Engineering Structures 303 (2024) 117506

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Cyclic elasto-plastic behaviour of replaceable ductile Split-T located at the

bottom beam flange
Haowen Zheng a, b, Guozhi Cui b, *, Shoichi Kishiki b
a
Sichuan Institute of Building Research, Department of Seismic Mitigation and Isolation, No. 55, North Section 3, First Ring Road, Chengdu 600081, China
b
Tokyo Institute of Technology, J3–1, 4259 Nagatsuta, Midori, Yokohama 226-8503, Japan

A R T I C L E I N F O A B S T R A C T

Keywords: This study proposed a novel replaceable ductile Split-T (RDST) component located at the bottom beam flange to
Replaceable ductile Split-T enhance the ductility and energy dissipation capacity of steel structures. The proposed RDST aims to effectively
Beam-to-column connection control damage and enable rapid recovery of structural functions in the aftermath of earthquakes. A series of
Plastic deformation capacity
cyclic loading tests under large deformation amplitude was conducted to investigate the plastic deformation
Cyclic behaviour
capacity of this new type of structural component. The influence of shear span ratio on the specimens’ hysteretic
behaviours was studied experimentally. Meanwhile, experimental phenomena such as the pinching effect and
snap-through buckling which compose the complicated deformation behaviour were evaluated quantitatively.
Consequently, methods for quantitatively evaluating the mechanical properties of this new type of structural
component were established based on the experimental results and previous discussions on T-stub-type structural
components. Moreover, a hysteretic model was developed to precisely describe the specimens’ complicated
hysteretic behaviour. It is expected to be the key element for further research on the structures with this new type
of beam-to-column connection.

1. Introduction instrumental in facilitating urban disaster recovery.

High-strength bolted connections have been widely used in recent
The steel frame structure is the most commonly used structural form years [12–14]. They are used for beam-to-column connections owing to
owing to its superior seismic performance. However, after the 1994 their flexible and convenient installation and disassembly. In terms of
Northridge earthquake and the 1995 Kobe earthquake, the possible the urban disaster recovery perspective, high-strength bolted connec­
brittle fractures around the heat-affected zone of welded joints were tions emerge as the more viable option over welded connections due to
found to severely challenge the reliability of steel frames [1–3]. These their replaceability.
issues necessitated extensive research on innovative approaches to T-stubs and steel angles are prevalently utilized as high-strength
enhance the ductility of steel beam-to-column connections. Methods bolted connections in structural engineering. Wang et al. [15] investi­
such as applying welded haunches [4,5] and cover plates [6–8] in the gated the mechanics of bolted connections using low yield point angle
reinforcement of beam ends and weakening the cross-section at a spe­ steels. Marreiros et al. [16] utilized steel angles as energy dissipaters in
cific location on the beam far away from panel zones [9–11] were widely rocking precast systems, applying both experimental and analytical
reported in previous studies. These improvements aimed to ensure that design methods. Wang et al. [17] developed an innovative self-centering
the stress concentration regions in the beams sidestepped the steel coupling beam for RC coupled walls, integrating super-elastic
heat-affected zone of welded joints, thereby mitigating the possibility of shape memory alloy bolts and steel angles to improve seismic resil­
brittle fractures and augmenting structural ductility. The construct­ ience through damage isolation and easier post-earthquake repair. In
ability of structure repair projects was not considered during these comparison, T-stubs represent a more robust bolted connection option
performance optimizations, making the functional recovery of damaged than angle steel connections. Previous studies revealed that there are
buildings more time-consuming. Developing a steel beam-to-column three possible failure modes for T-stub-type components due to T-flange
connection with high reliability and ductility, exceptional energy thickness and bolt configuration (Fig. 1, the parts susceptible to failure
dissipation capacity, and easily replaceable elements has been are coloured in red) [18]. The yielding mechanism of one of the failure

* Corresponding author.
E-mail address: cui.g.aa@m.titech.ac.jp (G. Cui).

https://doi.org/10.1016/j.engstruct.2024.117506
Received 9 July 2023; Received in revised form 28 December 2023; Accepted 7 January 2024
Available online 18 January 2024
0141-0296/© 2024 Elsevier Ltd. All rights reserved.
H. Zheng et al. Engineering Structures 303 (2024) 117506

Fig. 2, elastic Split-T plays the role of “rotation center” in the

beam-to-column connection system. Furthermore, the shear stiffness of
elastic Split-T is significantly higher than that of RDST components.
Installing the elastic Split-T connection provides superior shear resis­
tance capabilities to the system.
The remainder of this paper is structured as follows. Section 2 in­
Fig. 1. Failure modes of T-stub components. troduces a set of cyclic loading tests with ascending deformation
amplitude for 11 RDST specimens. In Section 3, the experimental results
modes (mode No. 3 in reference [18]) is typically characterized by the are systematically discussed. Further, the mechanical properties of RDST
plastic hinges that form in two local flange areas of the T-stub. The components, including initial stiffness, strain-hardening stiffness, and
plastic hinges notably enhance the ductility of the structure and add a yield strength, are quantitatively evaluated. Section 4 evaluates the
degree of energy dissipation to the system. Various studies have opti­ pinching effect and snap-through buckling phenomenon which compose
mized the design of T-stub components based on failure modes, aiming the complicated hysteretic behaviour of RDST components. Conse­
to concentrate damage, enhance energy dissipation, and facilitate quently, a hysteretic model of RDST components is proposed based on
replaceability. For instance, Latour et al. introduced T-stub components the above investigations.
with hourglass-shaped flanges [19], while Wang et al. [20] investigated
the impact of key geometric dimensions on the deformability and energy 2. Test program
dissipation of T-stub components. Men et al. [21] conducted research on
the seismic performance of a novel steel beam-column joint with T-stub 2.1. Test specimens
connection components, validating their replaceability through experi­
ments and finite element models. Cyclic loading tests were conducted on 11 RDST components. The
However, it is conceivable that column-flanges severely constrain representative dimensions and configuration of specimens are shown in
the space allowing for the deformation of the well-designed T-flanges. Fig. 3. All specimens are identical in length (L), width (b), height (h),
Therefore, research only focusing on shape optimization of T-flanges is thickness of T-web (tw), thickness of cover plates (tc), thickness of seat
insufficient to detect the available potential of T-stub-type structural plates (ts), and radius of the fillets (R1 and R2, respectively). RDST
components.
The present study proposes a new type of replaceable T-stub beam-
to-column connection based on failure mode 3. As the T-stub compo­
nents are generally called “Split-T” in Japan, this innovative T-stub-type
device is named “replaceable ductile Split-T” and referred to as RDST in
this paper. Fig. 2 demonstrates the configuration of the RDST compo­
nent. To ensure that the T-flange can deform under compression without
any obstruction, seat plates are inserted between the column and RDST
component. The locations on the T-flange where plastic hinges initiate
can be controlled by adjusting the dimensions of the cover plates.
The concrete slab within the composite beam has led to an uneven
strain distribution, with typically higher strain levels in the bottom
flange of the steel beam than in the top flange. Moreover, after the 1994
Northridge earthquake, it was observed that the brittle fractures around
the heat-affected zone of welded joints generally started to develop from
the bottom beam flanges [1–3]. Therefore, it is supposed that installing
the RDST components on the bottom flange of steel beams can signifi­
cantly improve the deformation capacity of the beam-to-column
connection, thereby improving the ductility and reliability of the
structure. On the other hand, owing to the relatively lower strain level
acting on the top flange of the steel beam, the T-stub component
installed on the top flange generally functions during the elastic stage. Fig. 3. Outline of replaceable ductile Split-T specimens and experi­
Therefore, it is named “elastic Split-T” in this paper. As illustrated in mental parameters.

Rotation center
Concrete slab
Elastic Split-T
Force
Shear
Column

top flange force

Deformation
Steel Beam Moment

Plastic Split-T

Bottom flange Force

Cover plate Deformation

Seat plate

Fig. 2. Configuration and deformation behaviour of the RDST components.

2
H. Zheng et al. Engineering Structures 303 (2024) 117506

components were cut from H-shaped steel RH-390 × 300 × 10 × 16 Table 2

(depth H × flange width B × web thickness × flange thickness), which List of specimens.
was fabricated from high-quality rolled steel SN400B (Fig. 3(a)). To Specimen lp [mm] tf [mm] s = lp /tf
prevent premature fracture resulting from excessive stress concentration
F-0526 9 5 1.8
during the experimental process, a fillet R2 with a 4 mm radius was F-0630 13 6 2.17
intentionally incorporated on the lower side of the T-flange during the F-0736 19 7 2.71
milling process. The tensile coupon test results of the specimens are F-0836 19 8 2.38
listed in Table 1. The parameters used in the experiment include the F-0841 24 8 3
F-0850 33 8 4.13
distance from the cover plate to the T-web l and thickness of the T- F-0855 38 8 4.75
flange’s working section tf. Table 2 lists the specimens’ code names and F-0949 32 9 3.56
their experimental parameters. The definition of shear span ratio s is F-1060 43 10 4.3
given by Eq. (1), as follows: F-1070 53 10 5.3
F-1164 47 11 4.27
lp
s= , (1)
tf

where lp is the distance between the ends of the fillets on the one-side T-
flange, which is used for describing the length of the area between the
plastic hinges that form in the one-side T-flange (area between sections
SR1 and SR2 in Fig. 3). The designed value of shear span ratio s ranges
from 2 to 5.5, increasing in increments of 0.5. However, slight dis­
crepancies between the actual and designed values of s arose due to
fabrication and installation variances. Only the actual measurement
results of s after the completion of specimen installation are presented in
Table 2. Simultaneously, the value of tf ranges from 5 mm to 11 mm.
Considering the limitations of the experimental setup and aiming for
precise control of the formation location of plastic hinges on the T-
flange, in general, specimens with smaller values of s also have corre­
spondingly smaller tf values.

2.2. Setup

A series of cyclic loading tests were conducted on RDST specimens

using the setup illustrated in Fig. 4. RDST components primarily dissi­
pate earthquake-induced energy through axial tensile and compressive
deformation. The experimental setups are designed to apply axial cyclic
loading to the specimens. The RDST component was connected to a
reaction frame via high-strength bolts F14T-M24. The T-web was con­
nected to a loading jig using super high-strength bolts F14T-M16 × 3 (σ u
=1400 N/mm2). No slips occurred between the loading jig and T-web
during the entire process of the cyclic loading test. The loading jig was
installed on an actuator. Therefore, the parallel horizontal movement
Fig. 4. Experimental setup.
during cyclic loading could be applied to the specimen using the actu­
ator until fracture. The loading direction is defined as positive (+) when
the specimen is tensioned (δ ≥ 0); conversely, negative (-) denotes that
the specimen is subjected to compressive loading (δ < 0).

2.3. Loading protocol and measurement plan

The RDST components were loaded with the loading history indi­
cated in Fig. 5. The deformation amplitude increased to ± 15 mm in 9
steps. Subsequently, the deformation amplitude was maintained con­
stant until the fracture of specimens. A built-in load cell in the actuator
measured the force F in the system. The specimen’s deformation δ shown
in Fig. 5 is the distance change from the seat plate’s lower surface to the
targets located on T-web. The δ values are obtained by calculating the
average of the measured value of displacement transducers δ1 and δ2 Fig. 5. Loading protocol.
(Fig. 4). Furthermore, δ can be converted into the rotation angle of the
beam end θb, which is defined as the ratio between the value of δ to the beam height (H). The target beam rotation angle here is set as 0.05 ra­
dians to investigate the behaviour of plastic Split-T under large
Table 1 deformation.
Tensile coupon test results.
σy [N/mm2] σu [N/mm2] σy /σu
2.4. Test results
T-flange 275 420 0.65
T-web 315 433 0.73
The loading tests were halted when the fracture of specimens

3
H. Zheng et al. Engineering Structures 303 (2024) 117506

occurred. Since the deformation behaviour of each specimen was represents the accumulated plastic deformation experienced by a spec­
observed to be similar in the cyclic loading tests, the images of specimen imen during the entire process of the cyclic loading test. When the
F-1060 in the last two loading cycles are shown as an example in Fig. 6 deformation levels are identical, specimens with a smaller s display a
with the corresponding F-δ relationship (Fig. 6(a)). Images of the spec­ comparatively larger rotation angle, resulting in a more pronounced
imen when δ reaches the peak value under both tension and compression strain on the section where plastic hinges develop on the T-flange. This
are indicated in Fig. 6(b) and Fig. 6(c), respectively, while Fig. 6(d) is the phenomenon is evident in the hysteric behaviour of the specimens, with
image of the specimen when the fracture occurs. As shown in Fig. 6(a), those possessing a smaller s exhibiting earlier fractures. In contrast,
an urge in F value is observed in the section of the F-δ relationship RDST specimens with a larger s showcase enhanced fatigue performance
marked in red. This sudden growth in stiffness caused by the increasing and plastic deformation capability under equivalent deformation am­
axial tension of the T-flange is defined as the “pinching effect” in the plitudes, coupled with superior adaptability to large deformations. Thus,
present study. Namely, the pinching effect specifically refers to the it is observed in Fig. 8 that both the values of Eu and δsum increase with
notable experimental phenomenon of a rapid increase in the T-flange’s increasing shear span ratio s of the specimens. The RDST components
bearing capacity under axial tension. Moreover, when the specimen is with relatively larger s are expected to have greater plastic deformation
loaded under compressive loading, a sudden reduction in the F value is capacity.
observed when deformation δ decreases to a specific value (the section The envelope curves of the specimens’ hysteric loops are employed
of the F-δ relationship marked in blue). This indicates that snap-through to depict the force-deformation (F-δ) relationship of the RDST compo­
buckling occurs because of the compressive loading that acts on the nents loaded under monotonic tensile loading (Fig. 9) in this study [22,
severely axially deformed T-flanges. Elastic deformation was absent in 23]. The experimental results of the RDST components’ initial stiffness
the bolts, cover plates, and seat plates; however, plastic deformation and Ke, strain-hardening stiffness Kp, yield strength Fy, yield deformation δy,
fractures were exclusively observed in the T-flanges. This indicates that peak strength Fu, and peak deformation δu can be obtained based on the
damage caused by cyclic loading is effectively concentrated on T- envelope curves illustrated in Fig. 9. The occurrence of a significant
flanges, and the other parts of the system are protected from irreversible inflection point (yield point) in the force-deformation relationship is
damage. Hence, the concept of RDST components is proven feasible and considered an indication that the specimen has entered the plastic
effective. deformation stage. The values of δy and Fy are identified as the X and Y
Although all the specimens exhibit similar hysteretic behaviours, the coordinates of the intersection of strain-hardening stiffness Kp and initial
ones with the smallest and largest shear span ratios s are considered to stiffness Ke, respectively [24]. These indicators that affect the RDST
be the most representative. Therefore, only the experimental results of components’ mechanical behaviour are systematically discussed in
force-deformation (F-δ) relationship for specimens F-0526 and F-1070 Section 3.
are illustrated in Fig. 7. Due to the length limit of the text, experimental The following section delves into an exploration of the impact of the
results of the remaining nine specimens have been shown in Appendix A. shear span ratio on the fundamental mechanical characteristics man­
For all specimens, hysteretic loops are stable and plump before δ be­ ifested in the envelope curve. The shear span ratio-strength yield ratio
comes approximately 3 mm; afterward, pinching effect and snap- relationship and shear span ratio-hardening ratio relationships are
through buckling interactively occur in the large deformation. The indicated in Fig. 10. Here, strength-yield ratio is defined as the ratio of
phenomena observed in large deformation compose the complicated maximum strength to yield strength (Fig. 10 (a)). Hardening ratio is
hysteretic behaviour, which is evaluated in Section 4. defined as the ratio of strain-hardening stiffness to initial stiffness
The experimental results of the plastic deformation capacity of RDST (Fig. 10 (b)). As previously mentioned, specimens with a larger s
components are illustrated in Fig. 8. The cumulative energy dissipation demonstrate superior adaptability to large deformation. Additionally,
Eu of a specimen, as shown in Fig. 8(a), is inferred by calculating the sum under significant deformations, the axial force impact on the T-flange is
of the area of each hysteretic loop, whereas δsum, shown in Fig. 8(b), more pronounced in specimens with a larger s. Therefore, when yield

Fig. 6. Deformation behaviour of RDST specimen F1060.

4
H. Zheng et al. Engineering Structures 303 (2024) 117506

Fig. 7. Hysteretic behaviour of RDST specimen.

Fig. 8. Plastic deformation capacity of RDST components.

strength and initial stiffness are at the same level, specimens with a
larger s exhibit greater strain-hardening stiffness Kp, ultimately resulting
in higher peak strength Fu.

3. Mechanical behaviour of RDST under monotonic loading

In this section, the mechanical behaviour of RDST components

loaded under monotonic loading is discussed based on experimental
results and finite element modelling (FEM).

3.1. Finite element modelling of RDST components

Fig. 9. F-δ relationship under tensile loading. Three-dimensional (3D) finite element models (FE models) are
developed for all specimens using Software ABAQUS /CAE 2021 to
simulate the behaviour of RDST components loaded under monotonic
tension loading. The FE model of specimen F-0841 is illustrated in
Fig. 11 (a) (as an example). Only half of a specimen is modelled to save
computational time. the results of coupon tests (Table 1) are directly
used in determining the material properties of the analytical model’s T-
flanges and T-webs. Young’s Modulus and Poisson’s ratio are assumed to
be 205 GPa and 0.3, respectively [18]. For all contact pairs, the
surface-to-surface contact is set to be “finite sliding”. The lower surfaces
of the seat plates are fixed; subsequently, X-axial displacement δ is
applied to a point that is coupled with the T-web inner surface.
The analytical and experimental results of the F-δ relationship,
Fig. 10. The mechanical performance of RDST components. development of the T-flange’s stress distribution, and ultimate state of
specimen F-0841 are compared and illustrated in Fig. 11 (b) (as an
example). Similar to the other specimens, the analytical results of the
corresponding FE model show high agreement with the experimental

5
H. Zheng et al. Engineering Structures 303 (2024) 117506

Fig. 11. Finite element modelling.

results. Thus, FE modelling is verified to be a practical approach for

investigating the deformation behaviour of RDST components loaded
under monotonic loading. The grey-coloured zones denote the areas
where the stress σ has already exceeded the yield stress of the material
σ y. Primarily, the T-flanges yield from the ends of the fillets (sections
SR1 and SR2 in Fig. 3), and plastic hinges gradually form at the current
location with increasing deformation δ. This is why lp is used as the
indicator for describing the distance between the plastic hinges on the
one-side T-flange in this study. Subsequently, the continually increasing
δ values result in the rapid growth of the T-flanges’ axial force; hence,
the plastic zones tend to develop along the T-flange and soon spread
through the entire lp region. Furthermore, as indicated in Fig. 11 (b),
fractures commonly occur in sections SR1 and SR2.
Fig. 12. Experimental results and estimated values of yield strength Fy.

3.2. Evaluation methods for RDST components’ yield strength and initial [( )− 1 ]
stiffness l3e le ( )
Ke = 2⋅ + δ ≤ δy (3)
E⋅b⋅t3f G⋅b⋅tf
The behaviour exhibited by RDST components prior to the formation
of plastic hinges is comparable with that of a conventional T-stub-type where E is the elastic modulus, and G is the shear modulus. le rep­
component, which fails in failure mode No. 3. Plastic hinges tend to form resents the effective length of RDST specimens, which is the difference
in the sections close to the inner ends of fillets (sections SR1 and SR2 in between l and 0.2 times the sum of the radius of fillets R1 and R2. δy
Fig. 3), and the force in system F is the sum of shear forces acting on the denotes the yielding deformation of RDST components that is calculated
T-flanges. Thus, the yield strength Fy can be evaluated using Eq. (2) by considering Eq.(2) and Eq. (3).
based on the yield strength evaluation method for T-stub-type compo­ The experimental results of the initial stiffness Ke are indicated in
nents’ failure mode No. 3, proposed by the Recommendation for Design Fig. 13 in comparison with the estimated values obtained using Eq. (3).
of Connections in Steel Structures [18]: The precision of the evaluation methods is verified, as shown in
Fig. 14. In this figure, the X- and Y-axes denote the estimated values (Fy
(cal.) and Ke(cal.)) and experimental results (Fy(exp.) and Ke(exp.)), respec­
2⋅(MSR1 + MSR2 ) b⋅t2
Fy = , or Fy = f σ y (2)
lp lp tively. Although the experimental results of Fy are slightly under­
Where MSR1 and MSR2 denote the full plastic moments of the sections estimated by Eq. (2), most of the experimental results of both Fy and Ke
where plastic hinges form. b is the width of RDST specimens. σ y is the
yield stress of the T-flange.
Compared with the experimental results of the yield strength of the
RDST components, the values estimated using Eq. (2) are indicated in
Fig. 12. Although the yield strength of specimens loaded under both
tension and compression loading show high consistency, there are still
some non-negligible differences. Here, the black and red markers indi­
cate the experimental results of yield strength under tensile loading (F+ y)
and compressive loading (F-y), respectively. The same approach is
applied to the illustration of other experimental results.
However, unlike slender rod systems, the impact of the shear stiff­
ness of the T-flanges cannot be ignored in RDST components owing to
their relatively small shear span ratio s. Thus, Eq. (3) is proposed here to
estimate the initial stiffness Ke of the system: Fig. 13. Experimental results and estimated values of initial stiffness Ke.

6
H. Zheng et al. Engineering Structures 303 (2024) 117506

Fig. 14. Precision verification.

are scattered over the range of 0.8–1.2 times the estimated values.

3.3. Behaviour of RDST components in the elastic-plastic stage

Analysing the results of FE modelling, the present study confirms the

effectiveness of the previously discussed failure mode No. 3 of conven­
tional T-stub-type components for describing the behaviour of RDST
components during the elastic-plastic stage. Fig. 15 indicates the
analytical results of the FE model described in Section 3.1.
F(FEM) and Fy(FEM) denote the FE modelling results of the force in the
system F and yield strength Fy, respectively. F’(FEM) is the estimated value
of F, calculated by substituting the FE modelling results of the moment
change at sections SR1 and SR2 in Eq. (2) F(FEM) is severely under­
estimated by Eq. (2), indicating that the failure mode No. 3 of T-stub-
type components can no longer accurately reflect the behaviour of RDST
specimens that are subjected to monotonic loading with large defor­
mation. On the other hand, incremental deformation leads to a sharp
increase in the axial force of T-flanges, and the rotation angle change of
T-flanges significantly affects the behaviour of RDST components in this
scenario. A new mechanical model is proposed, as shown in Fig. 16, and
defined as Eq. (4) for describing the behaviour of RDST components
after the yielding of the system. Not only the effect of shear force Q but
also the effect of axial force N in T-flanges on the value of F are
considered. The force in the system is now supposed to be the sum of the
components in the loading direction of the T-flanges’ axial force N and Fig. 16. Mechanical model for the RDST components loaded under mono­
shear force Q, as described as follows: tonic loading.
[ ( ) ( )]
F = 2⋅ N⋅ sin θ − θy + Q⋅ cos θ − θy , (4) Another phenomenon that can be observed in Fig. 15 is that the value
of F’(FEM) is close to that of Fy (FEM) and almost remains constant after the
where θ is given by Eq. (5), which describes the rotation angle of the T- yielding of the system. This indicates that the plastic hinge approach and
flanges, and θy denotes the value of rotation angle when δ reaches δ y. Eq. (2) can still be used for evaluating the shear force Q acting on the
δ one-side T-flange. Thus, the value of Q in Eq. (4) and Fig. 16 can be
θ= . (5) expressed as 0.5 times the estimated value of Fy in Eq. (2).
lp
Collating the experimental results, the effect of the axial force in the
one-side T-flange N on the force in the system F is presented in Fig. 17. In
this figure, the black line indicates the envelope curve of an RDST
specimen (F-0841), and the specimen’s yield deformation δy is marked
in red. The grey line indicates the effect of shear force Q, which is
inferred by finding the component in the loading direction of the esti­
mated value of Q. Thus, according to the definition of the mechanical
model of RDST components described in Eq. (4), FN —the difference
between the value of F and effect of shear force Q is the effect of axial
force N on the force in the system. Therefore, the value of N can be
calculated using (Eq. (6)), as follows:
F − 2⋅Q⋅ cos θ
N= . (6)
2⋅ sin θ
Representing the ratio of the T-flange’s rotation angle θ to yield
rotation angle θy on the X-axis, the growth of axial force N is presented in
Fig. 15. Analytical results obtained from FE modelling.

7
H. Zheng et al. Engineering Structures 303 (2024) 117506

( ) ( )
1.4⋅Ny ⋅ sin θ − θy + Fy ⋅ cos θ − θy ( )
Kp = δ > δy (7)
δ
The precision of Eq. (7) is verified, as shown in Fig. 19. Both strain-
hardening stiffness Kp and maximum strength Fu are precisely estimated
using Eq. (7). For most specimens, the experimental results range from
0.8 to 1.2 times the estimated value.
Consequently, a new mechanical model is established, as shown in
Fig. 16 and Eq. (4), for RDST components during the elastic-plastic
stage. By considering Eq.(2), Eq. (3), and Eq. (4), the backbone curve
for the hysteretic model of RDST components can be expressed as Eq.
(8).
( )
F = Ke ⋅δ δ ≤ δy (8-a)
( )
Fig. 17. Effect of axial force in the one-side T-flange N on the force in the F = Kp ⋅δ δ > δy (8-b)
system F.

4. Hysteretic behaviour and hysteretic model of RDST

components

A hysteretic model is proposed (described in this section) to precisely

predict the behaviour of RDST components under cyclic loading in this
section.

4.1. Hysteretic behaviour of RDST under large deformation amplitude

As indicated in Section 3.2, the increase in the T-flanges’ axial force

facilitates the increase in the F value after the formulation of plastic
hinges. The behaviour of the T-flanges in this case represents that of
tightened wires. By contrast, compared with the conventional T-stub-
type components, snap-though buckling is more likely to occur in RDST
Fig. 18. α-θ/θy relationship.
components owing to their relatively slender T-flanges and the effect of
seat plates. The sharp decrease in the F value caused by snap-through
Fig. 18. In this figure, the Y-axis denotes the axial force ratio α, which is buckling and subsequent surge in F derived from tightened T-flanges
defined as the ratio of N to the axial yield strength Ny of the lp region’s occur alternatively. This is why the pinching effect significantly in­
section. α-θ/θy relationships of all RDST specimens are indicated in grey. fluences the hysteretic behaviour of RDST components. In the present
Conversely, the average values of α are indicated in black. The average study, the points where the force in the system sharply increases and
values of α can reflect the trend change of the T-flange’s axial force. It is where snap-through buckling initiates are now referred to as pinching
worth noting that α rapidly increases before θ reaches approximately ten and snap-through buckling points (points P (δpinch, Fpinch) and S(i) (δsnap
times θy and tends to hold firm near 0.7. Subsequently, N in the T-flange (i), Fsnap(i)) of loading cycle No. i), respectively.
reaches approximately 0.7 times Ny after the yielding of the system. To The experimental results are indicated in Fig. 20 and Fig. 21. In these
simplify the definition of the mechanical model discussed in Fig. 16, the figures, the X- and Y-axes indicate the shear span ration s of the specimen
value of T-flanges’ axial force N in Eq. (4) is now expressed as 0.7 times and the pinching strength Fpinch or snap-through buckling strength Fsnap,
Ny . respectively. It is evident that both the values of Fpinch and Fsnap decrease
Overall, based on the investigation described above, the strain- with increasing s, indicating that for RDST components loaded under the
hardening stiffness Kp of RDST components can be estimated using Eq. identical loading level, the pinching effect is more significant, and snap-
(7), as follows: through buckling is more likely to occur in the ones with relatively
slender T-flanges. Moreover, Fpinch under tensile loading is highly
consistent with that under compressive loading. However, the value of

Fig. 19. Precision verification of Kp and Fu.

8
H. Zheng et al. Engineering Structures 303 (2024) 117506

4.1.2. Evaluation method for snap-through buckling of RDST components

The lp region is modelled as a rod that is hinged at both ends in the
mechanical model, described in Fig. 16. The slenderness ratio λ of RDST
specimens is considerably lower than that of the critical slenderness
ratio λcr, making the calculated value of critical buckling stress σcr
almost equal to the yield strength of the specimens σ y. Thus, the eval­
uation method of RDST components’ critical snap-through buckling
force Fcr can be expressed as Eq. (10). However, as mentioned earlier,
the experimental results of Fsnap
+
are usually lower than those of Fsnap
−
,
indicating the significant overestimation of the Fsnap
+
value by Eq. (10-a).
Accordingly, a coefficient of 0.8 is applied for describing the snap-
through buckling behaviour of RDST components during reloading
Fig. 20. Experimental results of the pinching point. phases (Eq. (10-b)). Snap-through buckling is considered to occur when
F in the system exceeds the evaluated Fcr in the loading direction.
F −cr = 2Ny sin θ + Fy cos θ (θ ≥ 0) (10-a)
( )
F+
cr = 0.8⋅ 2Ny sin θ + Fy cos θ (θ < 0) (10-b)

Compared with the experimental results of snap-through buckling

points S, Eq. (10) is indicated in Fig. 23. Estimation using Eq. (10) is a
simple and feasible prediction method for snap-through buckling.

4.2. Establishment of the hysteretic model of RDST components

Based on the backbone curve established in Section 3 and the above

discussion, a hysteretic model of RDST components is established in this
Fig. 21. Experimental results of snap-through buckling strength Fsnap. section. It is worth noting that for all specimens, hysteretic loops remain
stable and full before the peak deformation in the loading direction δpeak
Fsnap
+
is approximately 0.8 times that of Fsnap
−
. reaches the pinching deformation δpinch. Moreover, the effects of
pinching behaviour and snap-through buckling tend to be more pro­
nounced afterward. Therefore, the pinching effect and snap-through
4.1.1. Evaluation method for the pinching effect of RDST components
buckling are considered once the value of δpeak exceeds that of δpinch
To precisely describe the pinching effect on RDST components’
in the loading direction.
hysteretic behaviour, the
The hysterical model for the case δpeak < δpinch is illustrated in
relationship between the experimental results of δpinch and shear
Fig. 24. When δpeak ≥ δpinch, the pinching effect and snap-through
span ratio s of the specimens is illustrated in Fig. 22. Here, δpinch is
buckling are considered, as shown in Fig. 25. The backbone curve
converted into the T-flanges’ rotation angle at the pinching point θpinch,
established in Section 3 (solid blue line) is plotted in both Fig. 24 and
using Eq. (5). The relationship between θpinch and s can be evaluated by
Fig. 25. The yield strength Fy is defined in Eq. (2). The unloading and
an inversely proportional function of s. Consequently, a prediction
reloading stiffness remain the same as initial stiffness Ke (Eq. (3)),
approach for the RDST components’ pinching behaviour is obtained by
whereas the strain-hardening stiffness Kp is evaluated using Eq. (7). As
curve fitting (Eq. (9)), as follows:
cyclic strength and stiffness deterioration are not considered in this
θpinch =
δpinch 0.4
= . (9) study, the strength in the loading path increases along the backbone
lp s curve in later stages of deformation amplitude. The softening of the
material stiffness caused by the Bauschinger effect is reflected in the case
of δpeak < δpinch at points 3, 6, and 9. The stiffness decreases at points 6
and 9 when the value of strength reaches 0.5 times that of the previous
unloading points in the loading direction (points 2 and 5, respectively)
[25,26]. Then, the loading segments are directed towards the previous

Fig. 23. Comparison between Eq. (10) and test results of snap-through buck­
Fig. 22. Evaluation method of the pinching point. ling point S.

9
H. Zheng et al. Engineering Structures 303 (2024) 117506

Fig. 24. Hysteretic model of RDST components (δpeak < δpinch).

illustrated in Fig. 26 (as examples). The black and grey solid lines
represent the hysteretic model and the corresponding hysteretic curve,
respectively.
The precision of the hysteretic model established in the present study
is verified, as shown in Fig. 27. The evaluated and experimental results
of the pinching strength Fpinch, critical buckling force Fsnap, energy
dissipation of each loading cycle Ei, and cumulative energy dissipation
Esum are illustrated in Fig. 27. In this figure, the Esum value is calculated
by summing the energy dissipation during the loading cycles before the
initiation of strength deterioration, and the estimated value of Fsnap
(Fsnap(cal.)) is the vertical coordinate of snap-through buckling point. For
almost all specimens, the experimental results of energy dissipation
remain approximately 1.2 times the estimated values, and the experi­
mental results of pinching strength scatter within 0.8–1.2 times the
estimated values. The experimental results are slightly underestimated
but highly consistent with the estimated values.

5. Conclusion

A comprehensive study was conducted on the replaceable ductile

Fig. 25. Hysteretic model of RDST components (δpeak ≥ δpinch). Split-T structural proposed in this study. The main conclusions are as
follows:
unloading points. Moreover, in the case that a hysteretic loop does not
(1) A series of cyclic loading tests are conducted on the RDST spec­
pass through the Y-axis, such as the segments 8–11, the criteria
imens. The experimental results indicate that the RDST compo­
described earlier still apply. It should be noted that the strength of points
nent leads to excellent energy dissipation, and the plastic
3 (F3) on the first loop is defined as 0.5 times the yield strength in the
deformation capacity tends to increase with increasing shear span
loading direction (F-y) owing to the lack of reference points. Subse­
ratio of the component. Furthermore, installing the RDST
quently, the loading segment is directed towards the yield point.
component on the bottom flange can significantly improve the
However, the shape of hysteretic loops changes due to the pinching
ductility and reliability of the structure.
effect and snap-through buckling in the case of δpeak ≥ δpinch (Fig. 25).
(2) The analytical results obtained from finite element modelling
Owing to the Bauschinger effect observed during the reloading phases,
indicate that the behaviour of RDST components during the
the stiffness decreases to 0.1 times the initial stiffness Ke when the value
elastic stage is highly consistent with that of conventional T-stub-
of strength reaches 0.5 times Fpinch in the loading direction at points 3, 7,
type components. Therefore, the mechanical model proposed in
and 11. Then, the loading segments continue up until they intersect with
the Japanese Recommendation for Design of Connections in Steel
Eq. (10), which represents the critical snap-through buckling force Fcr.
Structure for steel structure remains in use in the behaviour
Subsequently, the loading segments point towards the pinching points
description of RDST components during the elastic stage. In
estimated using Eq. (9). The pinching effect is not considered when a
addition, the effect of T-flanges’ increasing axial force and rota­
hysteretic loop does not intersect with the Y-axis, as demonstrated by
tion angle on the force in system tends to be more significant after
segments 10–14. The loading segment that starts from point 14 is
the formulation of plastic hinges. To precisely predict the
directed towards the previous unloading points in the loading direction.
behaviour of RDST components during the elastic-plastic stage, a
The comparisons between the hysteretic model and experimental
novel mechanical model is proposed, as discussed in Section 3.
results of hysteretic curves of RDST specimens F-0841 and F-1070 are

10
H. Zheng et al. Engineering Structures 303 (2024) 117506

Fig. 26. Comparison between the hysteretic model and experimental results (F-δ) relationship.

Fig. 27. Precision verification of the hysteretic model.

(3) The pinching effect and snap-through buckling significantly hysteretic model proposed in this study will be completed with a sys­
affect the hysteretic behaviour of the RDST components. It can be tematic investigation of the cyclic hardening, cyclic strength, and stiff­
inferred from the experimental results that the pinching behav­ ness deteriorating behaviour. Additionally, future research will explore
iour and snap-through buckling tend to occur in the specimens the generation and impact of column rotation in beam-column joint
with relatively slender T-flanges when the RDST components are systems incorporating the plastic Split-T. Moreover, future research will
loaded under identical loading levels. employ a combination of experimental studies and finite element
(4) The RDST components exhibit stable hysteretic behaviour before parameterized analysis to systematically evaluate the fatigue and energy
the experienced peak deformation reaches the pinching defor­ dissipation performance of RDST components.
mation in the loading direction, whereas the pinching effect and
snap-through buckling complicate the shape of hysteretic loops CRediT authorship contribution statement
afterward. Therefore, a two-stage hysteretic model is composed
in this study to precisely predict the behaviour of RDST compo­ Haowen Zheng: Writing – review & editing, Visualization. Guozhi
nents under cyclic loading with large deformation amplitude. Cui: Writing – review & editing, Writing – original draft, Visualization,
Software, Data curation. Shoichi Kishiki: Writing – review & editing,
In the future, the hysteretic behaviour of RDST components under Supervision, Methodology, Funding acquisition, Conceptualization.
earthquake-induced excitations will be further studied. Further, the

11
H. Zheng et al. Engineering Structures 303 (2024) 117506

Declaration of Competing Interest Acknowledgments

The authors declare that they have no known competing financial We would like to express our sincere gratitude to Professor Akira
interests or personal relationships that could have appeared to influence Wada (Tokyo Institute of Technology), Professor Satoshi Yamada (The
the work reported in this paper. University of Tokyo), Professor Toru Takeuchi (Tokyo Institute of
Technology), and Dr. Ken Okada (Nikken Sekkei Ltd.) for their valuable
Data Availability insights and contributions to the experimental design of this study.

Data will be made available on request.

Appendix A

With the exception of specimens F-0526 and F-1070, the hysteresis curve experimental results for the remaining 9 specimens are depicted in
Fig. A1. All specimens exhibit similar hysteric behavior, with the occurrence of snap-through buckling and pinching effects observed in each case.
Notably, specimens F-0836, F-0949, F-1060, and F-1164 demonstrated strength degradation during the final loop loading, a phenomenon will be
systematically validated in future work.

12
H. Zheng et al. Engineering Structures 303 (2024) 117506

500 F [kN] 500 F [kN]

0 δ [mm] 0 δ [mm]
-20 0 20 -20 0 20

-500 -500

(a) Specimen F-0630 (s=2.17) (b) Specimen F-0736 (s=2.71)

500 F [kN] 500 F [kN]

0 δ [mm] 0 δ [mm]
-20 0 20 -20 0 20

-500 -500

(c) Specimen F-0836 (s=2.38) (d) Specimen F-0841 (s=3)

500 F [kN] 500 F [kN]

0 δ [mm]
0 δ [mm] -20 0 20
-20 0 20

-500
-500
(e) Specimen F-0850 (s=4.13) (f) Specimen F-0855 (s=4.75)
500 F [kN] 500 F [kN]

0 δ [mm] 0 δ [mm]
-20 0 20 -20 0 20

-500 -500
(g) Specimen F-0949 (s=3.56) (h) Specimen F-1060 (s=4.3)

500 F [kN]

0 δ [mm]
-20 0 20

-500
(i) Specimen F-1164 (s=4.27)
Fig. A1. Test results with different shear span ratio.

References [2] Mahin SA. Lessons from damage to steel buildings during the Northridge
earthquake. Eng Struct 1998;20:261–70.
[3] Akiyama H. Evaluation of fractural mode of failure in steel structures following
[1] Miller DK. Lessons learned from the Northridge earthquake. Eng Struct 1998;20:
Kobe lessons. J Constr Steel Res 2000;55:211–27.
249–60.

13
H. Zheng et al. Engineering Structures 303 (2024) 117506

[4] Yu QS, Uang CM, Gross J. Seismic rehabilitation design of steel moment connection [16] Marreiros R, Lúcio V, Pampanin S. Development of steel angles as energy
with welded haunch. J Struct Eng 2000;126:69–78. dissipation devices for rocking connections. Struct Concr 2018;19(6):1657–71.
[5] Lee CH, Jung JH, Oh MH, Koo ES. Cyclic seismic testing of steel moment [17] Wang B, Nishiyama M, Zhu S, et al. Development of novel self-centering steel
connections reinforced with welded straight haunch. Eng Struct 2003;25:1743–53. coupling beams without beam elongation for earthquake resilience. Eng Struct
[6] Engelhardt MD, Sabol TA. Reinforcing of steel moment connections with cover 2021;232:111827.
plates: benefits and limitations. Eng Struct 1998;20:510–20. [18] Architectural Institute of Japan: Recommendation for design of connections in steel
[7] Kim T, Whittaker AS, Gilani ASJ, Bertero VV, Takhirov SM. Cover-plate and flange- structure. 2012 (in Japanese).
plate steel moment-resisting connections. J Struct Eng 2002;128:474–82. [19] Latour M, Rizzano G. Design of X-shaped double split tee joints accounting for
[8] Deng E, Zong L, Ding Y, Dai X, Lou N, Chen Y. Monotonic and cyclic response of moment-shear interaction. J Constr Steel Res 2015;104:115–26.
bolted connections with welded cover plate for modular steel construction. Eng [20] Wang M, Dong K, Liu M. Damage control mechanism and seismic performance of a
Struct 2018;167:407–19. steel moment connection with replaceable low-yield-point steel double T-stub
[9] Popov EP, Yang T, Chang S. Design of steel MRF connections before and after 1994 fuses. Thin-Walled Struct 2020;157:107143.
Northridge earthquake. Eng Struct 1998;20:1030–8. [21] Men J, Li T, Zhou Q, Zhang H, Huang C. Structural fuse performance and
[10] Garoosi AM, Roudsari MT, Hashemi BH. Experimental evaluation of rigid earthquake-resilient performance of beam-column joints with replaceable T-stub.
connection with reduced section and replaceable fuse. Structures 2018;16: J Constr Steel Res 2023;206:107943.
390–404. [22] Kishiki S, Fu X, Tatsumi N. Seismic performance evaluation of non-structural light-
[11] Morshedi MA, Dolatshahi KM, Maleki S. Double reduced beam section connection. gauge-steel frame drywall partition subjected to in-plane cyclic loading. Thin-
J Constr Steel Res 2017;138:283–97. Walled Struct 2022;181:110099.
[12] Zhao Z, Liang B, Liu H, Li Y. Simplified numerical model for high-strength bolted [23] Liang W, Lan J, Zou Z, Zhu Z, Zhai S, Mo Y, et al. Seismic behavior of Chuan-dou
connections. Eng Struct 2018;164:119–27. timber frames with different infilled walls. Constr Build Mater 2022;327:126906.
[13] Zhang Y, Gao S, Guo L, Fu F, Wang S. Ultimate tensile behavior of bolted stiffened [24] Park R. Evaluation of ductility of structures and structural assemblages from
T-stub connections in progressive collapse resistance. J Constr Steel Res 2022;189: laboratory testing. Bull NZ Soc Earthq Eng 1989;22(3):155–66.
107111. [25] Akiyama H. Influence of Bauschinger effect on seismic resistance of steel
[14] Zhang Y, Chen B, Liu A, Pi Y, Zhang J, Wang Y, et al. Experimental study on shear structures. J Struct Constr Eng Trans AIJ 1990;418:49–57.
behavior of high strength bolt connection in prefabricated steel-concrete composite [26] Yamada S, Jiao Y. A concise hysteretic model of structural steel considering the
beam. Compos B: Eng 2019;159:481–9. Bauschinger effect. Int J Steel Struct 2016;16:671–83.
[15] Wang M, Tong Y, Sun Y, et al. Structural fuse design of bolted steel frame with low
yield point angle steel connection components. Thin-Walled Struct 2022;179:
109553.

14