mandated inelastic rotation angle (IRA)

© ASCE 04023114-1 J. Struct. Eng.


Fig. 1. Distortion at the end of stiffener and resulting highly stressed region at the web of the shear link with conventional vertical stiffeners.

required by the AISC Seismic Provisions (AISC 2016a). This fail- in link-to-column connections of D-braced EBFs (Okazaki 2004;
ure mechanism was thought to be the results of (1) concentrations Okazaki et al. 2004).
of stress and strain close to the k-area at the termination of the stiff- In a link-to-column connection (Fig. 3), the moment developed
ener welds (McDaniel et al. 2003; Dusicka et al. 2004; Richards at the column side is higher than the moment developed at the
2004), (2) the use of an excessively demanding loading protocol beam–brace side of the link due to the greater rotational restraint
(Richards and Uang 2003), and (3) the susceptibility of the k-area at the column face than at the beam–brace end of the link. In the
to low-cycle fatigue (Okazaki et al. 2005). Nonlinear finite element elastic range, the moment at the column face (M c ) is higher than the
analysis was carried out by Chao et al. (2006) who concluded that moment at the beam face (M b ); however, once the yielding occurs
the observed early ductile fracture at the weld–web interface was in the shear link, the ratio of M c =Mb gradually decreases to reach a
the result of the high triaxial constraints that develop at the weld value close to unity due to the moment redistribution after yielding
ends coupled with elevated local strain demands in this region. The the link web.
authors found that the current vertical stiffener layout inevitably For example, for shear links where e=ðM p =V p Þ ¼ 1.1 and
caused high local strain concentrations in the link web near the e ¼ 635 mm, Okazaki (2004) found that the M c =M b ratio ranged
end of the stiffeners, as shown in Fig. 1. Chao et al. (2006) ana- from 1.64 to 1.83 in the elastic range; however, these ratios gradu-
lytically showed that the conventional vertical stiffeners can be re- ally decreased to reach a ratio close to unity after the yielding in
placed by a horizontal stiffener (Fig. 2) to alleviate the local strain shear links occurred. Therefore, the link at the column side reaches
concentration at the web-to-stiffeners weld termination and the high its plastic moment strength before the beam–brace side. Once the
stress-triaxiality. column side enters its inelastic state, moment redistribution com-
mences, and the beam–brace side gradually reaches its plastic
Fracture of Flange Near Column Face moment strength. The unequal moments at both ends of a link-
to-column connection cause a combined high moment and shear
Current AISC 341 (AISC 2016a) recommends avoiding the
D-braced EBF configuration where the link is attached to the col-
umn, i.e., a link-to-column connection (Fig. 3). The load and de-
formation demand at a link-to-column connection in a D-braced
EBF is substantially different from that of a beam-to-column con-
nection in a special moment frame. A shear link at a link-to-column Stiffeners
connection sustains a much higher moment gradient than a typical Beam
moment connection. Prior experimental results have indicated that
moment connections, showing good performance in moment frame
beam-to-column connections, can demonstrate poor performance

Link-to-column
connection
Brace

Column

Fig. 2. Horizontal stiffener for shear links. (Reprinted from Chao et al. Fig. 3. Wide-flange shear link at a link-to-column connection in a
2006, © ASCE.) D-braced EBF.

© ASCE 04023114-2 J. Struct. Eng.


forces to rise in the link near the column face. Notably, Okazaki of the web and the transfer of shear forces to the flanges after yield-
et al. (2009) reported that nearly all link-to-column connections ing the web link. This can potentially reduce the overstrength of the
failed due to link flange fracture, including two connections that shear link, which in turn reduces the required forces for the non-
used improved details and had reached large link rotation angles. yielding members.
The reason that the fracture initiated from the link flange could be This study experimentally investigated the seismic performance
attributed to the fact that flanges in a wide-flange shear link rein- of wide-flange steel links with horizontal stiffener detailing pre-
forced with vertical stiffeners can carry a significant amount of sented by Chao et al. (2006). First, a proof-of-concept specimen
shear that was traditionally assumed to be completely carried by was tested to determine the viability of the proposed stiffener con-
the web (McDaniel et al. 2003; Richards 2004; Ji et al. 2016; figuration. Four large-scale specimens were then tested (two with a
Yin et al. 2020). Therefore, link flanges at the column face as well horizontal stiffener and two with vertical stiffeners). The specimens
as the welds connecting the link flange and the column face are were tested under two loading protocols: one outlined in Section
subjected to high combined bending and shear stresses. K2.4c of the AISC Seismic Provisions (AISC 2016a); the other
The shear distributions among the link flanges and web depend is a near-collapse (NC) loading protocol developed by Chao and
on their relative stiffness in resisting the shear. Past research has Price as documented in Price’s thesis (Price 2015). Finally, nonlin-
indicated that reducing the stiffness by partially separating it from ear finite element analyses of shear links with horizontal stiffener
the web can effectively shift the shear force into the beam web of a detailing were carried out to investigate the application of horizon-
free-flange moment connection (Choi et al. 2003). On the other tal stiffeners in deeper sections.
hand, Li et al. (2009) reported that vertical stiffeners added in
the plastic hinging zone of a reduced beam section moment con-
nection can effectively delay local buckling, but the von Mises Experimental Program
stress in the flange has a noticeable increase. In this study, the au-
The testing program was carried out to explore the feasibility of
thors hypothesize that the vertical stiffeners used in the conven-
using horizontal stiffeners in shear links attached to columns, such
tional shear links increase the stiffness of the link flanges, which
as the shear links used in D-braced EBFs. The current practice in
is one of the contributing factors causing the high cumulative plas-
AISC 341 (AISC 2016a) requires the use of vertical stiffeners to
tic strain in the flange and its fracture. This identification of in-
reinforce shear links. Section K2.4c of AISC 341 provides a load-
creased stiffness in the link flanges due to the vertical stiffeners
ing protocol to test link-to-column connections. To qualify as link-
also encourages replacing the vertical stiffeners with horizontal
to-column connections, the shear links must sustain their plastic
stiffeners to reduce the flange stresses at the column side. This
shear strength for at least one complete loading cycle at an inelastic
replacement would shift the failure mode from the brittle fracture
link rotation angle of 0.08 rad (8%), in accordance with Section
of link flanges at the column face to a safer ductile plastic defor-
K2.8 of AISC 341.
mation in the link flanges and the link web. The location of failure
initiation can also be shifted away from the link-to-column connec-
tion where severe stress conditions exist. Test Setup
Fig. 4(a) shows the test setup for W460 × 60 shear link specimens,
which was designed and constructed to generate internal moment
High Overstrength of Shear Links
and shear force distributions that are similar to that of an actual
Link overstrength is due to strain hardening in a link and is defined link-to-column connection in a D-braced EBF, where the moments
as the ratio between the maximum shear capacity, V u , and the plas- are unequal at both ends of the shear link, and the shear force is
tic shear strength, V p , in a link. Hence, the plastic shear strength is constant over the link length. The test setup consisted of a W360 ×
computed using V p ¼ 0.6Fy actual ðd − 2tf Þtw based on actual yield 287 column on one side of the shear link and a rigid L-arm on the
stress (AISC 2016a; Bruneau et al. 2011). d is the overall depth of other side of the link. One end of the link was rigidly connected to
the link, tf is the thickness of the flange, and tw is the thickness of the W360 × 287 column attached to a stiff H-frame having a col-
the web. The overstrength of a shear link must be estimated when umn of W840 × 527. This boundary condition imposed a high rota-
designing the nonyielding members outside of the links such as tional restraint like the column end of a shear link in a D-braced
beams, braces, columns, and connections, which are designed by EBF. The other end was connected to a rigid L-arm fabricated using
the maximum forces imposed on these members due to fully W360 × 287 sections supported by stabilizing links. Stabilizing
yielded and strain-hardened links. For economic reasons, AISC links were provided in-plane on the outside of the L-arm to restrain
341 (AISC 2016a) suggests an average overstrength value of 1.25 its rotation. Axial deformation was allowed in the W460 × 60 shear
for wide-flange links, although past research has generally recom- link specimens because of the slightly larger hole size than the pin
mended an overstrength factor of 1.5 (Popov and Engelhardt 1988). size in the stabilizing links, which provided greater rotational flex-
Numerous studies have found that the actual overstrength of wide- ibility to that end of the link, similar to the beam end of a shear link
flange links can be 20% to 60% greater than the value specified by in a D-braced EBF. The moment ratio between the column end and
AISC 341. Three major factors have been suggested (McDaniel the beam end [Fig. 4(a)], M c =M b , will be discussed later. The
et al. 2003; Richards 2004; Okazaki and Engelhardt 2007; Della L-arm was attached to a closed-loop servo-controlled actuator,
Corte et al. 2013; Ji et al. 2016; Kazemzadeh Azad and Topkaya which applied forces to the link’s midspan. A lateral bracing system
2017; Liu et al. 2020) as overstrength causes: (1) axial restraints at was used for the L-arm to prevent any undesired torsion in the
the link ends provided by other structural members, especially at specimen during testing. Notably, the test setup for the proof-of-
large link rotation angles; (2) shear resistance of flanges especially concept (POC) shear link specimen [Fig. 4(b)] was different from
relatively thick flanges; and (3) very large plastic strains due to the one shown in Fig. 4(a). The stabilizing links in the test setup, as
considerable cyclic strain-hardening of steel in the link web. To shown in Fig. 4(b), provided a high axial restraint and did not im-
improve the overstrength-causing factors (2) and (3), the use of pose unequal end moments that occur in a D-braced EBF. The shear
horizontal stiffeners eliminates the stiffening effect on the link link specimen was connected to the test setup using an extended
flanges by conventional vertical stiffeners and allows link web in- end-plate moment connection with 19 mm diameter A193 B7 rods.
elastic buckling. These factors minimize the cyclic strain-hardening B7 rods were used because A490 bolts are not recommended for

© ASCE 04023114-3 J. Struct. Eng.


Test setup for W460×60 links 1335 kN
Actuator

Reaction
Column
Rigid L-Arm

Reaction e/2
1/2"
H-Frame
End-plate
Column Column

1 5/8"
Stabilizing

1168
Links Shear Link

LVDT
e
1"

Column
Beam End
End

Reaction Floor Lateral bracing LVDT

(a)

Test setup for POC-AISC 445 kN Actuator


W150×29.8 link

Reaction
Column
Stabilizing
Reaction Link
Column Reaction
Column

Stabilizing
Links

Link Specimen Stabilizing


Link

Reaction Floor

(b)

Fig. 4. Details of the test setups: (a) test setup for W460 × 60 shear link specimens; and (b) test setup for W150 × 29.8 shear link specimen.

retightening. In addition, while B7 rods and A325 bolts have a sim- the demand imposed on the active link during a design-level ground
ilar strength, the cost of B7 rods is approximately 60% less than motion (Richards 2004). The displacement history is symmetrical,
that of A325 bolts with the same length. Link rotation angles were which subjects the link to equivalent displacements in each direc-
computed using the relative displacement between two ends of the tion. The near-collapse (NC) loading protocol simulates the dis-
link divided by the link length, minus the rotation of the L-arm, placement demands during a near-collapse ground motion, where
which is typically negligible. a structure is near collapse under a maximum considered earth-
quake (MCE) level ground motion. This type of ground motion
is required for nonlinear time-history (NTH) analysis carried out
Loading Protocols following FEMA P695 (FEMA 2009). In the near-collapse situa-
The two loading protocols used in this study are shown in Fig. 5 tion, the response is featured by a ratcheting effect, as shown in
and Table 1. The AISC loading protocol follows the required load- Fig. 5. NTH analyses indicate that the shear link’s rotation angle
ing history prescribed in Section K2.4c of AISC 341 and simulates history is unsymmetrical with one-sided displacements in one di-
rection when an EBF is subjected to the high impulses generated
from near-collapse conditions (Price 2015). The development of the
25 175 near-collapse loading protocol developed by Price (2015) resem-
Total Link Rotation Angle (rad)

20 bles the AISC loading protocol up to 0.75% total link rotation an-
125
15 gle. Once this rotation angle is achieved, the specimen undergoes
Displacement (mm)

10 75 large cyclic rotation angles in one direction to simulate the ratch-


5 eting effect experienced by an EBF in a near-collapse condition.
25
0 Notably, prior shear link testing used a loading protocol with ran-
-5 -25 domly selected large displacements in each direction during the
-10 -75 early cycles, the shear link specimen achieved a 12.5% IRA before
Near-Collapse
-15 failure, which is higher than that attained under the AISC loading
AISC -125
-20 protocol (Okazaki and Engelhardt 2007).
-25 -175
0 5 10 15 20 25 30 35 40 45
Cycle Number Specimen Details
One POC shear link specimen and four large-scale specimens
Fig. 5. AISC loading protocol and near-collapse loading protocol.
were tested to investigate the feasibility of using horizontal

© ASCE 04023114-4 J. Struct. Eng.


stiffeners in shear links. Table 2 shows the details of the five link specimen (POC-AISC) utilizing the horizontal stiffener
specimens; Table 3 shows their material properties from coupon detailing was constructed with a W150 × 29.8 section made of
tests carried out according to ASTM E8 (ASTM 2016). The first ASTM A992 steel and tested under the AISC loading protocol.
part of the specimen’s name (HSD or CSD) represents the stiffener The other four specimens were constructed with W460 × 60 sec-
configuration utilized. Thus, HSD corresponds to horizontal stiff- tions made of ASTM A992 steel from the same heat to reduce
ener detailing, and CSD corresponds to conventional stiffener variation in material properties and ensure a more direct compari-
detailing. The second part (AISC or NC) represents the loading son between specimens. The length of W460 × 60 specimens was
protocol used: AISC represents the AISC loading protocol, and similar to that of the shear link specimens tested in prior research
NC represents the near-collapse loading protocol. A POC shear (Okazaki and Engelhardt 2007). During the test, the loading se-
quence was based on the total rotation angle (TRA) of the speci-
mens. Full-field 3D strain and displacement measurements were
taken using digital image correlation (DIC) equipment for all spec-
Table 1. AISC loading protocol and near-collapse loading protocol imens. This noncontact system has a measuring strain accuracy of
Loading protocol No. of cycles Total link rotation angle 0.01% (mm=mm).
AISC 6 0.375% Fig. 6 shows the details of the test specimens. POC-AISC
6 0.500% was fabricated using a W150 × 29.8 section with a 10 mm thick
6 0.750% horizontal stiffener located at midheight on one side of the speci-
6 1.0% men. The length of the specimen, 559 mm, corresponds to an
4 1.5% e=ðMp =V p Þ ratio of 1.22 when V p and M p were calculated from
4 2.0% the measured yield strength, well below the 1.6 upper bound for
2 3.0% shear yielding links. HSD-AISC and HSD-NC were fabricated us-
1 4.0% ing W460 × 60 sections with a 13 mm thick longitudinal stiffener
1 5.0%
located at midheight, on one side of the specimen. The thickness of
1 7.0%
1 9.0% the stiffeners for these specimens was determined to be slightly
1 11.0% thicker than the minimum thickness (10 mm) required by the AISC
1 13.0% 341 (AISC 2016a), designed to resemble the flange thickness of the
Near-collapse 6 0.375% (−0.375%)a W460 × 60 section. CSD-AISC and CSD-NC were designed to
6 0.750% (−0.750%) compare the behavior of the horizontal stiffener configuration to
1 0.750% (−4.00%) the conventional stiffener detailing. They were fabricated using
1 4.0% (2.00%) W460 × 60 sections with four 10 mm thick vertical stiffeners
1 6.0% (3.00%)
spaced at 140 mm on center. The 700 mm length of HSD and
1 8.0% (4.00%)
1 10.0% (5.00%)
CSD specimens corresponds to an e=ðMp =V p Þ ratio of 1.10 when
1 12.0% (6.00%) V p and M p are calculated from measured yield stress.
1 14.0% (7.00%) As shown in Fig. 6, extended end-plate moment connections
1 16.0% (8.00%) were used at the ends of the shear link to facilitate the removal
1 18.0% (9.00%) of the tested specimens and avoid damage to the column and
1 20.0% (10.0%) L-arm of the test setup. The design of the 25 mm thick extended
1 22.0% (11.0%) end-plate (ASTM A572 steel) and the 19 mm diameter fasteners
a
Value in parentheses denotes the link rotation angle for the opposite was based on AISC’s Design Guide 4 (AISC 2015). A prequalified
excursion. complete-joint-penetration (CJP) groove weld of TC-U4b-GF

Table 2. Test specimens


Link length Intermediate stiffeners
Stiffener type Specimen Section Loading protocol e (mm) e=ðM p =V p Þ Thickness/quantity
Horizontal POC-AISC W150 × 29.8 AISC 559 1.22 10 mm=1 at midheight
HSD-AISC W460 × 60 AISC 700 1.10 13 mm=1 at midheight
HSD-NC W460 × 60 NC 700 1.10 13 mm=1 at midheight
Conventional CSD-AISC W460 × 60 AISC 700 1.10 10 mm=4 at 140 mm
CSD-NC W460 × 60 NC 700 1.10 10 mm=4 at 140 mm

Table 3. Material properties from coupon tests


Flange Web
Specimen Section Loading protocol Fy (MPa) Fu (MPa) Fy (MPa) Fu (MPa)
POC-AISC W150 × 29.8 AISC 354 463 360 424
HSD-AISC W460 × 60 AISC 399 532 403 462
HSD-NC W460 × 60 NC 399 532 403 462
CSD-AISC W460 × 60 AISC 399 532 403 462
CSD-NC W460 × 60 NC 399 532 403 462

© ASCE 04023114-5 J. Struct. Eng.


700

50

50

Section 2-2 CSD Link HSD Link Section 1-1

CJP welding and access hole POC Link Section 3-3

Fig. 6. Details of test specimens (unit: mm).

(AISC 2016b) was used to connect the link flange to the extended Test Results
end-plates as per AISC Design Guide 4. Notably, Part 12 of AISC
360 (AISC 2016b) states that, when CJP groove welds are used, Table 4 summarizes the key results obtained from each specimen.
weld access holes should not be used. However, to facilitate the Specimen CSD-AISC was not able to be failed due to its high over-
fabrication, access holes were allowed on the link specimens ac- strength, which led to strength beyond the tensile capacity of the
cording to the fabricator’s usual practice. These access holes actuator. The overstrength factor (V u =V p ) is calculated as the ratio
had curved transition edges from the web to the flange and a radius between the maximum measured shear force, V u , and the plastic
prescribed in Section J1.6 of AISC 360 (AISC 2016b). However, shear strength of the link V p ¼ 0.6Fy ðd − 2tf Þtw (AISC 2016a)
the length and height of these access holes were 32 and 16 mm, based on the measured yield stress (Fy ¼ 403 MPa) of wide-flange
respectively, which were slightly smaller than those required by section’s web. The overall depth d of the link was 455 mm
AISC 360 (length = 38 mm; height = 19 mm). Testing results in- for W460 × 60 and 158 mm for W150 × 29.8. The thickness of
dicated that the access holes had no adverse effect on the perfor- the flange tf was 13.3 mm for W460 × 60 and 9.3 mm for
mance of the shear links tested in this study. Before the CJP groove W150 × 29.8. The thickness of the web tw was 8 mm for W460 ×
welds were placed, 8 mm backing fillet welds were first placed on 60 and 6.6 mm for W150 × 29.8. The IRA, γ p , of the specimens
the beam web side of the beam flanges. The root of the bevel was was determined as follows:
then backgouged to remove any contaminants from the 8 mm
backer filler welds. Backgouging of the root was not applied in γ p ¼ γ t −V=k0 ð1Þ
the area of the flange directly above the beam web because the
backing fillet weld was not present. Fillet welds were also used where γ t = the link’s total rotation angle (TRA); V = the applied
to weld the link web to the end-plate. The gas metal arc welding load; and k0 = the link’s initial elastic stiffness measured from the
process and E49 electrode were used for the CJP groove and fillet shear force versus TRA response. Table 4 summarizes the attainedP
welds. No backing bar or weld tabs were used in the welding pro- IRA, γ p ; the ultimate IRA, γ pu ; and the cumulative IRA, γp;
cess. All welds were not ultrasonically tested. Notably, all welds maximum measured shear force, V u ; the overstrength factor,
were not demand-critical because the end-plate was designed ac- V u =V p ; the failure mechanism; and the cumulative dissipated en-
cording to AISC’s Design Guide 4 (AISC 2015) rather than AISC ergy of each specimen.
358 (AISC 2020) for moment connections in special moment Following Sections F3.6e.1(b) and K2.8 of AISC 341 (AISC
frames. Nonetheless, no premature weld failure was observed in 2016a), the attained IRA in Table 4 was defined based on the last
the test specimens. All specimens were whitewashed on one side complete loading cycle in which the link shear force exceeded the
of the specimen, and the DIC system was set up on the other side of plastic shear strength of the link, V p , which was computed based on
the specimen to observe the yielding patterns and strain distribution the measured yield stress of the web where Fy actual ¼ 403 MPa
throughout the test. (AISC 2016a). The ultimate IRA, γ pu , is the maximum IRA that

© ASCE 04023114-6 J. Struct. Eng.


Table 4. Summary of test results
Link shear (kN)
Cumulative dissipated
Specimen γ p (%) γ pu (%) Σγ p (%) Vue V u =V p f Failure mechanism energy (kN-m)
CSD-AISCa 8.1 10.2 154 1,242 1.52 Unable to fail but necking in the k-area was observed 1,115d
CSD-NCa 21.2 21.2 121 1,218 1.49 Flange and web fracture near end plate connection 836
POC-AISCb 9.5 11.8 160 433 2.21 Flange and web fracture near end-plate connection 275
HSD-AISCc 10.4 12.5 247 1,005 1.23 Web fracture after severe WLB 873d
HSD-NCc 21.3 24.4 127 1,102 1.35 Web fracture after severe WLB 924
a
W460 × 60 link with conventional vertical stiffeners.
b
W150 × 29.8 link with proposed horizontal stiffener.
c
W460 × 60 link with proposed horizontal stiffener.
d
Up to the completion of positive excursion of 11% total rotation angle (approximately 10% inelastic rotation angle).
e
V u is the average of the maximum shear forces in both directions (positive and negative).
f
V p ¼ 0.6 Fy actual , ðd − 2tf Þtw based on actual yield stress; V p ¼ 817 kN for W460 × 60 link and V p ¼ 196 kN for W150 × 29.8 link.

the specimens could achieve without the shear strength dropping CSD-NC
below the nominal shear strength. Similar to CSD-AISC, the first yielding of CSD-NC was observed
at the first peak of 0.75% TRA at a load of 730 kN. At this point,
the majority of the specimen web had yielded. A concentration of
Shear Forces versus Inelastic Link Rotation Angle, shear strain was observed near the end of the vertical stiffener
Failure Mechanism, and Overstrength welds. The majority of the web area in the panels had a shear strain
Fig. 7 shows the hysteretic responses of the shear force versus the of less than 3,500 microstrain, but the shear strain around the end of
IRA relationship of the tested specimens. Specimens POC-AISC, the vertical stiffener welds was in excess of 9,000 microstrain.
CSD-AISC, HSD-AISC, CSD-NC, and HSD-NC were able to Figs. 9(a and b) show the shear strain and the out-of-plane displace-
achieve IRAs of 9.5%, 8.1%, 10.4%, 21.2%, and 21.3%, respec- ment distributions for CSD-NC at the positive peak of 18% TRA.
tively. None of the tested specimens showed strength degradation According to the DIC results shown in Fig. 9(a), the highest ob-
below the plastic shear strength (V p ) until after achieving 8% IRA served shear strain was in excess of 0.149 mm=mm in the web
for the AISC loading protocol and 19% IRA for the NC loading steel adjacent to the flange connected to the end-plate interface.
protocol. The strain along the vertical stiffener welds was approximately
0.105 mm=mm. From Fig. 9(b), it is evident that the web buckling
CSD-AISC had increased slightly. The two panels closest to the left end-plate
The first yielding of CSD-AISC was observed at the first peak of experienced the largest magnitude of web buckling of 3.6 mm.
the 0.75% TRA by the DIC system, as shown in Fig. 8(a), at a load Buckling in the other three panels was much less at approximately
of 820 kN. At this point, the majority of the web had yielded. A 1.9 mm. Strength degradation started at the positive peak of 22%
strain concentration was located along the vertical stiffener welds in TRA after the flange of the specimen began to crack, although the
the outside panels and the middle panel. The vast majority of the fracture did not penetrate all the way through the flange. The high-
web had a shear strain of less than 3,750 microstrain but the shear est DIC-observed shear strain was approximately 0.222 mm=mm
strain in these areas of concentrated strain was in excess of 7,800 in the web steel adjacent to the flange to the end-plate interface. It is
microstrain. Necking near the end of welds was noticed at 7% important to note that the flange fracture occurred before the onset
TRA, and a large strain concentration was located in the k-area of any significant web buckling. Fig. 9(c) shows the fracture propa-
adjacent to the end-plate connection. Fig. 8(b) shows the DIC re- gation at the end of the test (prior to obtaining 24% TRA). At this
sults and the necking at the end of the stiffener welds at the first point, the test was stopped due to significant strength degradation.
peak of 11% TRA. The shear strain in the region of the web ad- The fracture severed the flanges adjacent to the end-plate and
jacent to the end-plate connection of the compression flange was in caused the web to fracture. The fracture was more serious at the
excess of 0.111 mm=mm. The increased strain in this region was “column” end [Fig. 4(a)] than at the “beam” end. CSD-NC exhib-
accompanied by an increase in necking at the end of the stiffener ited a stable response up to 21.2% IRA, which is significantly be-
welds. At this point in the test, no evidence was found of fracture or yond the 16% IRA suggested by the authors for the MCE level
buckling at any location on the specimen. The test was concluded ground motions (Price 2015).
after the first peak of the 11% TRA due to excessive overstrength of
the shear link causing the capacity to exceed the strength limita- POC-AISC
tions of the tensile capacity of the actuator. By the time the testing Fig. 10(a) shows the web and flange yielding at 12.7% TRA of
was stopped, the specimen exhibited a stable response up to 10.2% specimen POC-AISC. Cracking at the intersection of the flange
ultimate IRA and 8.1% IRA, beyond the 8% IRA mandated by and the end-plate at the column end was first noticed at 9%
AISC 341 (AISC 2016a). The DIC results indicated the presence TRA [Fig. 10(b)]. This crack did not cause strength degradation.
of a significant strain concentration near the end of the stiffener At a TRA of 11%, the load-carrying capacity did not drop, but the
welds, but the specimen was not allowed to continue until failure. crack propagated further into the flange; meanwhile, the buckling
The excessive strain demands accompanied by a significant amount of the flange at the beam end was noticed adjacent to the endplate
of necking observed in this area was a good indicator that the speci- [Fig. 10(c)]. The specimen failed due to rapid propagation of the
men would have failed due to fracture at the end of the stiffener crack through the entirety of the flange and into the web of the
welds if allowed to continue until failure. However, due to an ob- specimen while attempting to reach 13% TRA. POC-AISC attained
served overstrength of 1.52, the specimen could not be tested until a 9.5% IRA, which is greater than the 8% IRA required by AISC
failure. 341 (AISC 2016a).

© ASCE 04023114-7 J. Struct. Eng.


1500 1500
Vu= 1280 kN γp= 8.1% Vu= 1218 kN γp= 21.2% V+u
+
1000 V u 1000

Shear Force (kN)

Shear Force (kN)


Vp Vp
500 500

0 Stop due to
0
actuator
capacity
-500 -500
-Vp -Vp
-1000
V ˗
u
-1000 V˗ u
-1500 -1500
-12 -8 -4 0 4 8 12 -4 0 4 8 12 16 20 24
(a) Inelastic Link Rotation Angle (%) (b) Inelastic Link Rotation Angle (%)

600
Vu= 433 kN γp= 9.5%
400 V+u

Shear Force (kN)


200
Vp
0

-200 -Vp
-400
V˗ u
-600
-12 -8 -4 0 4 8 12
(c) Inelastic Link Rotation Angle (%)

1500 1500
Vu= 1005 kN γp= 12.5% V+u Vu= 1102 kN γp= 21.3%
V+u
1000 1000 Vp
Shear Force (kN)

Shear Force (kN)


Vp
500 500

0 0

-500 -500
-Vp -Vp
-1000 -1000
V˗ u V˗ u
-1500 -1500
-12 -8 -4 0 4 8 12 -4 0 4 8 12 16 20 24
(d) Inelastic Link Rotation Angle (%) (e) Inelastic Link Rotation Angle (%)

Fig. 7. Cyclic responses of the specimens: (a) CSD-AISC; (b) CSD-NC; (c) POC-AISC; (d) HSD-AISC; and (e) HSD-NC. V u ¼ ðV þ −
u þ V u Þ=2, and
γ p is the maximum inelastic rotation angle corresponding to the full cycle reached by the shear link before the shear strength dropped to less than the
plastic shear strength, V p , which was computed based on actual yield stress.

Necking near
end of weld

Stiffener

(a) (b)

Fig. 8. CSD-AISC: (a) shear strain distribution from DIC measurement at first yielding (0.75% TRA); and (b) shear strain distribution from DIC
measurement and necking at the end of the stiffener weld at the first peak of the 11% TRA.

Specimen POC-AISC had a high overstrength of 2.21, which is area-to-web area (bf tf =ðd − 2tf Þtw ) of the POC-AISC speci-
attributed to the following factors: men is higher than the other specimens.
1. POC-AISC had a shallow section with a depth of 157 mm; 3. As the web was stiffened with a horizontal stiffener, the link
therefore, the link web did not buckle even though the web sus- was able to resist greater shear force up to approximately twice
tained large plastic strains [Fig. 10(a)]. the plastic shear strength. Nonlinear finite element analysis
2. The plastic shear strength (V p ¼ 0.6Fy ðd − 2tf Þtw ) does not (details of the FE models are discussed later in this paper)
consider the contribution of flanges. However, the ratio of flange shows the von Mises stresses at the web is approximately

© ASCE 04023114-8 J. Struct. Eng.


Beam end

Through fracture in
flange and web

Column end

(a) (b) (c)

Fig. 9. CSD-NC: (a) shear strain distribution from DIC measurement at 18% TRA; (b) out-of-plane displacement (mm) distribution from DIC
measurement at 18% TRA; and (c) fracture at the end of the test (immediately before 24% TRA).

Location of
fracture initiation

(a) (b) (c)


Experimental FEM without axial restraint Experimental FEM with axial restraint
600 600

400 400
Shear force (kN)

Shear force (kN)

200 200

0 0

-200 -200

-400 -400

-600 -600
-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
Inelastic Link Rotation Angle (%) Inelastic Link Rotation Angle (%)
(d) (e)

Fig. 10. POC-AISC: (a) web and flange yielding at 12.7% total rotation angle; (b) flange fracture at the “column end,” plate connection (POC-AISC);
(c) flange buckling at the “beam end”; (d) comparison between results from experiment test and FE model without axial restraint; and (e) comparison
between results from experiment test and FE model with axial restraint.

690 MPa, which is twice the nominal yield strength of the steel HSD-AISC
(345 MPa). Therefore, the flexural stress demand in the link The first yielding of the specimen was observed at the first peak
flanges of the link-to-column connection was significantly in- of the 0.75% TRA, as shown in Fig. 11(a), at a load of 725 kN.
creased, which led to the flexural cracks near the column flange The first yielding of the specimen was concentrated directly above
[Fig. 10(b)]. the stiffener and extended from the beam-end end-plate to approx-
4. As shown in Fig. 4(b), the test setup for POC-AISC imposed imately 115 mm from the centerline of the specimen. This yielding
high axial restraint, which in turn led to additional strain- appears to be from a uniformly distributed diagonal tension field
hardening and shear strength. Figs. 10(d and e) compare the load along the weld lines of the stiffener. The maximum shear strain
versus link rotation results between the experimental test and FE in the specimen was 3,470 microstrain, located at the boundary
model without and with axial restraint. It shows that the addi- between the beam-end end-plate and the horizontal stiffener.
tional strain-hardening and shear strength predicted by the FE Fig. 11(a) shows that the maximum principal plane was oriented
analysis agree well with the experimental result when an axial at approximately 45 degrees throughout the entire length of the
restraint is included in the FE model. specimen. The short lines indicate the direction of the maximum

© ASCE 04023114-9 J. Struct. Eng.


(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

Fig. 11. HSD-AISC: (a) shear strain distribution from DIC measurement at first yielding (0.75% TRA); (b) shear strain distribution from DIC
measurement at 7% TRA; (c) out-of-plane displacement (mm) distribution from DIC measurement at 7% TRA; (d) HSD-AISC at the end of
the 7% TRA; (e) out-of-plane displacement (mm) distribution from DIC measurement at 9% TRA; (f) HSD-AISC at the end of the 9% TRA;
(g) HSD-AISC at the end of the 13% TRA; (h) HSD-AISC at the end of testing; and (i) photograph showing no damage at link flanges.

principal strain. Fig. 11(b) shows the shear strain distribution; Figs. 11(e and f) show the DIC results and a photograph of the
Fig. 11(c) shows the out-of-plane displacement of HSD-AISC at buckled web of HSD-AISC at the end of 9% TRA at a load of
the first peak of 7% TRA at a load of 990 kN. The highest observed 845 kN. The load-carrying capacity diminished slightly at this point
shear strain at this point in the test was in excess of 0.050 mm=mm. in the test but is still 103% of V p . Web buckling was able to be seen
The distribution of shear strain differs from the earlier cycles, by the naked eye. The comparison between the DIC system and a
which were similar in pattern to that shown in Fig. 11(a). The shear traditional camera’s photograph of the specimen’s web shows the
strain distribution at the 7% TRA had begun to become distorted accuracy of the DIC system. The maximum positive and negative
due to the onset of web buckling. The orientation of the maximum out-of-plane displacements were in excess of 26 and 27 mm, re-
principal strain was distorted around the location of the buckling spectively. The buckling was more severe at the end of 9% TRA
bands. It is evident from Fig. 11(c) that several buckling bands had than at the first peak. The buckling mode was consistent with the
begun to form at this rotational angle. The maximum displacements typical buckling mode of plates with a length-to-depth ratio of 2
of the two largest buckling bands were approximately 4.2 and (Bulson 1970). Two distinct buckling bands developed on each side
5.0 mm and located in the top panel on either side of the specimen. of the horizontal stiffener. The development of a buckling mode
It is important to note that, although web buckling was noticed by consistent with two adjacent long plates indicates that the stiffener
the DIC system, this degree of web buckling could not be seen by thickness was sufficient, as it successfully provided adequate re-
the naked eye [Fig. 11(d)]. straint to prevent buckling in the midheight area of the specimen.

© ASCE 04023114-10 J. Struct. Eng.


If the stiffener thickness was not adequate, the web would buckle displacement of the buckling bands on the top panel was higher
in a manner analogous to specimens previously tested without than those on the bottom panel. It is important to note that, despite
stiffeners (Hjelmstad and Popov 1983). Fig. 11(g) displays a photo- the severe buckling, which had occurred in the specimen web, the
graph of the HSD-AISC web buckling at the end of 13% TRA with horizontal stiffener maintained the separation of the top and bottom
a load of 810 kN, which is only 1% smaller than V p . The buckling panels of the web into two distinct panels with an out-of-plane dis-
of the web continued to increase in severity, which by now was placement of around zero along the web’s midheight. This shows
obviously causing the overall height of the specimen to decrease. that the dimensions used for stiffener width and thickness were ap-
The severe buckling had begun to cause the horizontal stiffener to propriate for this specimen.
warp. On the way to the first peak of the 15% TRA, the load- The link web started to fracture along the weld lines of the hori-
carrying capacity of the specimen dropped due to the development zontal stiffener at the intermediate 10% TRA (unloading excursion
of a significant fracture in the specimen web as shown in Fig. 11(h). after 20% TRA), as shown in Fig. 12(f). Fig. 12(g) shows the DIC
Fig. 11(i) indicates that, unlike shear links with conventional results for the direction and magnitude of the maximum principal
vertical stiffeners where failure is due to link flange fracture strain. The figure shows that the maximum principal strain at the
(Okazaki et al. 2009), no link flange damage occurs at link-to- fracture location (along the horizontal stiffener) was oriented
column connections. perpendicular to the rest of the web. When the web of each of the
The specimen exhibited stable results up to 12.5% IRA, signifi- panels experienced a large magnitude of buckling, it induced large
cantly beyond the 8% IRA mandated by AISC. At this point, the tensile strains in the web steel adjacent to the horizontal stiffener.
loading dropped substantially due to the initiation of a low-cycle Over several cycles, this could have caused fractures to initiate in
fatigue fracture in the specimen’s web. The load increased at each this location. At 22% TRA, the web buckling was concentrated into
step until approximately 8% IRA was reached. After this point, the two bands in the top panel and one band in the bottom panel. The
load-carrying capacity decreased slightly with each subsequent bands had a maximum out-of-plane displacement of approximately
step but remained above V p until after 12.5% IRA was obtained. 38 mm. The measured load was 1,060 kN. Although the specimen
The hysteretic behavior of this specimen indicates that the hori- load dropped 105 kN, the sustained load was 30% above V p . The
zontal stiffener detailing provides a ductile alternative for shear fracture along the horizontal stiffener had propagated, extending
links. The ultimate shear force developed in the specimen was approximately 110 mm in length. Fig. 12(h) shows that, at 24%
1,005 kN, which corresponds to an overstrength factor of 1.23, TRA, due to excessive local buckling of the web (greater than
which is 20% less than that of CSD-AISC and is below the 1.50 51 mm), the load-carrying capacity of the specimen dropped to
overstrength factor used for the EBF design specified in AISC 900 kN or 75% of the ultimate load and was still 10% above
(2016a). V p . The fracture along the web–stiffener interface had extended
to 230 mm long and had begun to propagate into the specimen’s
HSD-NC web, as shown in Fig. 12(i). HSD-NC exhibited stable behavior up
The first yielding of the specimen was observed at the first peak of to 21.3% IRA, significantly beyond the 16% IRA recommended by
the 0.75% TRA as shown in Fig. 12(a), at a load of 740 kN. The the authors for a suitable collapse level performance (Price 2015).
first yielding of the specimen was concentrated in several bands, The load increased at each step until approximately 19% IRA
the most notable of which spanned from the stiffener–end-plate in- was reached. After this point, the load decreased slightly with each
terface on the left side of the specimen to the midheight area of the subsequent step but remained above V p , until the load-carrying
bottom panel on the right side of the specimen. The specimen’s capacity diminished after 21.3% IRA was obtained. The hysteretic
maximum shear strain was 6,940 microstrain located along the behavior of this specimen showed that the horizontal stiffener
crest of the aforementioned most notable band. Fig. 12(b) shows detailing provides adequate performance for the expected demands
the shear strain distribution; Fig. 12(c) shows the out-of-plane dis- imposed under an MCE level ground motion. HSD-NC had an
placement of HSD-NC at 12% TRA under a load of 1,125 kN. Each overstrength factor of 1.35, which is 10% less than that of
of the web panels experienced significant yielding, which made the CSD-NC and is below the 1.50 overstrength factor used for the
separate bands no longer distinguishable in the panels. The strain design of EBF as specified in AISC (2016a).
demands in the specimen’s web typically exceeded 0.07 mm=mm Test results indicate that HSD-AISC and HSD-NC web frac-
with the highest strain being around 0.09 mm=mm. Fig. 12(c) tures were different. In HSD-AISC, the displacement history
clearly shows three main buckling bands, as they began to form was fully reversed, which created alternate tension field action
at this point during the test, seen on both the top and bottom panels and compression buckling in two opposite diagonal directions.
of the web. The maximum displacements of the bands on the top Consequently, the web eventually fractured due to low-cycle
panel were approximately 4.4, 4.5, and 8.9 mm from left to right in fatigue. Conversely, the displacement history of HSD-NC re-
the figure. The maximum displacements of the bands on the bottom mained fully reversed only up to a small TRA of 0.75%. At large
panel were approximately 5.2, 2.9, and 4.5 mm from left to right. rotation cycles, the specimen was pushed toward one direction
Notably, the onset of web buckling occurred much later for this only, thereby creating buckling bands and increasing only in
specimen (at 12% TRA) than for the HSD-AISC specimen at one diagonal direction. Notably, although HSD-NC was unloaded
7% TRA. The shear strain demands and the buckling pattern of to positive 10% TRA (intermediate) after achieving positive 20%
the two panels are almost identical with only minor discrepancies. TRA, some of the compression buckling bands still had a maxi-
Figs. 12(d and e) show the DIC results and a photograph of the mum out-of-plane displacement in excess of 25 mm. This caused a
buckled web of HSD-NC at the 20% TRA under a load of different web fracture location along the weld lines of the horizon-
1,165 kN. This is the ultimate load-carrying capacity of the speci- tal stiffener.
men. The buckling mode shape remained the same, but the buck-
ling itself increased in severity. The maximum out-of-plane Strain Demand of Flanges
displacements of the three buckling bands in the top panel from Strain gauges at the column end were mounted on the link top
left to right are 12.0, 32.3, and 27.9 mm. The maximum out-of- flange 38 mm from the end-plate connection (Fig. 4) and were
plane displacements of the three buckling bands in the bottom panel analyzed to determine the strain at this location for HSD-AISC,
from left to right are 15.0, 14.8, and 30.2 mm. The out-of-plane CSD-AISC, HSD-NC, and CSD-NC [Fig. 13(a)]. CSD-AISC

© ASCE 04023114-11 J. Struct. Eng.


(a) (b) (c)

Fracture along the horizontal


stiffener weld lines

(d) (e) (f)

(g) (h) (i)

Fig. 12. HSD-NC: (a) shear strain distribution from DIC measurement at first yielding (0.75% TRA); (b) shear strain distribution from DIC mea-
surement at 12% TRA; (c) out-of-plane displacement (mm) distribution from DIC measurement at 12% TRA; (d) out-of-plane displacement (mm)
distribution from DIC measurement at 20% TRA; (e) HSD-NC at 20% TRA; (f) fracture of HSD-NC along the horizontal stiffener weld lines at
intermediate 10% TRA; (g) maximum principal strain of HSD-NC at the intermediate 10% TRA; (h) HSD-NC at the 24% TRA; and (i) fracture of
HSD-NC along the horizontal stiffener weld lines at intermediate 24% TRA.

developed a significantly higher strain than HSD-AISC. CSD- implementing horizontal stiffener detailing. In other words, the
AISC and HSD-AISC had a maximum strain of 6.0% and 2.9%, lower overstrength is achieved by horizontal stiffener detailing
respectively. CSD-NC and HSD-NC developed a maximum strain when compared with conventional stiffener detailing.
of 1.3% and 0.75%, respectively. The horizontal stiffener detailing This observation was further verified by FE analyses, as shown
alleviates the demand on the flange at the end plate connection in Fig. 13(b), which provide the average cumulative plastic strains
as evidenced by the cumulative and maximum plastic strains at 40 mm from the FE model’s flange-to-column connection for
experienced by the specimens. The horizontal stiffener detailing HSD-AISC and CSD-AISC. The cumulative strains of both spec-
alleviated the strain demand on the flanges at the end plate connec- imens are identical until the onset of web buckling in HSD-AISC
tion as evidenced by the cumulative and maximum plastic strains started at 7% TRA. The rate of increase of CSD-AISC becomes
experienced by the specimens. This could indicate failure mecha- much higher beyond a cumulative strain of 0.2. The cumulative
nisms of shear links where specimens under the same loading plastic strains are 0.67 and 0.43 for the CSD-AISC and HSD-AISC,
protocol (with larger maximum and cumulative strain on the respectively, at 11% TRA. This analysis indicates that the lower
flanges) failed by fracture of the flanges when conventional vertical flange strains in HSD links are directly related to the stiffener ar-
stiffeners were used. In addition, prior studies have also shown that rangement as the response of HSD-AISC and CSD-AISC was iden-
the high overstrength in short shear links is partly the result of tical until the buckling in the web of HSD-AISC occurred. This
flange participation in shear resistance. The observation from this web buckling would not happen if conventional vertical stiffeners
research indicates that flange participation can be decreased by were used.

© ASCE 04023114-12 J. Struct. Eng.


6 6 2 6
4 4 4
1
Strain (%)

Strain (%)

Strain (%)

Strain (%)
2 2 2
0 0 0 0
-2 -2 -2
-1
-4 CSD-AISC -4 HSD-AISC CSD-NC -4 HSD-NC
-6 -6 -2 -6
-12 -6 0 6 12 18 -18 -12 -6 0 6 12 18 -6 0 6 12 18 24 -6 0 6 12 18 24
Inelastic Rotation Angle (%) Inelastic Rotation Angle (%) Inelastic Rotation Angle (%) Inelastic Rotation Angle (%)
(a)
HSD-AISC CSD-AISC
0.8
e = 700

Column Side
A
Average Cumulative Plastic

Beam Side
Strain (Section A-A)

TRA = 11%
0.6
Top view of

A 40
specimen CSD-AISC
or HSD-AISC
0.4

Note: Onset of web buckling of HSD-AISC

TRA = 9%
0.2

TRA = 7%
commenced at TRA = 7% and grew at
TRA = 9 and 11%.

0
0 5 10 15 20 25 30 35 40
Cycle Number
(b)

Fig. 13. (a) Strains on the flanges throughout the test; and (b) average cumulative plastic strains at 40 mm from the FE model’s flange-to-column
connection for the HSD-AISC and CSD-AISC specimens.

Finite Element Analysis conducted on Model 1 with shear link geometry identical to
the CSD-AISC specimen. Four FE models with different material
A 3D nonlinear finite element (FE) model was developed to sim-
ulate the tested shear links using the ABAQUS 6.14 program properties were tested, and the hysteresis shear force response was
(ABAQUS 2016). This model was also used to investigate other compared with the experimental results, as shown in Fig. 14(d).
parameters such as the M c =M b ratio and the effect of deeper link The calibration process suggested that Material 4 had a better
sections. Fig. 14 summarizes an example of the calibration process agreement with the test results in terms of the overall response
of FE models, which begins from the test results of steel coupons and hardening behavior (i.e., the Bauschinger effect) subjected
taken from the web of shear links [Fig. 14(a)]. Because the coupon to cyclic loading. Therefore, Material 4 was used for the simula-
test is monotonic testing in nature, the stress-strain data cannot tion of CSD-AISC and HSD-AISC specimens. Fig. 15 shows
accurately simulate the hardening and plastic deformation of shear that the experimental response of HSD-AISC and CSD-AISC
links subjected to cyclic loading. Therefore, this study used the have good agreement with the FEA results. Fig. 16 illustrates
combined isotropic and kinematic cyclic hardening model to that the FE model is able to simulate the web buckling in link
consider the plastic deformation of shear links under cyclic loading. HSD-AISC at large TRA. Fig. 17 shows the von Mises stress
Fig. 14(b) shows the estimated stresses using the combined and plastic strain distributions of the simulated CSD-AISC
isotropic and kinematic cyclic hardening model in the ABAQUS using Model 1 at 11% TRA. The figure illustrates the concentra-
6.14 material library, σ ¼ σ0 þ α þ β, where σ0 ¼ 403 MPa tion of plastic strain near the end of vertical stiffeners, which
is the initial yield stress taken from the coupon test, α ¼ agrees with the observed necking near the end of vertical stiffener
C=γð1−eð−γϵÞ Þ is the kinematic hardening component, and β ¼ weld.
Q∞ ð1 − eð−bϵÞ Þ is the isotropic hardening component. ϵ is the plas- As discussed in this paper’s introduction, unequal moments oc-
tic strain. The kinetic hardening parameters, C and γ, and the iso- cur at the column end (Mc ) and beam end (Mb ) in the shear link of a
tropic hardening parameters, Q∞ and b, are material parameters D-braced EBF. The two moments gradually equalize after the link
obtained by calibration using the cyclic behavior of the experimen- enters an inelastic state. Due to higher rotational restraint at the
tally tested shear links. Fig. 14(b) shows the estimated stresses us- column end than at the beam end, the initial M c can be much larger
ing the isotropic/kinematic hardening model with four material than M b , especially for a short link; hence, the accumulated strain
models, Materials 1 to 4. Each model has various parameters near the column face increases faster than that at the beam end,
(C, γ, Q∞ , and b). The stress-plastic strains of Materials 1 to 4 which tends to cause premature failure of link flange at the column
were used for the materials properties of Model 1, which represents end. The test setup used in this study was designed in such a man-
the 3D nonlinear finite element simulation of a shear link with the ner that the unequal moments were generated to simulate the severe
test setup used in this study. The test setup provides shear link demand at the link-to-column connection. Model 1 simulates the
boundary conditions close to that of a link–column connection experimental test setup used in this study; Model 2 simulates a
in a D-braced EBF [Fig. 14(c)]. The calibration process was one-bay two-story D-braced EBF. The models were analyzed to

© ASCE 04023114-13 J. Struct. Eng.


Experimental results of CSD-AISC
Material 1 Material 3
529
Material 2 Material 4
1500

Stress, MPa 403


1000

Shear Force (kN)


500

-500

0.002 0.1 -1000

Strain, mm/mm -1500


-12 -8 -4 0 4 8 12
(a)
Link Inelastic Rotation Angle (%)
(d)
Material 1 Material 3 Note: The rigid L-arm,
Material 2 Material 4 stabilizing links, and rigid Loading
600 column are restrained in Direction
Combined Stress (σ), MPa

U1 and UR3 to prevent the


500 out-of-plane deformation
U1 = 0
400 U2 = 0
U3 = 0
σ0 = 403 MPa U1 = 0
300 UR1 = 0
U2 = 0 2
b Q∞ γ C UR2 = 0
200 U3 = 0
Material 1 10 18.4 90 834 3 UR3 = 0
Material 2 10 7.95 90 1307 UR1 = 0
100 1
Material 3 3 5.67 75 1540 UR2 = 0
Material 4 0 0 65 1741 UR3 = 0
0
0 0.02 0.04 0.06 0.08 0.1 UR: Rotation
Plastic strain (ε), mm/mm Model 1
U: Displacement
(b) (c)

Fig. 14. Model calibration process for shear link Model 1: (a) tensile stress-strain relation from coupon steel obtained from the link web; (b) materials
1 to 4 are used to calibrate the hardening and plastic parameters using the combined isotropic and kinematic stresses model; (c) use materials 1 to 4 to
analyze Model 1 for CSD-AISC; and (d) link shear force versus IRA response based on four material models and experiment results.

Experimental Test Model 1 Experimental Test Model 1


1500 1500
1000 1000
Shear force (kN)

Shear force (kN)

500 500
0 0
-500 -500
-1000 -1000
-1500 -1500
-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
(a) Inelastic Link Rotation Angle (%) (b) Inelastic Link Rotation Angle (%)

Fig. 15. Experimental results and Model 1 analysis: (a) HSD-AISC; and (b) CSD-AISC.

compare the moment ratio, M c =M b . Fig. 18 shows the M c =M b ra- by the strain measurements from two strain gauges (SG1 and SG2)
tio versus the TRA responses of the link in Model 1. Fig. 18 also attached at the flange near the column end of CSD-AISC and an-
shows the first-story link in Model 2 subjected to monotonic other two strain gauges (SG3 and SG4) mounted at the flange near
displacement-controlled loading. While the M c =M b ratio of Model the beam end. Fig. 18 shows the measured average strain near the
1 and Model 2 are unequal at small link rotations, the Mc =Mb ratio column end, ðSG1 þ SG2Þ=2, versus the average strain near the
approaches unity once the shear link yields at large link rotations beam end, ðSG3 þ SG4Þ=2. Plots A and B illustrate the average
when the moment redistribution occurs from the column end to the strain at the flange near the column end versus the beam end up
beam end. Fig. 18 indicates that the Mc =M b ratio in Model 2 is to 2% and 5% TRA, respectively. The hysteresis curves in Plot
greater but comes closer to that of Model 1 after 0.375% TRA. A show the maximum recorded strain value at the column end
Therefore, Model 1, which simulates the experimental setup of (εc;A ) and beam end (εb;A ) at 2% TRA were 0.0137 and 0.0082,
this study, behaves similar to a link-to-column connection in the respectively, with a strain ratio (εc;A =εb;A ) of 1.67. However, in
D-braced EBF of Model 2. This can also be indirectly demonstrated Plot B, this ratio dropped to 1.02 at 5% TRA.

© ASCE 04023114-14 J. Struct. Eng.


(a) (b)

Fig. 16. FEA results of HSD-AISC using Model 1: (a) comparison of buckling bands in tested specimen with the FE model at 9% TRA; and
(b) increase in web buckling and horizontal stiffener at 11% TRA.

Necking near
end of weld

Stiffener
Necking near
end of weld

(a) (b)

Fig. 17. FEA results of CSD-AISC using Model 1: (a) von Mises stress distribution at 11% TRA; and (b) plastic strain distribution and necking of
tested specimen at the end of the stiffener weld at 11% TRA.

εc,A and εb,A : average strain at column and beam ends at 2% TRA, respectively
εc,B and εb,B : average strain at column and beam ends at 5% TRA, respectively
3
Lateral displacement
Average strain at column

A
end, (SG1+SG2)/2

2.5 εc,A εc,A


εb,A = 1.67
End moment ratio (Mc / Mb)

Lateral
displacement
2
εb,A

Average strain at beam end, (SG3+SG4)/2


1.5
εc,B B
Average strain at column

8.9 m
end, (SG1+SG2)/2

εc,B
1
e = 700 εb,B = 1.02
Column End
Beam End
0.375%

SG3 SG1 εb,B


0.5
SG4 SG2
38 Top view of specimen 38 0.005
A B CSD-AISC Average strain at beam end, (SG3+SG4)/2
0
0 2 4 5 6 8 10 12 14
Total Rotation Angle, TRA (%)

Fig. 18. Mc =M b ratio versus of TRA responses of Model 1 and Model 2 and average strain ratios (for tested CSD-AISC specimen) at the column and
beam ends.

© ASCE 04023114-15 J. Struct. Eng.


27 in. 30 in. 33 in.
obtained from the near-collapse loading protocol, which features
One horizontal stiffener a ratcheting effect, can be used for structural collapse simulations
2 Two horizontal stiffeners according to FEMA P695. Two large-scale, W460 × 60 specimens
utilizing the conventional (vertical) stiffener detailing were tested
under these loading protocols to compare the performance. The
1.5
tests confirmed that horizontal stiffener detailing is a viable alter-
VFEA, 0.08 / Vp

native to conventional stiffener detailing. The following conclu-


1 sions can be drawn from this study:
1. Current AISC 341 (AISC 2016a) requires full-depth inter-
W690x140

W760x161

W840x210
mediate web stiffeners to effectively resist web buckling
0.5 and to limit strength degradation due to flange local buckling
and lateral-torsional buckling. This study shows that the
0
horizontal stiffener configuration developed a ductile failure
690 760 840
mechanism resulting from gradually increasing local buckling
Link depth (mm) in the web. Large deformation of the flange was observed at
large link rotation angles; yet, no torsional buckling was ob-
Fig. 19. Effects of link depth and quantity of one-sided horizontal stif- served. The strength degradation of links with the horizontal
feners on the link shear capacity. stiffener detailing was much more gradual due to the ever-
increasing web buckling, which was the opposite of the sud-
den brittle fractures occurring at the flange-to-column face of
links with conventional vertical stiffener detailing.
Horizontal Stiffeners for Deeper Links 2. Unlike shear links with conventional vertical stiffeners where
As the depth of the link increases, the web-buckling amplitude failure is due to link flange fracture (Okazaki et al. 2009), no
could be too large if only one horizontal stiffener is used. Conse- link flange damage occurred in the W460 × 60 link specimens
quently, an FEA was carried out to investigate if more horizontal with horizontal stiffener detailing. As the failure occurs on the
stiffeners are needed when the link depth becomes deeper. Model 1 web, the effect of the large moment at the link-to-column con-
was used for shear links with depths of 690, 760, and 840 mm. The nection in the shear link of a D-braced EBF does not have a
considered wide-flange sections were W690 × 140, W760 × 161, major impact on the behavior of a shear link with a horizontal
and W840 × 210. The length of each link was selected to maintain stiffener.
a link length ratio expressed as e=ðM p =V p Þ ¼ 1.1, which equals 3. Replacing conventional vertical stiffeners with a horizontal
the used link length ratio for the tested specimens. In each model, stiffener allows web buckling in multiple bands, which softens
the shear force corresponding to the IRA of 0.08 rad (V FEA;0.08 ) was the flanges and web. This in turn reduces the strain demands in
recorded for shear links with one or two horizontal stiffeners. the link flange and the overstrength of the link.
Fig. 19 indicates that links with a depth of 690 mm have a 4. Strain gauge data showed that the horizontal stiffener detailing
V FEA;0.08 =V p ratio of 1.3, and a ratio of 1.46 for links with one decreased the maximum strain demands by nearly 50% at the
horizontal stiffener and two equally spaced horizontal stiffeners, flange to end-plate interface.
respectively, where V p is computed based on the measured yield 5. The overstrength factor of the W460 × 60 link specimens that
stress of the web (= 403 MPa). The shear strength improvement utilized the horizontal stiffener detailing was lower than the
made possible by using two horizontal stiffeners is also observed conventional stiffener detailing (i.e., 1.23 versus 1.52), which
for links with depths of 760 and 840 mm. The analysis indicates the decreased the likelihood of undesired inelastic activity in the
840 mm deep links with one horizontal stiffener had a V FEA;0.08 =V p nonyielding members outside of the link. The overstrength for
ratio below 1.0. Therefore, it is recommended to use two equally the CSD (1.52) was similar to that recognized by AISC (1.50).
spaced horizontal stiffeners for short links having a depth of more Based on a simple ratio of the overstrength of the HSD to
than 760 mm. CSD, it is now possible to reduce the design overstrength from
1.25 to 1.05 for shear links utilizing the horizontal stiffener
detailing (AISC 2016a). This decrease in the overstrength fac-
Summary and Conclusions tor is attributed to a lower shear resistance from the flanges in
the horizontal stiffener detailing than in the vertical stiffener
Current AISC 341 (AISC 2016a) recommends avoiding a D-braced detailing, which has a higher flange restraint and flange shear
EBF configuration, where the link is attached to the column, be- stiffness (Ji et al. 2016).
cause the shear link at a link-to-column connection sustains a much 6. The ultimate rotation angle of the W460 × 60 specimen with
higher moment gradient than a typical moment connection. Shear the conventional stiffener detailing under the AISC loading
links reinforced by conventional vertical stiffeners tend to develop protocol was not reached, but both stiffener details exceeded
high strain demands at the link flanges due to the stiffened link the 8% IRA mandated by AISC, reaching an IRA of at least
flanges. A horizontal stiffener detailing, which can relieve the high 8.1% and 12.5% for the CSD and HSD specimens, respec-
flange strain demand, was investigated in this study. Five shear link tively. Both stiffener details were capable of sustaining large
specimens were fabricated and tested to investigate the perfor- plastic rotations, 21.3% (HSD) and 21.2% (CSD) when sub-
mance of a horizontal stiffener detailing for D-braced EBF shear jected to the near-collapse loading protocol.
links. Tests were performed using a setup that was able to produce 7. DIC results showed that there was essentially no out-of-plane
a high unequal moment at both ends of a shear link, similar to that displacement at the location of the horizontal stiffener after the
at the link-to-column connection in a D-braced EBF. The horizontal onset of buckling indicating that the dimensions of the stiffener
stiffener detailing on a proof-of-concept W150 × 29.8 specimen were sufficient. The analytical study agreed that the stiffener
and two large-scale W460 × 60 specimens under the AISC loading thickness can use the current AISC requirement (the greater
protocol and a near-collapse loading protocol. The response of 0.75tw or 10 mm) with satisfactory performance. According

© ASCE 04023114-16 J. Struct. Eng.


to the design of the experimental specimens and analytical ASTM. 2016. Standard test methods for tension testing of metallic materi-
models, the stiffeners shall extrude from the web an approxi- als. ASTM E8/E8M-16. West Conshohocken, PA: ASTM.
mate distance of (bf =2−tw =2). Bruneau, M., C.-M. Uang, and R. Sabelli. 2011. Ductile design of steel
8. The test setup produced different end restraints. The end mo- structures. 2nd ed. New York: McGraw-Hill.
ments were greater at the column end than at the beam end Bulson, P. S. 1970. The stability of flat plates. London: Chatto and
Windus.
[Fig. 4(a)], which simulates the boundary conditions at a
Chao, S.-H., K. Khandelwal, and S. El-Tawil. 2006. “Ductile web fracture
link-to-column connection in a D-braced EBF. Nonlinear
initiation in steel shear links.” J. Struct. Eng. 132 (8): 1192–1200.
FEA was carried out to show that the unequal end moments https://doi.org/10.1061/(ASCE)0733-9445(2006)132:8(1192).
during the elastic stage were similar to that expected in links Choi, J., B. Stojadinovic, and S. C. Goel. 2003. “Design of free flange
with one end connected to a column. The same analysis shows moment connection.” Eng. J. AISC 40 25–41.
that the end moment gradually equalizes as the link yield pri- Della Corte, G., M. D’Aniello, and R. Landolfo. 2013. “Analytical and
marily in shear. Besides, the experimental data of strain gauges numerical study of plastic overstrength of shear links.” J. Constr. Steel
show that the recorded strain ratio at the link column end to the Res. 82 (Mar): 19–32. https://doi.org/10.1016/j.jcsr.2012.11.013.
beam end is 1.67 at 2% TRA. The ratio gradually decreased Dusicka, P., A. M. Itani, and I. G. Buckle. 2004. “Evaluation of conven-
nearly to unity beyond 5% TRA. tional and specialty steels in shear link hysteretic energy dissipators.” In
9. It is recommended to use one-sided, two equally spaced Proc., 13th World Conf. on Earthquake Engineering. Tokyo:
horizontal stiffeners for shear links having a depth of more International Associations of Earthquake Engineering.
than 760 mm. Engelhardt, M. D., and E. P. Popov. 1989. “On design of eccentrically
10. The proposed horizontal stiffener detailing has the potential to braced frames.” Earthquake Spectra 5 (3): 495–511. https://doi.org/10
.1193/1.1585537.
decrease welding in the shear link (due to the reduced quantity
FEMA (Federal Emergency Management Agency). 2009. Qualification of
of stiffeners and using a one-sided stiffener only) while exhib-
building seismic performance factors. FEMA P695. Washington, DC:
iting adequate seismic performance, maintaining a lower over- FEMA.
strength factor, and utilizing a simplistic design approach. Galvez, P. 2004. “Investigation of factors affecting web fractures in shear
11. Welds used in all specimens were not demand-critical because links.” Master’s thesis, Dept. of Civil Architectural and Environmental
the end-plate was designed according to AISC Design Guide 4 Engineering, Univ. of Texas at Austin.
(AISC 2015) rather than AISC 358 (AISC 2020). Because no Hjelmstad, K. D., and E. P. Popov. 1983. Seismic behavior of active beam
premature weld failure was observed in the test specimens, it links in eccentrically braced frames. Rep. No. UCB/EERC-83/15.
indicates using the welding requirements in AISC Design Berkeley, CA: Earthquake Engineering Research Center, Univ. of
Guide 4 are satisfactory for a shear link end-plate. California, Berkeley.
12. In all the tests, the link was connected to the test setup through Ji, X., Y. Wang, Q. Ma, and T. Okazaki. 2016. “Cyclic behavior of very
bolted extended end-plates. The experimental results indicate short steel shear links.” J. Struct. Eng. 142 (2): 04015114. https://doi
that the proposed details can be used as a replaceable link .org/10.1061/(ASCE)ST.1943-541X.0001375.
when attached to the beam and column through bolted ex- Kasai, K., and E. P. Popov. 1986. “Cyclic web buckling control for shear
link beams.” J. Struct. Eng. 112 (2): 362–382. https://doi.org/10.1061
tended end-plates.
/(ASCE)0733-9445(1986)112:2(362).
Kazemzadeh Azad, S., and C. Topkaya. 2017. “A review of research on
steel eccentrically braced frames.” J. Constr. Steel Res. 128 (Jan):
Data Availability Statement 53–73. https://doi.org/10.1016/j.jcsr.2016.07.032.
Li, F.-X., I. Kanao, J. Li, and K. Morisako. 2009. “Local buckling of RBS
Some or all data, models, or code that support the findings of this beams subjected to cyclic loading.” J. Struct. Eng. 135 (12): 1491–
study are available from the corresponding author upon reasonable 1498. https://doi.org/10.1061/(ASCE)ST.1943-541X.0000073.
request. Liu, X.-G., J.-S. Fan, Y.-F. Liu, M.-Z. Zheng, and J.-G. Nie. 2020. “Theo-
retical research into cyclic web buckling and plastic overstrength of
shear links.” Thin-Walled Struct. 152 (Jul): 106644. https://doi.org/10
Acknowledgments .1016/j.tws.2020.106644.
Malley, J. O., and E. P. Popov. 1984. “Shear links in eccentrically braced
This work was partially supported by the American Institute of frames.” J. Struct. Eng. 110 (9): 2275–2295. https://doi.org/10.1061
Steel Construction Milek Fellowship. /(ASCE)0733-9445(1984)110:9(2275).
McDaniel, C. C., C.-M. Uang, and F. Seible. 2003. “Cyclic testing of
built-up steel shear links for the new bay bridge.” J. Struct. Eng.
129 (6): 801–809. https://doi.org/10.1061/(ASCE)0733-9445(2003)
References
129:6(801).
ABAQUS. 2016. ABAQUS user’s manual V. 6.14. Pawtucket, RI: Dassault Okazaki, T. 2004. “Seismic performance of link-to-column connections in
Systèmes Simulia Corp. steel eccentrically braced frames.” Ph.D. dissertation, Dept. of Civil Ar-
AISC. 2015. Extended end-plate moment connections—Seismic and wind chitectural and Environmental Engineering, Univ. of Texas at Austin.
applications. 2nd ed. Chicago: AISC. Okazaki, T., G. Arce, H. C. Ryu, and M. D. Engelhardt. 2005.
AISC. 2016a. Seismic provisions for structural steel building. ANSI/AISC “Experimental study of local buckling, overstrength, and fracture
341-16. Chicago: AISC. of links in eccentrically braced frames.” J. Struct. Eng. 131 (10):
AISC. 2016b. Specification for structural steel buildings. ANSI/AISC 1526–1535. https://doi.org/10.1061/(ASCE)0733-9445(2005)131:10
360-16. Chicago: AISC. (1526).
AISC. 2020. Prequalified connections for special and intermediate steel Okazaki, T., and M. D. Engelhardt. 2007. “Cyclic loading behavior of EBF
moment frames for seismic applications, including supplements No. 1 links constructed of A992 Steel.” J. Constr. Steel Res. 63 (6): 751–765.
and No. 2. ANSI/AISC 358s2-20. Chicago: AISC. https://doi.org/10.1016/j.jcsr.2006.08.004.
Arce, G. 2002. “Impact of higher strength steels on local buckling and over- Okazaki, T., M. D. Engelhardt, A. Drolias, E. Schell, J.-K. Hong, and C.-M.
strength of links in eccentrically braced frames.” Master’s thesis, Dept. Uang. 2009. “Experimental investigation of link-to-column connections
of Civil Architectural and Environmental Engineering, Univ. of Texas at in eccentrically braced frames.” J. Constr. Steel Res. 65 (7): 1401–1412.
Austin. https://doi.org/10.1016/j.jcsr.2009.02.003.

© ASCE 04023114-17 J. Struct. Eng.


Okazaki, T., M. D. Engelhardt, M. Nakashima, and K. Suita. 2004. Richards, P. W. 2004. “Cyclic stability and capacity design of steel eccentri-
“Experimental study on link-to-column connections in steel eccentri- cally braced frames.” Ph.D. dissertation, Dept. of Structural Engineer-
cally braced frames.” In Proc., 13th World Conf. on Earthquake Engi- ing, Univ. of California, San Diego.
neering. Tokyo: International Associations of Earthquake Engineering. Richards, P. W., and C.-M. Uang. 2003. Development of testing protocol for
Popov, E. P., and M. D. Engelhardt. 1988. “Seismic eccentrically braced short links in eccentrically braced frames. Rep. No. SSRP-2003/08.
frames.” J. Constr. Steel Res. 10: 321–354. https://doi.org/10.1016 San Diego: Dept. of Structural Engineering, Univ. of California,
/0143-974X(88)90034-X. San Diego.
Price, B. 2015. “Investigation on innovative shear link configurations and Yin, W.-H., F.-F. Sun, H.-J. Jin, and D.-Z. Hu. 2020. “Experimental and
optimal design for earthquake resistant steel eccentrically braced analytical study on plastic overstrength of shear links covering the full
frames.” Master’s thesis, Dept. of Civil Engineering, Univ. of Texas range of length ratio.” Eng. Struct. 220 (Oct): 110961. https://doi.org/10
at Arlington. .1016/j.engstruct.2020.110961.

© ASCE 04023114-18 J. Struct. Eng.

You might also like