Journal of Constructional Steel Research 149 (2018) 64–77

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Ductility and energy dissipation behavior of G20Mn5QT cast steel shear

link beams under cyclic loading
Lewei Tong a,c, Yunfeng Zhang b,⁎, Lyu Zhang c, Heng Liu b, Zhiping Zhang c, Ruipeng Li b
a
State Key Laboratory of Disaster Reduction in Civil Engineering, Tongji University, Shanghai 200092, China
b
Department of Civil & Environmental Engineering, University of Maryland, College Park, MD 20742, USA
c
Department of Structural Engineering, Tongji University, Shanghai 200092, China

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

Article history: This paper presents the results of experimental and numerical simulation study of replaceable cast steel link
Received 20 February 2018 beams under cyclic loading, including the ductility, strength, stiffness and energy dissipation of seven shear
Received in revised form 15 June 2018 link specimens with and without circular perforations. Monotonic test of a G20Mn5QT cast steel link specimen
Accepted 6 July 2018
was first carried out. Cyclic load testing of four G20Mn5QT cast steel link specimens are performed to investigate
Available online xxxx
their failure mode, deformation and ultimate load carrying capacity, ductility, stiffness, and energy dissipation
Keywords:
performance. Among these, two link specimens from the same cast steel batch were machined to form two dif-
Link beam ferent web perforation patterns in their web. Cast steel links without web openings show good ductility while
Cast steel perforating the link web offer a tool to reduce the link strength at the cost of ductility. The finite element analysis
Finite element analysis results of the G20Mn5QT cast steel shear links with and without web perforations are also presented and it is
Replaceable link found that finite element simulation results are in good agreement with the experimental results.
Experimental testing © 2018 Elsevier Ltd. All rights reserved.
Hysteresis behavior
Seismic fuse
Eccentrically braced frame

1. Introduction structures and extending its life span [7]. Conventional welded steel
link beams are found to develop cracks in the welded intersection
Growing interests in seismic design have been given to a strategy zone connecting transverse stiffeners to the link beam web plate [13–
that involves the use of fuse-type replaceable energy dissipation devices 15].
with highly ductile performance while other members of the primary The use of steel castings for structural applications has been becom-
structural system are designed to have little or no damage under design ing more popular [2]. Use of cast steel joints is reported to significantly
level earthquakes [5,6]. For example, nonlinear behavior of the eccentri- improve the fatigue life of otherwise highly complex welded connec-
cally braced frames (EBFs) is limited to replaceable link beams. The de- tions. Testing of cast steel connectors that fit between a tubular brace
sign of EBFs with replaceable links has been gaining attention in the past and a gusset plate are reported by [8] for application in seismic load
ten years. Using the replaceable link concept, structural designers have resisting braced frames. Their laboratory test results from static and cy-
more flexibility to choose a proper section size for the steel link beam clic testing of concentrically loaded brace-connector assemblies showed
that best meets the capacity design requirements. The replaceable link that the use of a cast steel connector is a viable means of connecting to
design, which features a bolted extended end plate for ease of removal tubular brace members for seismic applications. Cast steel components
post a major earthquake, has been used in several 2011 Christchurch have also been shown in previous research [1,16,17] to be capable of
earthquake rebuild projects [12]. Link beams in EBFs are capable of de- producing excellent ductility and stable energy dissipation as well as
veloping large plastic deformation while the other structural members the versatility in shape afforded by the casting process. For example,
remain elastic. Using shear links is usually preferred because of its Fleischman and Sumer [1] developed a cast steel modular connector
highly ductile performance and they have been used in structures for use as an energy dissipating element in steel moment-resisting
such as buildings and bridges [9]. Because the plastic deformation will frames. The modular connectors are meant to attach between the col-
be confined within the link beam, damaged links can be quickly umn and the flanges of a beam, and during earthquake the modular
inspected and replaced following a major earthquake, significantly re- connector's ‘arms’, repeatedly deform plastically in double curvature
ducing the disruption time from inspecting and repairing damaged flexure to dissipate seismic energy. This detail removes the need for
field welding in the vicinity of the column panel zone and focuses in-
⁎ Corresponding author at: Tel: (301) 405-1955; Fax: (301) 405-2585. elastic demand to the arms of the cast modular connector. Half-scale
E-mail address: [email protected] (Y. Zhang). prototype modular components that were tested exhibited improved

https://doi.org/10.1016/j.jcsr.2018.07.009
0143-974X/© 2018 Elsevier Ltd. All rights reserved.
 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77 65

Table 1
Test specimen labeling and quantity.

Label Quantity Web Perforation Pattern Loading

Link w/o perforation L0 1 No Perforation Monotonic

L1 2 Cyclic
L2

Perforated Link† L-C-1 1 Cyclic

L-C-2 1

†
Hole diameter is 20 mm for all web holes. Spacing between holes in L-C-1 is 60 mm and additional holes are added to L-C-2 to grid diagonals.

Fig. 1. Dimensions of cast steel link beam specimens without perforation (unit: mm).

ductility and exceeded qualifying rotational capacities for seismic-resis- under cyclic load. For fuse devices, link beams made of hot-rolled sec-
tant, moment-resisting frames. Tong et al. [17] have also conducted ex- tions are commonly used for lower cost and avoiding potential welding
perimental study of steel beam-to-column joints equipped with weld- defects. Yet only limited sections of hot-rolled steel are available. One al-
free cast steel connectors used as replaceable energy-dissipating com- ternative solution to tune the properties of the link beams is to make
ponents. These joints were also shown to meet the classification criteria perforations over the link beams' web area in order to achieve the de-
of semi-rigid joints in terms of stiffness, and the ductility, which was sired mechanical properties. Usually stress concentrations would be in-
governed by fracture of energy dissipating elements, was found to be troduced near web perforations of shear links, which would reduce the
satisfactory.
A total of seven G20Mn5QT cast steel links were tested in this exper-
imental study to investigate their hysteretic behavior and failure mode Table 3
Material properties of G20Mn5QT cast steel.

Table 2 Test Ultimate Strength fu Yield Strength fy Percent Elongation

Chemical composition of G20Mn5QT cast steel for link specimens(%). Coupon (MPa) (MPa) (%)

Web 612.1 463.4 20.3

Material C Si Mn P S Cr Ni Mo
Flange 614.4 461.2 25.7
G20Mn5QT 0.19 0.45 1.38 0.0145 0.0075 0.175 0.035 0.03 Untreated 568 401 32.5
66 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77

Fig. 2. Experimental test setup for cast steel link: (a) schematic illustration; (b) photo.

ductility of the shear links and result in the shear links more susceptible perforations, links circular perforations on web plate. It is also noted
to ductile fracture under cyclic loading. Two types of cast link beam that links with slits are found to have lower ductility due to premature
specimens with the same dimensions were investigated for the ductil- cracking at the corner of slits and therefore their test results are not re-
ity, strength, stiffness and energy dissipation of the shear link specimens ported here. Finite element (FE) simulations are conducted on two cast
and effect of perforation patterns: shear link beams without steel shear yielding link beams; and the results of FE simulations and ex-
periments are compared. The cyclic void growth model (CVGM) is ap-
plied to determine the ductile fracture initiations in the finite element
models of the investigated link beams, which are viewed as the control-
ling limit states of the finite element models. The cyclic load behaviors
of the cast link beams are evaluated in terms of the following parame-
ters: elastic stiffness, nominal shear strength, ultimate shear strength,
ductility, and energy dissipation capacity.

2. Experimental program: test setup and specimens

Experimental tests were carried out on seven cast steel link beam
specimens as labeled in Table 1. A total of eight identical link beam spec-
imens were fabricated with one casting batch of G20Mn5QT cast steel.
However, links with slits are found to have lower ductility and therefore
omitted here. Dimensions of the five test specimens are shown in Fig. 1.
One cast link beam was used for making material coupon testing. Heat
treatment (including quenching and tempering) was also applied to
all cast steel link specimens after casting. The chemical composition of
Fig. 3. Load history for cyclic loading. the material G20Mn5QT cast steel is provided by the casting
 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77 67

Fig. 4. Instrumentation plan: (a) displacement sensor location; (b) strain gage location for link specimens without perforation.

manufacturer, as listed in Table 2. The material properties listed in Table Considering the fact that the mechanical property may vary to
3 were obtained based on coupon tests on three test bars directly cut certain degree due to uncertain factors such as air voids distribution
from the web of one cast steel link beam. in cast steel, two identical link specimens were first tested for the
unperforated link specimens to verify this. The last two specimens
2.1. Specimen description with circular perforations have different perforation patterns based
on the previous test results from the first five link specimens. Speci-
As seen in Table 1, two link beam specimens were perforated in their mens were named according to the following rule: The first letter L
web while the other three specimens (L0 to L2) were not perforated. refers to the Link beam, C refers to the Circular perforation. The num-
Each link beam is made up of four components that are cast into a ber denotes the order of the specimen being test in the two similar
monolithic link specimen: [1] flanges that measure 180 mm specimens. For example, L-C-1 is the first link beam with circular
× 600 mm × 15 mm; [2] end plates measuring 250 mm × 400 mm holes in the web.
× 30 mm; [3] web plate measuring 185 mm × 600 mm × 12 mm; [4] The link beams are usually designed to serve as fuse devices in ec-
traverse stiffeners that measure 185 mm × 84 mm × 14 mm. Fillet tran- centrically braced frames (EBF). In EBF structure systems, short links
sition with various radius (R14 or R20) are made at the intersections of (e b 1.6Mp/Vp, referred to as “shear link” or short link) are generally pre-
these plate components. Provisions on the traverse stiffeners in the Chi- ferred, since they are more ductile than the long links (e N 2.4Mp/Vp, re-
nese code for seismic design of buildings (GB 50011-2010) are adopted. ferred to as “flexural link” or long link) with the same section sizes. The
In this study, the link depth is 215 mm, thus stiffeners are provided only yield strength of cast steel fy is 466 MPa from material coupon test re-
on one side. Since link beam web shear yielding controls the hysteresis ported later, and the plastic moment and plastic shear capacity of the
behavior of the test specimen under cyclic loading, controlling the web link beam without perforation can be calculated as,
thickness and its uniformity over the web area is important to shear link
- Plastic moment:
behavior. The steel casting manufacturer followed the Chinese code for
fabricating the cast steel link beam specimens and so quality assurance
procedure was implemented to ensure geometrical imperfection is Mp ¼ W p f y ¼ 642675  466 ¼ 2:99  108 N  mm
small. The effect of web geometry imperfection on the cast steel link ¼ 2:99  105 kN  mm ð1Þ
beam hysteresis behavior is thus neglected in the analytical and FE
model. - Shear force corresponding to the plastic moment,
68 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77

Fig. 5. Tensile coupon test of G20Mn5QT cast steel: (a) picture: (b) measured stress-strain curve; (c) fractured coupon specimen FD2 from link web and fractured section.

V pm ¼ Mp =ðe=2Þ ¼ 2:99  105 =300 ¼ 997kN ð2Þ performance. The target cyclic drift history (Fig. 3) for the test is from
ANSI/AISC 341-10 [10]. It is comprised of five cycles each to peak drift
ratio of 0.375%, 0.5% and 0.75%. After this the peak drift is increased to
1%, 1.5%, 2% and 3%, each with two to four cycles of corresponding
- Plastic shear capacity, peak drift. Whether it is monotonous loading or cyclic loading, the spec-
imens are loaded to failure. The failure refers to the following two
criteria: [1] specimens have obvious cracking; [2] the load bearing ca-
pacity of specimens drops below 90% of the ultimate bearing capacity
V p ¼ 0:6Aw f y ¼ 0:6  2220  466 ¼ 6:21  105 N ¼ 621kN ð3Þ
observed. Monotonic loading was applied to the Specimen L0 in a cer-
tain direction until the specimen failed. The cyclic loading is variable
Therefore, the (e*Vp)/Mp is 1.25, below the 1.6 upper limit value for amplitude loading, and the loading rate is controlled at 6 mm/min.
the shear link without perforation. Those link specimen with perfora- The control factor of the loading mechanism, that is, the deformation
tions in the web should have an even lower value of (e*Vp)/Mp while of specimen, is the link rotation angle γ, which is defined as the ratio
having their plastic moment value unchanged. of the relative displacement between the two end plates over the length
of the link beam. Before formal loading, the specimen was loaded for
2.2. Test setup preliminary test observation, and the preloading amplitude was set as
30% of the yield deformation.
A schematic view of the experimental test setup are shown in Fig. 2
(a) and 2(b), respectively. The test setup used in this experiment is
composed of L-type loading beam, ground beam, two pinned struts to 2.3. Instrumentation
form a parallel mechanism with the L beam and ground beam, and reac-
tion frame. The reaction frame and ground beam is connected together Instrumentation was installed to record the cyclic load response of
with high strength rods and steel beams. This design ensures that the the link beam specimens. The applied load was measured directly
link beam specimens have the similar boundary conditions as in real ec- from the actuator's load cell. Horizontal displacement transducers
centrically braced frames the end plates of the link beam specimens were placed at the two end plates of the link beams, and ground beam
would not rotate during the loading test. A servo-hydraulic actuator to determine the deformation of link beams and monitor slide if any.
was used to load the test specimen at the same height as the middle Displacement transducers were also installed on the two diagonals of
length point of the link beam specimen. The bottom end plate of the the link web to measure the shear displacement (see Fig. 4a). Numerous
link beam was connected to the ground beam using four high-strength uniaxial strain gauges were installed at a variety of locations along the
bolts and spot weld was later provided to prevent the link from sliding link beams (see Fig. 4b) to determine the axial forces and bending
at large drift.
This research aims to investigate the hysteresis behavior of replace- Table 4
able steel link beams. Firstly, a specimen is subjected to monotonic load- Calibrated plastic strain hardening constants for G20Mn5QT cast steel in finite element
ing, and the deformation behavior of the specimen is observed for simulation.
comparison and prediction of cyclic loading test. The remaining speci- C1 (MPa) γ1 k (MPa) R0 R∞ b
mens are subjected to cyclic loading, and the experimental phenomena
5872.7 31.51 483 0 21 1.2
are observed for comparative analysis of their energy consumption
 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77 69

Fig. 6. Meshed finite element model for two cast steel link beams: (a) boundary constraints; (b) Specimen L1 (or L2); (c) Specimen L-C-2.

moments of each specimen. The strain gauges were bonded parallel to first, hoping to adjust the link beam's load capacity and yield/fracture
the longitudinal axis and on the outer faces of the flanges of the link initiation zone through strategically arranging the slots and circular
beam specimens. Rosette strain gauges were installed on the webs of holes in the link web; for the stiffener, the purpose of strain gage instru-
the link beams, and the corners of the slit and circular perforations to mentation is to determine the stress condition.
measure the local strains and to record the occurrence of yielding. The
layout of displacement transducers and strain gauges for Specimen L1 3. Material properties of G20Mn5QT cast steel
is shown in Fig. 4. For the flange, strain gage data is used to understand
the stress distribution at different locations along the length of the link The steel used in this study was made of G20Mn5QT cast steel,
beam; for the link web, the goal for the strain gage installation is to un- which was manufactured by Jiangsu Langyue Maoda Casting Company
derstand the stress distribution, and where and when yielding occurs in Eastern China. After the production of cast steel specimens, surface
70 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77

1200

800

400

Shear Force (kN)

0

-400

-800

L1
L2
-1200
-0.2 -0.1 0 0.1 0.2
Link Rotation Angle (rad)

(a)

800

600

400
Shear Force (kN)

200

-200

-400

-600
L-C-1
L-C-2
-800
-0.1 -0.05 0 0.05 0.1
Link Rotation Angle (rad)

(b)

Fig. 8. Hysteresis curves of cast steel link specimens under cyclic loading.
Fig. 7. Cast steel link specimen L0 under monotonic loading: (a) pushover curve; (b)
picture of specimen at link rotation angle γ of 0.345 rad. (c) area with small crack.
specimen directly, including three samples of web and two samples of
flange, and all of them have the same size. The size of the material coupon
processing (cleaning, polishing, shot blasting), heat treatment measures 100 mm × 15 mm × 12 mm. In accordance with “metal tensile
(quenching and tempering), nondestructive testing (magnetic particle test part one: Test Method at room temperature” (GB/T 228.1-2010),
testing), machining (end plate surface milling, bolt hole drilling, and strain rate is controlled at 0.0025 mm/mm/s. The CMT5205 microcom-
slot cutting, etc.) were conducted. puter-control electronic universal testing machine was used to perform
In accordance with the provisions of the code for determining the uniaxial tensile tests on material samples and testing results are given
chemical contents of low alloy steel, the chemical composition of the in Table 3. The uniaxial tensile stress-strain curve of a typical material
cast steel is listed in Table 2. The mechanical properties of cast steel can coupon is shown in Fig. 5b. Fractured section of tension coupon test spec-
be controlled by controlling the casting process and heat treatment tech- imen FD2 after test was completed is shown in Fig. 5c. The elongation and
nology. Considering that the heat treatment and shot blasting may affect yield ratio of the cast steel meet the requirements of the code, but its
the properties of cast steel, bar samples were cut from one link beam strength is slightly greater and the ductility is smaller than Q345 steel.
 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77 71

F F F l
σ true ¼ ¼ ¼  ¼ σ nom ð1 þ εnom Þ ð5Þ
A A0 l0 =l A0 l0

The l0 is the initial gauge length and l is the length after loading. A0 is
the original cross-sectional area of the gage segment and A is the de-
formed cross section area. A0⁎l0 is equal to A*l because the material vol-
ume over the gage length is conserved.
Compared with the material properties of the cast steel material
(from the same batch, but without heat treatment and machining),
the yield strength of the processed cast steel link specimen is 18% higher
than the untreated specimen and the ultimate strength of the processed
specimen is 9% larger than the untreated specimen. However, the
treated specimen has a lower elongation value (20% vs. 32%) than the
untreated specimen.

4. Experimental test results

The lateral load vs. displacement curve of Specimen L0 under mono-

tonic loading is shown in Fig. 7a. In the figure, it is seen that the cast link
had a very large deformation capacity up to 0.35 rad (referred to as
chord rotation angle which is defined as lateral displacement divided
by link length) and the test had to be terminated before the specimen
lost its load bearing capacity when the actuator reached its maximum
stroke. Its yield load and corresponding chord rotation angle are 630
kN and 0.009 rad respectively. At link rotation angle of 0.35 rad, no
cracks were observed in the link web and its intersection with flanges
and stiffeners as shown in Fig. 7c. However, minor cracks were seen at
the intersection of tension-side flanges and end plates when the test
was terminated.
The lateral load vs. displacement hysteresis curves of Specimens L1 &
L2 under cyclic loading are shown in Fig. 8a. It is seen that the two cast
link specimens without web perforation have fat hysteresis loop up to
the maximum link rotation angle of 0.17 rad when the test was termi-
nated due to fracture-induced load bearing capacity degradation (see
Fig. 8a). First visible crack in Specimen L1 was reported to occur near
the intersection of flanges under tension and end plates at 0.09 rad
and this crack propagated as load cycles increased. At the link rotation
angle of −0.13 rad, a crack parallel to the flange was found in the link
web near the flange and transverse stiffener. However, the load bearing
capacity of both link specimens did not drop when these cracks
emerged and before they developed into large through cracks. Through
crack was seen at the intersection of end plates and flanges in Specimen
L1 and L2 at link rotation angle of 0.15 rad and 0.17 rad respectively,
when the tests were halted. Thus, Specimens L1 and L2 exhibited similar
behavior, except for small difference in ultimate deformation capacity
(0.15 rad vs. 0.17 rad in terms of ultimate link rotation angle). This is
also reflected in the two hysteresis curves in Fig. 8a that almost overlap
with each other except for ultimate deformation stage.
Table 5 shows the mechanical properties of all cast steel link speci-
Fig. 9. Pictures of cast steel link specimens at ultimate loading stage: Specimen L1, γ = mens under cyclic loading. The yield strength of Specimen L1 is lower
+0.17 rad; (b) γ = −0.07 rad, Specimen L-C-1; (c) γ = +0.11 rad, Specimen L-C-2. than Specimen L2 by about 62 kN. It is seen that the ductility coefficient
of Specimens L1 & L2 is around 14. The ductility coefficient for all link
It is found from experiments that the elastic modulus of G20Mn5QT beam specimens are also shown in Fig. 11a. Ductility measures the plas-
cast steel is 207 GPa. The percent elongation (at fracture) is 20.3 and the tic deformation capacity of structural members before failure. The duc-
overstrength ratio (ultimate stress/yield stress) is 1.32, which meets the tility ratio μ of the cast steel shear link specimens is determined from the
Chinese design code requirements for steel structure construction. The backbone curves, which was calculated as the ratio of the ultimate link
nominal stress and strain curve in Fig. 5b was obtained from uniaxial rotation γu to the yield link rotation γy. The backbone curve is the enve-
(monotonic) tensile test, the other curve in Fig. 5b is the true stress lope of hysteretic curve under cyclic loading, obtained by connecting the
and strain curve. Before the tensile coupon specimen necks, the nominal peak points at each load cycle. Link rotation γ is defined as the ratio of
strain εnom and stress σnom can be transformed into true strain εtrue and relative displacement δ between link end sections and link length
true stress σtrue as below, (equal to 600 mm for all specimens). δu is the ultimate displacement,
corresponding to the load level with 15% drop from the peak load capac-
Zl  
dl l ity Vmax. δy denotes the yield displacement of the link specimens, which
ε true ¼ ¼ ln ¼ ln ð1 þ ε nom Þ ð4Þ is calculated using the Park method [18]. Therefore, this yield point is
l l0
l0 the equivalent yield point, usually with a yield load greater than the
72 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77

Fig. 10. Measured response of Specimen L1: (a) strain at flange; (b) shear deformation; (c) equivalent strain of link web at three panel center locations.

first yield load identified from the strain gage measurement on the link L1 is shown in Fig. 10c. It is seen that the stain in web was small in the
web. elastic stage but increased rapidly after yielding at 550 kN and the plas-
Hysteresis energy dissipated by a link in each loading cycle is equal tic deformation of link web was fairly large after 900 kN.
to the area enclosed by the hysteresis loop of that load cycle. The hyster- Two approaches to reducing the link beam strength and concentrat-
esis loops for all links are shown in Fig. 8. Therefore, energy dissipation ing the plastic deformation zone to locations to near the perforations
capacity of link beam specimens under cyclic loading is equal to the hys- were considered in this study: web slitting or web perforation. The
teresis loop area. The maximum energy dissipation of Specimens L1 & test results of slit link beam specimens are not reported here due to its
L2 (and all other cyclically loaded specimens in this study) is shown in poor ductility. For the perforated cast steel link specimens, circular
Fig. 11b. The energy dissipation capacity of Specimen L2 is greater holes of a diameter of 20 mm are perforated over the link web. Geome-
than that of Specimen L1 by about 19%, which can be explained by the try and perforation patterns of the two perforated cast steel links are
larger displacement of Specimen L2 in the last load cycle (0.15 rad vs. shown in Table 1. The cast steel link specimen L-C-1 has the horizontally
0.17 rad). aligned circular holes spaced 60 mm (Sdiag = 60 mm) apart; while the
The measured strain response near the flange's end section of Speci- cast steel link specimen L-C-2 has diagonally aligned circular holes
men L1 is shown in Fig. 10a. The strain at flange end increases linearly spaced 42.4 mm (Sdiag = 42.4 mm) apart. Although Specimen L-C-2
with load in initial elastic stage but then increases rapidly after flange has more holes than L-C-1, its energy dissipation and deformation ca-
yielded at around 760 kN, which is much greater than its web yield pacity (ductility) is found to be better than L-C-1. It is seen from Table
load of 590 kN. Therefore this link beam specimen is a shear link. This is 5 that the elastic stiffness and yield load capacity of L-C-1 are fairly
also confirmed by the results shown in Fig. 10b, which reveals that the close to that of L-C-2, however, the ductility of these two perforated
total deformation of the link specimen is dominated by its shear deforma- link specimen is different. This indicates that perforation pattern is im-
tion. The shear deformation is calculated using measured displacement portant to the ductile behavior of cast steel link beam with web perfora-
data from two diagonal displacement transducers in the link web. The tions. Diagonal patterns appear to be better than rectangular grid
strain in the stiffeners of Specimens L1 & L2 is quite small in elastic pattern even more holes are created in the link web. This can also be
stage; however, after specimen yields, it increases rapidly with the in- seen in the results summarized in Fig. 11.
crease of cyclic displacement amplitude but stays within the effective When loading Specimen L-C-1, first visible crack was reported at the
range (b3%) of strain gage before the specimen failure point. loading cycle of 0.02 rad, near one of the second row holes (next to
In order to compare the strain in the link web of Specimens L1 and flange) in the middle web panel. During the loading cycle of ±
L2, equivalent strain (von Mises strain) is calculated as, 0.07 rad, neighboring cracks were connected together and the experi-
ment testing was terminated due to significant load capacity drop. Pic-
pffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 ture of Specimen L-C-1 at link rotation angle of 0.07 rad is shown in
ε¼ ðε1 −ε 2 Þ2 þ ðε1 −ε 3 Þ2 þ ðε3 −ε 2 Þ2
3 Fig. 9b. For Specimen L-C-2, crack was first seen at the load cycle of
0.03 rad, which formed near the center hole in the middle web panel.
where ε1, ε2, and ε3 are principal strains calculated using measured During the load cycle of ±0.11 rad, three longitudinal cracks were
strain rosette data from the link web. The equivalent strain of Specimens seen to connect the holes in the middle web panel and the experiment
 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77 73

Fig. 11. Performance comparison of six cast steel link specimens under cyclic loading: (a)
ductility coefficient; (b) Maximum energy dissipation (from last load cycle). Fig. 12. Hysteresis loops of cast steel specimens: (a) Specimen L2; and (b) Specimen L-C-2.

test had to be terminated due to the 16% load capacity drop. Picture of strength, ductility, fracture initiation and strength degradation due to
failed specimen L-C-2 at link rotation angle of 0.11 rad is shown in fracture or local buckling in flange and web. Because L-C-1 exhibits a
Fig. 9c. less ductile behavior than L-C-2, finite element analysis are only con-
ducted for cast steel link beam L-C-2 and the link beam without web
5. Finite element analysis perforation (L1 or L2). The material properties for G20Mn5QT cast
steel used in this study were calibrated from experimental data. Von
A general finite element analysis software - ANSYS Academic 15.0 is Mises yield criterion is assumed for the material, which is associated
employed to perform finite element simulations on the shear links. The with the flow rule characterized by the combined kinematic and isotro-
finite element modeling analysis is conducted to provide insight on the pic hardening. This combined plastic strain hardening rule with the
cyclic loading behaviors of the investigated shear links, including nonlinear isotropic hardening model and the Chaboche model is applied
74 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77

Fig. 13. Specimen L2 at link rotation angle γ= +0.07 rad: (a) Accumulated plastic strain
contour plot from FE analysis; (b) Picture of flaked white wash pattern.

to characterize the stress strain relationship of G20Mn5QT cast steel

after material yielding, as expressed in Eq. (6) below.

C1  
σ¼ ð1−e−γ1 εp Þ þ k þ R∞ 1−e−b^εp ð6Þ
γ1 Fig. 14. Specimen L-C-2 at link rotation angle γ= +0.07 rad: (a) Accumulated plastic
strain contour plot from FE analysis; (b) Picture of flaked white wash pattern.
The calibrated material parameters for the combined hardening
model are listed in Table 4. Ductile fracture initiation due to ultra-low-
cycle fatigue is assumed to be the controlling limit state for the cast are analogous to the boundary constraints of a double curvature beam.
steel shear links, as observed in experimental tests. The ductile fracture The applied boundary constraints to meshed model are shown in Fig. 6a.
initiation point was estimated using the cyclic void growth model Two end plates are added to the two end sections of the cast steel
(CVGM) proposed by Kanvinde and Deierlein [19], as expressed shear link model to simulate the boundary conditions same as the ex-
Eq. (7) below. periments. Each end plate was constrained by eight bolts with a diame-
ter of 33 mm as shown in Fig. 6a. The constraint from bolts is simulated
X Z ε2 X Z ε2
by applying restraints to the nodes around the bolt-hole region in the
expðj1:5T jÞdε p − expðj1:5T jÞdε p
tensile cycles ε1 compressive cycles ε1 endplates. On the left side of the shear link, the selected nodes were
Damage ¼   only restrained in the out-of-web-plane direction and all rotational
VGI critical  exp −λεaccumulated
p DOFs as shown in Fig. 6a, while all DOFs are fully restrained at the
ð7Þ other right-side end plate. The boundary constraints are applied to the
nodes around the bolts area in both end plates of the shear link.
where T is the stress triaxiality, which is defined as the ratio of the hy- In the FEM analysis, residual stress was not modeled for the follow-
drostatic stress to the Von Mises stress; εp is the equivalent plastic ing reasons: these cast steel link beams underwent a heat treatment
strain; and εaccumulated
p is the accumulated plastic strain to the beginning process recommended by the steel casting foundry after casting to re-
of each tensile cycles. Ductile fracture initiation is expected in the shear move the residual stress; the cast steel link beam specimens have a
link when the damage index exceeds one within its critical volume. The short length (600 mm) and simple geometry. Therefore, it is believed
critical void growth index (VGIcitical) and the decay parameter λ used for that residual stress is low in the cast steel link beam and residual stress
the cast steel in this study are assumed to be 1.1 and 0.6 respectively. was not measured in the cast steel link beam specimens. An FEM study
The Shell181 element in the ANSYS software is utilized for finite ele- of the same cast steel link beam specimen L2 but with several different
ment modeling. The meshed model for the cast steel shear link is plotted assumed residual stress patterns as initial stress is also conducted to
in Fig. 6. Large displacement option has been taken into account in FE study its effect on the hysteresis behavior of steel link beam under
analysis to simulate geometric nonlinear effect (e.g., local buckling) in ultra-low-cycle fatigue loading. The FE analysis results show that resid-
the simulation. Mesh refinement studies were conducted to determine ual stress has little influence on the hysteresis behavior and crack initi-
the level of refinement necessary to achieve reasonable accuracy for cyclic ation point of the cast steel link beam.
behavior study of the cast steel shear links and fracture initiation point es- The same loading protocol as the one adopted for the experimental
timation. Local refined mesh is applied to the hot spot area with high frac- tests is applied to the cast steel shear links in FE analysis. The hysteresis
ture index. The boundary constraints applied to the finite element model loops of the cast steel shear link specimens determined from FE
 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77 75

Fig. 15. Specimen L2 with haunch stiffener at link rotation angle γ= +0.07 rad: (a) meshed FE model with haunch stiffener; (b) Accumulated plastic strain contour plot from FE analysis;
(c) hysteresis comparison to the original Specimen L2.

simulation and experiment testing on specimens L2 are plotted in Fig. Accumulated plastic strain contours for the unperforated link L2 at
12. Clearly, the FE analysis results agree well with the experimental link rotation angle of 0.07 rad are presented in Fig. 13a. Accumulated
test results. The mechanical properties of the cast steel specimen includ- plastic strain is defined as a path-dependent summation of plastic strain
ing the elastic stiffness K0 (unit: kN/mm), yield strength Vp, and yielding rate over the loading history. Fig. 13b shows the white wash flaking pat-
link rotation angle γy are determined from the backbone curves of the tern; clearly, the link web has already developed substantial plastic de-
cast steel shear links. In the FE analysis, γy corresponds to first signifi- formation at this point. For the cast steel shear link specimen L2 without
cant yielding in link web. It is seen that while the initial stiffness of the web perforation subjected to cyclic loading, crack was first observed in
two test specimens from the FE results are 212 kN/mm and 174 kN/ the experiment to occur at the connection between flange and end plate
mm for Specimens L2 and L-C-2 respectively. Because Specimen L-C-2 during the loading cycle with link rotation angle γ equal to 0.09 rad.
has a perforated web (otherwise same as Specimen L2), its initial stiff- From the FE simulation result, first crack initiation is estimated to
ness should be lower than L2. occur at the interface of flange and endplates when the link rotation
76 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77

Table 5 this paper. A total of five shear link specimens with and without circular
Properties of cast steel link specimens under cyclic loading from experiments and finite el- web perforations was fabricated and tested in the lab. Heat treatment
ement analysis (number shown in parenthesis).
including quenching and tempering was also applied to all cast steel
Specimen K0 Vp (kN) δy Vmax δu Ductility link specimens after casting. The cast steel link beam is designed as
(kN/mm) (mm) (kN) (mm) Coeff. shear link for the shear link without perforation. Those link specimens
L1 160.4 645.5 6.3 1012.7 90 14.28 with web perforations have an even lower value of (e*Vp)/Mp while
L2 164.3 643.6 6.2 1013 90 14.63 having their plastic moment value unchanged. The failure mode, defor-
(212.3) (686) (4.7)
mation and ultimate load carrying capacity, ductility, stiffness, and en-
L2 w HS⁎ (215.5) (686.5) (4.6) – – –
L-C-1 149.2 502.6 4.8 720.1 42 8.75 ergy dissipation performance are presented. Monotonic tensile test of
L-C-2 145.7 513.6 5.1 744.4 66 12.94 G20Mn5QT cast steel material coupons were conducted to obtain the
(173.9) (498) (4.3) mechanical properties of cast steel. Based on the monotonic tensile
⁎ Cast steel link beam with haunch stiffener (finite element simulation results only). property test of cast steel, it is known that the strength of cast steel is
higher and the ductility is less than Q345 steel. The link rotation angle
is found to be over 0.345 rad from monotonic load test (test was halted
angle equal to 0.07 rad, which matches the crack location and timing ob- at this point because the servo-hydraulic actuator reached its stroke).
served in the experiment. Blue dots in Fig. 12 indicate the first crack ini- The dominating failure mode of a shear link should be web shear
tiation in corresponding cast steel link. For a crack to be visible in the yielding and subsequent ultra-low-cycle fatigue induced fracture. All
experiment, it has to be wide enough; and thus the experimental ob- link beam specimens in this study exhibit this behavior. It is seen that
served crack occurrence time would be later than the first crack initia- the two cast link specimens without perforation have fat hysteresis
tion time corresponding to a much smaller crack size in FE analysis. loop up to the maximum link rotation angle of 0.17 when the test was
FE analysis derived accumulated plastic strain contours for the per- terminated due to fracture-induced load capacity drop. Specimens L1
forated link L-C-2 at link rotation angle of 0.07 rad are presented in and L2 exhibited similar behavior.
Fig. 14a. Hot spot areas with large accumulated plastic strain values It is seen that web perforation pattern has a significant impact on the
occur in stripes connecting the holes, which can be clearly seen in cyclic load behavior of the link specimens with circular perforations in
both figures. This observation support the above experimental observa- the web. Web perforation patterns with diagonal holes (e.g., L-C-2) per-
tions as this contour pattern is consistent with the white wash flaking form in a more ductile manner than those without diagonal holes (e.g.,
pattern at the same link rotation angle as shown in Fig. 13b. For the L-C-1). Even though Specimen L-C-2 has more web holes than Specimen
cast steel shear link specimen L-C-2 with web perforation, cracks were L-C-1, their elastic stiffness and yield load capacity are fairly close to
first observed at the load cycle of 0.03 rad, which formed at the center each other while the ductility of these two perforated link specimen is
hole in the middle web panel. These cracks subsequently merged into very different.
a large crack at the loading cycle of 0.11 rad. This is consistent with Finite element simulations are also conducted on two cast steel link
the FE simulation results, in which the crack initiation was firstly pre- beams with and without web perforations and the results of finite ele-
dicted at the middle of the orthogonal line of the holes in the middle ment analysis and experiments are compared. The cyclic void growth
web panel at link rotation angle equal to 0.02 rad. model (CVGM) is adopted to determine the ductile fracture initiation
For unperforated link specimens L1 and L2 under cyclic loading, crack point in the finite element models of the investigated link beams. The
initiation was found to first occur near the intersecting line of link flange cyclic load behaviors of the cast link beams from finite element analysis
and end plates based on both experimental and FE analysis results and are found to be consistent with the experimental results. It is also found
later the specimen failed because this crack developed into a large crack from finite element analysis that adding curved haunch stiffeners can
that separates the flange from end plates. Since one of the advantages effectively reduce the fracture potential near the end plates in cast
for using cast steel links is its flexibility of casting into free form, curved steel shear links since one of the advantages for using cast steel links
haunch stiffener plates with a radius of 50 mm and thickness of 12 mm is its flexibility of casting into free form without significantly increasing
(shown in Fig. 15a) are added to the unperforated link specimen L2 in the casting cost.
FE model and its cyclic load behavior is simulated with this FE model.
The accumulated plastic strain contour and hysteresis loop from the FE
Acknowledgements
analysis are shown in Fig. 15b and c respectively. By comparing with
that in Fig. 13a, it is clearly seen that the accumulated plastic strain near
The experimental part of the current work was supported partially
the intersection line of flanges and end plates are significantly reduced
by Tongji University (Grant No. 0200144802). However, the opinions
while its hysteresis loop is almost the same as that of unperforated link
and conclusions expressed in this article are solely those of the writers
L2 without haunch stiffener. FE analysis also shows that its first facture
and do not necessarily reflect the views of the sponsors.
initiation occur at much larger link rotation angle compared to cast
steel link L2. This suggests that adding curved haunch stiffeners can effec-
References
tively reduce the fracture potential near the end plates in cast steel shear
links without significantly increasing the casting cost. [1] R. Fleischman, A. Sumer, Optimum arm geometry for ductile modular connectors,
As shown in Table 5, the stiffness of the link beam with haunch stiff- ASCE J. Struct. Eng. 132 (5) (2006) 705–716.
[2] S. Herion, J.-C. de Oliveira, J.A. Packer, C. Christopoulos, M.G. Gray, Castings in tubular
eners is 215.5 kN/mm, which is slightly larger than 212.3 kN/mm for the structures – the state of the art, J. Struct. Build. 163 (6) (2010) 403–415.
link beam without haunch stiffener. Yield strength of the cast steel link [3] W. Richards Paul, Uang Chia-Ming, Effect of flange width-thickness ratio on eccen-
beam with and without haunch stiffeners are 686.5 and 686 respec- trically braced frames link cyclic rotation capacity, ASCE J. Struct. Eng. 131 (10)
(2005) 1546–1552.
tively. This can be explained by the fact that the curved haunch stiffener
[4] G. Corte Della, M. D'Aniello, R. Landolfo, Analytical and numerical study of plastic
with a tapered design pattern measures only 5 cm on each side (only overstrength of shear links, J. Constr. Steel Res. 82 (2013) (2013) 19–32.
8.3% of the overall beam length) and therefore would only slightly in- [5] D. Dubina, A. Stratan, F. Dinu, Dual high-strength steel eccentrically braced
frames with removable links, J. Earthq. Eng. Struct. Dyn. 37 (15) (2008)
crease the overall stiffness and strength.
1703~1720.
[6] N. Mansour, Development of the design of eccentrically braced frames with
6. Conclusions Rreplaceable Sshear Llinks, Doctoral Dissertation, Department of Civil Engineering,
University of Toronto, Toronto, Canada, 2010.
[7] Mansour Nabil, C. Christopoulos, Tremblay Robert, Experimental validation of re-
Both experimental testing and finite element analysis results of five placeable shear links for eccentrically braced steel frames, J. Struct. Eng. 137 (10)
G20Mn5QT cast steel link beams under cyclic loading are presented in (2011) 1141–1152.
 L. Tong et al. / Journal of Constructional Steel Research 149 (2018) 64–77 77

[8] J.C. De Oliveira, J.A. Packer, C. Christopoulos, Cast steel connectors for circular hollow [16] M.G. Gary, C. Christopoulos, J.A. Packer, Cast steel yielding brace system for concen-
section braces under inelastic cyclic loading, ASCE J. Struct. Eng. 134 (3) (2008) trically braced frames: concept development and experimental validation, ASCE J.
374–383. Struct. Eng. 140 (4) (2013).
[9] C.C. McDaniel, C.-M. Uang, F. Seible, Cyclic testing of built-up steel shear links for the [17] L.W. Tong, Y. Chen, Y. Chen, C. Fang, Cyclic behaviour of beam-to-column joints with
New Bay bridge, ASCE J. Struct. Eng. 129 (6) (2003) 801–809. cast steel connectors, J. Constr. Steel Res. 116 (2016) 114–130.
[10] AISC, Seismic Provisions for Structural Steel Buildings, ANSI/AISC, Chicago, IL, 2010. [18] R. Park, Ductility evaluation from laboratory and analytical testing, Proc. Ninth
[12] A.J. Fussell, K.A. Cowie, G.C. Clifton, N. Mago, Development and Research of Eccentri- World Conference on Earthquake Engineering, August 2–9, Tokyo-Kyoto, Japan,
cally Braced Frames with Replaceable Active Links, Proc. NZSEE Conference, 2014. 1988.
[13] P. Dusicka, A.M. Itani, I. Buckle, Cyclic behavior of shear links of various grades of [19] A.M. Kanvinde, G.G. Deierlein, Cyclic void growth model to assess ductile fracture
plate steel, ASCE J. Struct. Eng. 136 (4) (2010) 370–378. initiation in structural steels due to ultra low cycle fatigue, ASCE J. Eng. Mechanics
[14] S.H. Chao, K. Khandelwal, S. El-Tawil, Ductile web fracture initiation in steel shear 133 (6) (2007) 701–712.
links, ASCE J. Struct. Eng. 132 (8) (2006).
[15] T. Okazaki, G. Arce, H.C. Ryu, M.D. Englehardt, Experimental study of local buckling,
overstrength and fracture of links in eccentrically braced frames, ASCE J. Struct. Eng.
131 (10) (2005) 1526–1535.