A Study On Detailing Gusset Plate and Bracing Members in Concentrically Braced Frame Structures

Advances in Computational Design, Vol. 3, No.
3 (2018) 233-267
DOI: https://doi.org/10.12989/acd.2018.3.3.233 233
A study on detailing gusset plate and bracing members in

concentrically braced frame structures
M.S. Hassan1,2, S. Salawdeh1,2, A. Hunt3, B.M. Broderick4 and J. Goggins1,2
1
Civil Engineering, School of Engineering, National University of Ireland Galway, Ireland
2
MaREI Centre for Marine and Renewable Energy Ireland, Ryan Institute, National University of Ireland
Galway, Ireland
3
AECOM, St Albans, United Kingdom
4
Department of Civil, Structural & Environmental Engineering, Trinity College Dublin, Ireland
(Received June 20, 2017, Revised July 11, 2018, Accepted July 12, 2018)
Abstract. Conventional seismic design of concentrically braced frame (CBF) structures suggests that the
gusset plate connecting a steel brace to beams and/or columns should be designed as non-dissipative in
earthquakes, while the steel brace members should be designed as dissipative elements. These design
intentions lead to thicker and larger gusset plates in design on one hand and a potentially under-rated
contribution of gusset plates in design, on the other hand. In contrast, research has shown that compact and
thinner gusset plates designed in accordance with the elliptical clearance method rather than the
conventional standard linear clearance method can enhance system ductility and energy dissipation capacity
in concentrically braced steel frames. In order to assess the two design methods, six cyclic push-over tests on
full scale models of concentric braced steel frame structures were conducted. Furthermore, a 3D finite
element (FE) shell model, incorporating state-of-the-art tools and techniques in numerical simulation, was
developed that successfully replicates the response of gusset plate and bracing members under fully reversed
cyclic axial loading. Direct measurements from strain gauges applied to the physical models were used
primarily to validate FE models, while comparisons of hysteresis load-displacement loops from physical and
numerical models were used to highlight the overall performance of the FE models. The study shows the
two design methods attain structural response as per the design intentions; however, the elliptical clearance
method has a superiority over the standard linear method as a fact of improving detailing of the gusset plates,
enhancing resisting capacity and improving deformability of a CBF structure. Considerations were proposed
for improvement of guidelines for detailing gusset plates and bracing members in CBF structures.
Keywords: seismic design; concentrically braced frames; steel structures; earthquake engineering; gusset
plates; steel hollow sections; finite element modelling
1. Introduction
Concentrically braced frame (CBF) steel structures, in which the diagonal members intersect
the centre line of beams and columns, forms one of the most effective systems for providing
Corresponding author, Ph.D. E-mail: [email protected]
Copyright © 2018 Techno-Press, Ltd.

http://www.techno-press.org/?journal=acd&subpage=7 ISSN: 2383-8477 (Print), 2466-0523 (Online)
234 M.S. Hassan, S. Salawdeh, A. Hunt, B.M. Broderick and J. Goggins
seismic resistance in both low- and high-rise buildings. They resist build-up forces during
earthquakes by vertical in plane truss mechanisms consisting of alternating compression and
tension forces in the bracing members. In strong earthquakes, however, these bracing members can
be subjected to limit states, such as local buckling and fracture. As a result, the key focus of earlier
research was mainly on investigating the cyclic axial response of steel braces (Tremblay 2002,
Goggins 2004, Haddad et al. 2004, Goggins et al. 2006, Goggins and Salawdeh 2013, Haddad
2015, Ryan et al. 2017). It was generally concluded that global and local slenderness were two
important parameters influencing the local buckling and fracture. Consequently, slenderness limits
were proposed in detailing bracing members to obtain better fatigue lives during earthquakes.
More recently, interest has shifted to examination of details influencing gusset plate behaviour
in CBF structures. Gusset plates are a structurally efficient means of connecting steel braces with
beams and columns of the CBF, but their design must accommodate the unique hysteretic response
behaviour of the steel braces. However, there is a lack of sufficient guidance on detailing and
dimensioning of the gusset plates in CBF structures, primarily due to absence of relevant data in
the literature. To regulate the design process, Eurocode 8 (EC8) (CEN 2004) suggests detailing of
gusset plates in accordance with available methods or accompanied with experimental evidence in
the absence of a suitable method. Physical laboratory tests can be time consuming, difficult and
costly. To avoid this hindrance, structural engineers typically use established design methods
available. Of them, the Standard Linear Clearance (SLC) method is one of the most commonly
used due to its simplicity and ease. This method, however, neglects important seismic effects due
to bracing steel members and frame actions both in plane and out-of-plane (Bjorhovde and
Chakrabarti 1985, Hu and Cheng 1987, Brown 1989, Cheng et al. 1994). As a result, the ability of
this method to develop and predict the performance of test models of CBF under realistic
earthquake loading was found to be poor and was below the required performance (Lehman et al.
2008). In addition, it has been established that the method leads to gusset plates that are typically
thick, large and not the most cost-effective solution. On the other hand, a balanced design
approach, in which the non-linear off-set rule or so called Elliptical Clearance (EC) method is used
to balance the tensile yield resistances of the structural steel brace and gusset plate (Lehman et al.
2008). Salient features of the two design philosophies are highlighted in Table 1.
It is widely accepted that CBF structures display attractive stiffness and strength under low to
moderate magnitude earthquakes but can perform less favourably in strong earthquakes compared
to some alternative structural systems (for example, moment resisting frames (MRF)) due to
possessing lower ductility and less energy absorption capabilities. With the aid of improved gusset
plate design, these deficiencies can be overcome in CBF structures. The relative stiffness of the
gusset plate compared to the brace member will influence the location and extent of buckling in
the system.
This paper provides essential information on the balanced design approach and investigates the
performance of this method relative to the conventional design approach, which is currently used
by structural engineers. Realistic gusset plate details and bracing members were used to generate
experimental data. This experimental data is utilised into a finite element framework to develop
models of test specimens. The paper is a part of a larger research project (Broderick et al. 2015,
Salawdeh et al. 2017, Goggins et al. 2018) that investigates the influence of various gusset plate
details with bracing members on the inelastic response of CBF structures under earthquakes. In
this paper, the inelastic cyclic behaviour of brace gusset-plate members is studied with the aid of
finite element models. In particular, the fatigue performance of the models in relation to
standardised seismic design methods for brace gusset-plate members are investigated. The models’
A study on detailing gusset plate and bracing members in concentrically braced frame structures 235
Table 1 Comparison of two gusset plate design philosophies

Elliptical Clearance (EC) design Standard Linear Clearance (SLC) design
✓ Both brace and gusset plate evolve system
✓ Braces mainly derive system performance.
performance.
✓ Gusset plate may experience inelastic strains during
✓ Gusset plate remains elastic during brace tension.
brace tension.
✓ No such sharing mechanisms exist for this system.
✓ The ww ratio characterises the balance of the brace
✓ During seismic action, brace usually suffer inelastic
and gusset plate yielding mechanisms.
deformations.
✓ The deformation of the system is shared by two
✓ Gusset plate has a single role in this system that is to
elements during seismic action.
accommodate brace end rotation.
✓ Gusset plate has dual role in this system, (i) it
accommodates brace end rotation, (ii) contributes in
ductility and energy dissipation capacity and
✓ Uneconomical, larger and thicker gusset plate typically
increases brace fatigue-life.
arise.
✓ Thinner and compact gusset plates provide reduced
structural cost of dissipative members.
✓ Large gusset plates remain stiff and strong during seismic
✓ Weak gusset plates reduce probability of hinge
action and facilitate formation of plastic hinges at joints
formation at joints formed by beam-column by
formed by beam and column non-dissipative members.
engaging themselves instead.
predictions are compared to physical quasi-static tests conducted in the laboratory. The
performance of the models is evaluated in terms of initial yield strength, initial-post buckling
compressive strength, ductility capacity, energy dissipation capacity, number of cycles to global
and local buckling and fatigue capacity.
2. Physical laboratory test
Hunt (2013) carried out quasi-static tests on full scale models of a single storey plane CBF
structure. Six brace gusset-plate specimens designed with realistic structural details were tested in
a displacement-controlled set-up. These specimens were tested as part of a CBF structure to
accurately represent realistic boundary conditions. The scope of the tests includes specimens that
have following specifications:
• Brace section size
o S40 40x40x2.5 (mm)
o S60 60x60x2.5 (mm)
• Connection configuration
o CA Gusset connection to beam and column flange
o CB Gusset connection to beam only
• Gusset plate design
o G1 Conventional design with Standard Linear Clearance (SLC)
o G2 Balanced design with Elliptical Clearance (EC)
The brace slenderness ratios were determined in accordance with Eurocode 3 (EC3) (CEN
1993) guidelines and covers the broad range of allowable steel braces permitted by EC8 (CEN
2004) for braced steel frames structures that is 𝜆̅ ≤ 2.0. Also, all braces were designated as Class
1 cross-sections as per EC3 (CEN 1993). Rectangular gusset plates were used with four cross-
section sizes, identified as (lh x lv x tp), of 285x240x8, 270x230x4, 265x240x8 and 250x230x4mm.
Table 2 Summary of test models (Hunt 2013)

Brace
Brace section Connection Gusset plate tp bw Ih Iv
Specimen
size type clearance (mm) (mm) (mm) (mm)
length, Lb ww
(mm)
S40-CA-G1 40x40x2.5 CA SLC (3tp) 8 155 285 240 2368 0.38
S40-CA-G2 40x40x2.5 CA EC (8tp) 4 155 270 230 2503 0.75
S40-CB-G1 40x40x2.5 CB SLC (3tp) 8 155 265 240 2368 0.38
S40-CB-G2 40x40x2.5 CB EC (8tp) 4 155 250 230 2503 0.75
S60-CA-G1 60x60x2.5 CA SLC (3tp) 8 175 285 240 2368 0.52
S60-CA-G2 60x60x2.5 CA EC (8tp) 4 175 270 230 2467 1.03
The specific details of the gusset plates are given in Table 2 and Fig. 1. All specimens are
classified as G1 or G2 as per the two design methods for the gusset plates; G1 refers to the
specimen which is designed as per the conventional design approach, in which the linear off-set
rule or so called Standard Linear Clearance (SLC) method was used to design the gusset plate and
G2 refers to the specimen which is designed as per the balanced design approach, in which the
non-linear off-set rule or so called Elliptical Clearance (EC) method is used (Fig. 1). The
difference between the design approaches lies in the fact that the former design approach supports
thicker and larger gusset plates to accommodate brace end rotation, while the latter supports
thinner and compact gusset plates to attain gusset plate yielding proportional to the yielding of
brace element in tension and, to accommodate brace end rotation in compression. This
proportional yielding, which is inter-related with yielding of bracing members, formed the basis of
a design approach known as the balanced approach (Lehman et al. 2008). In this approach, the
primary yielding mechanism is carried by brace action, while the secondary mechanisms by gusset
plate action. The ratio, which is defined as ww, of the yield resistances of the two mechanisms
determines the allowable proportional yielding between the two elements in accordance with
properties associated with the geometry and material strength of steel braces and gusset plates, as
R y,brace Fy Anet,brace
βww = (1)
R y,gusset Fy,gusset bw t p
where R y is the ratio of the actual yield stress to the nominal yield stress of steel material, Fy is the
yield capacity of brace, Anet,brace is the net brace area effective for gusset plate connection, bw
and t p are the width and thickness of the gusset plate, respectively and the Fy,gusset is the yield
capacity of gusset plate.
EC8 (CEN 2004) does not provide strong rules on the detailing of gusset plate connections
connecting steel braces with beams and columns in the CBF. Hence, two types of detailing
connections were used in tests, one with the brace gusset plate connected to the beam only,
denoted as CB and the other with the gusset plate connected to both the beam and column, denoted
as CA. In Table 2, specimens are identified by the size of the brace, either S40 or S60, gusset plate
connecting detail with beam and column as CA, or with only beam as CB and the gusset plate
design as G1 or G2, including sequential number 1, 2 etc. The variable “bw” is the Whitmore width,
which is a function of weld length and width of brace section. The corresponding ratio βww was
calculated in accordance with Lehman et al. (2008). All specimens were tested on a plane CBF
structure supported by pinned end connections, as shown in Fig. 2.
(a) S40-CA-G1 (b) S40-CA-G2 (c) S40-CB-G1
(d) S40-CB-G2 (e) S60-CA-G1 (f) S60-CA-G2

Fig. 1 Detailing of gusset plates in test models
Fig. 2 Test set-up, including frame details
2.1 Loading protocol
The specimens involved in this study are particularly designed for seismic regions where they
may undergo large inelastic strains causing fatigue in less than thirty cycles of loading. A loading
Fig. 3 Profile of loading protocol in accordance Fig. 4 Plot of hysteresis loops obtained from test
with ECCS (1986) guidelines model of S40-CB-G2-4 specimen
protocol that can develop such a large plastic strain in compression and tension periods was
required. In this context, the ECCS (ECCS 1986) recommended cyclic protocol was employed.
This protocol implies using one cycle of symmetric displacement magnitude of 0.25, 0.5, 0.75 and
1δy, follow by three cycles of displacement magnitudes of 2, 4, 6 δy, etc., where δy is the yield
displacement (Fig. 3). The loading was applied uniaxially at the joint formed by the beam and
column members with the aid of a hydraulic actuator (Fig. 2).
2.2 Hysteretic behaviour
A brief description on the response of specimen to ultra-low cycle fatigue loading is presented.
Fig. 4 shows that the response of the specimen is non-symmetric, but essentially elastic for initial
cycles of loading. The first major event is global buckling and the load corresponding to it is
labelled with initial buckling load.
After occurrence of global buckling, axial compressive strength degraded as plastic rotation
demand increases at mid-length brace tube. Further strength reduction occurs when identical
loading magnitude is applied as a result of out-of-the-plane deformation and to a lesser extent due
to the Baushinger effect. At each tensile excursion, permanent elongation occurs in the specimen,
which upon load reversal causes additional thrust in compression and, thereby, the effect of U-
shaped pinching occurs. The effect of U-shape pinching varies with the global slenderness ratio;
the lower is the ratio of global slenderness, the earlier is the local buckling and the resulting U-
shaped pinching effect in the brace buckling behaviour, provided that the section has a low width-
to-thickness ratio. All braces were deemed as Class 1 cross-sections (CEN 1993) with intermediate
global slenderness properties. Formation of local buckles in plastic hinges occurred initially at
corners of the brace tube mid-length under compression loading. On the tension side, the
maximum force resisted by specimen is marked with ultimate load in Fig. 4. The label “fatigue”
indicates the point where the specimens had fractured in the physical test.
3. Finite element modelling framework
The numerical models were developed in ABAQUS v.6.13 (ABAQUS 2013), which is a
general-purpose commercial finite element package. In order to give simplicity and ease to
modelling, two assumptions were made:

(i) Connection between the gusset plate and frame has infinite stiffness (fully rigid).
(ii) The testing/supportive frame has negligible influence on cyclic response of the brace
specimen.
3.1 Material model
Material modelling is a complex task when dealing with two components having different
nominal yield strengths and fabrication processes. As such, the gusset plates are typically
fabricated from hot-rolled carbon steel material, while the steel braces considered in this study
were cold-formed carbon steel. Cyclic loading was employed in this study, which has two
important effects on the behaviour of steel material; (i) cycling hardening of the material in tension
and (ii) reducing yield strength in compression (Baushinger effect). A material model capable of
representing two-cyclic strength degradation phenomenon was required. To this end, three material
models were reviewed;
(i) combined orthotropic kinematic hardening model,
(ii) isotropic hardening model,
(iii) combined nonlinear isotropic/kinematic hardening model.
Briefly, the combined orthotropic-kinematic hardening model tends to model elasticity of steel
by mechanics of orthotropic material in which properties of a material are investigated in three
dimensions. As such, the material response differs in three stress space. To solve this, the Hill
criterion [21] (expansion of von Mises criterion) is employed for predicting yield strength of
material in relation to properties of the material in the three axes. The condition of the two-
symmetry plane must be satisfied at the points of integration to meet orthotropic criterion.
Composites and/or brittle material typically exhibit such behaviour and, hence, it is suited to them.
Steel behaves similar to an isotropic-like material, in which properties remain unchanged in all
three-space configurations. Indeed, isotropy of steel can be modelled as orthotropic but with the
expense of nine elastic constants, which requires extensive knowledge and tedious amount of work
for accurate quantification. In conjunction, a kinematic hardening model is used to model shifting
of a yield surface while loading material cyclically. In this way, the two models work collectively
into a unified format. Upon application, it has produced an acceptable hysteresis loop of steel
braces (see, for example, Haddad et al. (2004)).
In contrast, the isotropic elastic and hardening model requires two elastic constants to model
the elasticity of a steel material. These elastic constants are the modulus of elasticity, E and the
Poisson ratio, . The model considers inelastic response of material as isotropic on account of
hardening. Hence, it neglects the contribution of Baushinger effect. Constitutively, a von Mises
criterion, which is differentiable over the yielding surface, along with associative plastic flow rule
is used to model hardening of steel. Upon application, hysteresis loops obtained from this model
were found to be moderately non-representative of physical test data (see, for example, English
and Goggins (2012)). The reason behind non-representation was associated to the limitation
(stationery expansion and contraction) of the model. This limitation was overcome by using a
combined nonlinear isotropic/kinematic hardening model. With this model, hardening of material
can be modelled in both isotropic and kinematic states using the work of Lemaitre and Chaboche
(1986) and the flow rule of plasticity. It has two important components; one is associated with
isotropic hardening capturing the size of the yield surface as a function of plastic strain, while the
second is associated with the kinematic component capturing shifting of a yield surface through
(b) FE model incorporated (c) FE model incorporated

(a) test model with Nip et al. (2010) with Fell (2008) parameters
parameters
Fig. 5 Plots of base shear versus lateral frame displacement
Table 3 Material model for steel braces and gusset plates

Cyclic hardening parameters Kinematic shifting parameters
Structural element 𝜎|𝑜 𝑄∞ 𝑏𝑖𝑠𝑜 𝜎|𝑜 𝐶𝑘𝑖𝑛 𝛾
MPa MPa MPa MPa
All steel braces Table 6 32.9 0.133 Table 3 87100 394
All gusset plates Table 6 87 1.5 Table 3 30200 188
back-stress variable known as the Baushinger variable. The two components work collectively and
should be defined as part of a bulk material property. Each component has three non-linear
parameters defined accordingly (ABAQUS 2013) as:
• Isotropic hardening parameters:
o 𝛔|𝐨 It is the yield stress at zero equivalent plastic strain
o 𝐂𝐤𝐢𝐧 It is the kinematic hardening parameter
o 𝛄 It determines the rate at which the back stresses vary as the plastic strain increases.
• Kinematic hardening parameters:
o 𝛔|𝐨 It is the yield stress at zero equivalent plastic strain
o 𝐛𝐢𝐬𝐨 It is the hardening parameter that defines the rate at which the size of the yield
surface changes as plastic strain increases
o 𝐐∞ It defines the maximum change in the size of the yield surface
Hunt (2013) carried out tests on coupons of gusset plates and steel braces under static tensile
loading. This set of coupon test were used to establish the yield strength. The modelling
parameters proposed by Nip et al. (2010) and Fell (2008) for hot-rolled carbon steel and cold-
formed carbon steel material in relation to combined nonlinear isotropic and kinematic hardening
model were assessed by simulating initial runs of the finite element model for specimen S40-CB-
G1-3. It was found that the Fell (2008) model did not accurately predict the strain hardening
response of the brace, which was accurately modelled by Nip et al. (2010), as shown in Fig. 5. It
was also found that by changing parameters of the Fell (2008) model and the Nip et al. (2010)
model for hot-rolled gusset-plate steel material, while keeping properties of the cold-formed
bracing members from Nip et al. (2010) model, the response of the FE model did not alter
significantly. Hence, the Nip et al. (2010) model was adopted with parameters corresponding to
40x40x2.5mm cold-formed brace steel member (Table 3). In the absence of measured data for the
(a) physical laboratory test model (b) shell element based model (c) solid element based model
Fig. 6 Plots of base shear versus lateral frame displacement for specimen S60-CA-G1-5
elastic model, a standardised E value equals to 190000 MPa and a Poisson ratio value equals to 0.3
were used in the models. The actual yield strengths were retained from tensile coupon tests (Hunt
2013).
3.2 Element and mesh size
Specimens involved in this study had a periodic buckling and yielding response under fully
reversed cyclic axial loading. An element capable of capturing the effects of geometric changes
and material degradation was required. In this context, two types of models were developed;
(i) Solid element-based model, which comprises three dimensional cubic solid elements
(C3D8R) and
(ii) Shell model, which comprises four node doubly curved elements, designated as S4R.
Beside conventional pros and cons of the two models, the biggest advantage of using the solid
model is the possibility of incorporating extended finite element method (XFEM) to its modelling
framework. With XFEM, a crack can be modelled independently of the requirement of a dense
mesh by methods of partition of unity and cohesive element superimposition. On the other hand,
shell element-based models are popular in structural analysis of problems involving bending
behaviour such as bending response of brace gusset-plate specimen.
Fig. 6 shows the responses of the two FE models and a physical laboratory test model under
cyclic axial loading. As evident, the solid element-based model fails to represent the actual
hysteretic response, which is more closely captured by the shell model. Moreover, the solid
element-based model consumes modal runtime three times greater than the shell model (13 hours)
when simulating an identical model using an Intel core 7 CPU. Also, the solid element-based
model had an occupied storage capacity greater than 2.4 times the storage capacity required by the
equivalent shell model. The poor performance of the solid element-based model is attributed to
several factors, such as the elements (i) are stiff, (ii) lock shear in transverse direction, (iii) possess
one integration point, (iv) consume greater model run time and (v) possess less degrees of freedom
compared with shell elements.
An element size of b/22 × b/22 was employed at the mid-length of brace, where b is the width
of the brace tube. This element size was extended in both directions from the mid-length equal to
1.5 times the dimension of the larger face of the brace section, h, to ensure that the mesh is
sufficiently large to accommodate the half wave length of the locally buckled shape, which was
Fig. 7 FE model of S40-CA-G1-1 specimen with mesh detailing in measuring unit (mm)
Table 4 Results of linear buckling analyses and their comparison to physical laboratory test and FE outputs
in relation to response of specimen under fully reversed cyclic axial loading
Specimen k 𝜆̅ Ncr Fc/Ncr Fc/Ncr Fc
(kN) Test FE FE/Test
S40-CA-G1-1 0.68 1.41 -63.08 0.61 0.88 1.44
S40-CA-G2-2 0.84 1.83 -37.39 0.57 0.62 1.09
S40-CB-G1-3 0.72 1.49 -56.76 0.74 0.92 1.24
S40-CB-G2-4 0.85 1.86 -36.13 0.90 0.91 1.00
S60-CA-G1-5 0.86 1.10 -143.89 0.79 0.86 1.09
S60-CA-G2-6 0.94 1.27 -109.66 0.82 0.92 1.12
Mean 0.74 0.85 1.16
Cov 0.16 0.12 0.12
approximately assumed equal to the larger face of the section (Nip et al. 2010, Nip et al. 2013).
The gusset plates at each end of the braces were modelled and meshed with elements of size lh/95
× lh/95, where lh is the width of the gusset plate (Table 2). This mesh size was extended 1.5 times
the dimension of the larger face of the brace tube beyond the gusset plate to accommodate the half
wavelength of the locally buckled shape in this region. An element size of b/8 × b/22 was used for
the uncritical remainder region of the model. This element size ensures that the model has a
sufficient number of elements to form continuum sinusoidal waves in regions that are non-critical
(Fig. 7).
In tests, the lower gusset plate was fixed to the test frame at the bottom edge only and the upper
gusset plate was connected to the frame in accordance with connecting details designed, such as
CA and CB (refer to Table 2). Identical boundary conditions are used in FE models to those in the
physical tests with fixed constrained end condition.
3.3 Linear buckling analysis

A linear buckling analysis was carried out to determine the global slenderness of specimens.
The specimens were provided with identical boundary conditions to that of the physical tests and
were then loaded in axial compression. The resulting elastic critical buckling load was estimated
corresponding to the second Eigen buckling mode shape. Table 4 contains the results of the linear
buckling analysis in terms of global slenderness ratio, 𝜆̅, defined in Eq. (2), (where A is the brace
tube cross-section, fy is yield strength of brace), effective buckling length factor, k, elastic critical
buckling load, Ncr and ratios of Ncr relative to the initial buckling loads, Fc, obtained from physical
tests and FE models of specimens when subjected to cyclic axial loading.
A fy
̅=√
λ (2)
Ncr
Brace gusset-plate specimens that were detailed as per conventional design have a k value of
between 0.68 and 0.86, while specimens that were detailed according to the balanced design
methodology have a k value between 0.84-0.94. In conventional design, the use of thicker and
larger gusset plate leads to a stiffer connection restricting the buckling wave length of specimen,
whereas the thinner and compact gusset plate that resulted from the balanced design approach
were closer to an idealised pin connected (that is, towards k = 1.0). Furthermore, braces connected
to both beam and column (i.e., type CA) have a smaller k value than those connected to the beam
only (Type CB) due to increased stiffness of the gusset plate by connecting to both beam and
column (Table 4).
In Table 4, it can be seen that the ratio of elastic critical buckling load (Ncr) to buckling load
values (Fc) from both the physical tests and FE models are below unity, indicating that geometrical
imperfection had influence on the global buckling response of specimen. The lowest ratio was
found for test specimen S40-CA-G1-1, which was due to deformations that were observed in the
lower channel section connecting the lower portion of the brace to the reaction frame during
testing. This channel section was replaced with a strengthened one for later tests. On the other
hand, a very good correlation is found between the results obtained from linear buckling analysis
and the numerical models with a mean value of 0.85 and coefficient of covariance (Cov) of 0.12.
This good agreement is partially due to the geometrical imperfection model employed.
3.4 Geometrical imperfection model
Geometrical imperfections are the largest source of uncertainty in building the numerical model
due to their non-uniform distribution that can be difficult to predict and measure. Design codes
typically take imperfections into account when determining the buckling capacity of steel
members under axial compressive loading. The magnitude of these imperfections are determined
based on a number of parameters, such as the cross-sectional shape and size, grade of steel and
manufacturing method (see, for example, EC3 (CEN 1993). Design codes, however, do not
provide guidelines on the magnitude and distribution of imperfections for building models in a FE
framework. Thus, Goggins (2004) intended on contributing to this limited data by proposing a
range of measured imperfections for application to simulation of single-story single-bay braced
steel frame. In addition, Salawdeh and Goggins (2013) went further by investigating imperfections
primarily on steel bracing members. However, the scope of the investigation was confined around
half-sine wave shape imperfection along with limited imperfection magnitudes. To overcome this
issue, Hassan et al. (2018) investigated imperfection from the broader aspect by assessing three of
the most applied methods of imperfection with a broader range of magnitudes of imperfection.
Their results showed that FE models gave best predictions of physical tests when they were
modelled with a combined equivalent notional lateral load and sinusoidal wave imperfection,
where the former imposed a global imperfection with an initial lateral deformation at mid-length
of the brace, 1, equal to L/1000 and the latter imposed local imperfections whose magnitude, 2,
was determined by an expression given by Gardner et al. (2010) for cold-formed steel
ω2 /t = 0.034 (fy /σcr )0.5 (3)
where t is the plate thickness, 𝑓𝑦 is the material yield stress and 𝜎𝑐𝑟 is the plate critical buckling
stress.
The global imperfection governs the magnitude of the initial buckling capacity of the specimen,
while the local imperfection improves the response of the structural element during softening stage.
Therefore, the model of Hassan et al. (2018) was used in this paper, in which the magnitude of
global imperfection,1 was set equal to Lb/1000 (to comply with buckling curves in EC3 (CEN
1993), where Lb is brace tube length, while the magnitudes of local imperfections, 2 were derived
using Eq. (3). To incorporate the imperfection model into the finite element model, a code was
written using the programming language python. The code details are given in Hassan et al. (2018).
3.5 Finite element solver
ABAQUS (2013) offers various solvers to solve non-linear structural response problems using
large deformation theory. Among them, the general-purpose static solver and the low cycle fatigue
solver were assessed. A low-cycle fatigue solver solved fatigue issues associated with linear
geometric changes, which can be solved with general static solver, including fatigue issues of
nonlinear geometric changes. Hence, a general static solver was used with activated non-linear
option to stimulate inelastic response of structural steel elements. A tabular format is employed to
apply cyclic loading incrementally. The increment size was set at 0.0085mm per cycle with the
number of increments varying according to the loading history, but with a maximum number of
increments set at 106. The maximum number of iterations was set at 20 iterations (default in
ABAQUS is 5) to deal with finding solutions during non-linearity behaviour of the system,
particularly in the stages of brace buckling and yielding. Buckling and yielding of braces develop
plastic hinges at corners of the mid-length brace tube section, which can form cracks upon
maturity, leading to a reduced cross-section capacity and ultimate failure.
3.6 Fatigue-life model
As fatigue was not explicitly captured in the models, a predictive model for fatigue was
required. A predictive model can be micro-mechanical, meso-scale or macro-scale models. All
forms of fatigue model consider appropriate constitutive parameters to justify the scale of damage
and the resulting fatigue. However, very limited models address earthquake type fatigue. Of them,
the Coffin and Manson model (Manson 1953, Coffin 1954), which is a reflective of the meso-scale
damage model and the Kavinda and Deierlein (2004) model, which is reflective of micro-
mechanic model, were assessed. It was found that, although, micro-mechanical models assess
fatigue very locally, they are complicated to implement into computational plasticity due to their
modelling requirement, such as fine mesh, characteristic length feature, etc. In contrast, meso-scale
models operate over coarser to fine meshes, requires less modelling parameters, saving cost and
time of post-processing. Therefore, the Coffin and Manson meso-scale model (Manson 1953,
Coffin 1954) was used in this study. Another reason of opting for this model is the fatigue
quantification with strain magnitudes, which makes the model competent for dealing with fatigue
in strain-controlled tests. The model provides the number of reversals to failure for a given plastic
strain, as defined by Eq. (4).
εp
=ε′f (2Nf )C (4)
2
where p/2 is the plastic strain amplitude, ε′f is the fatigue ductility coefficient which is the
material parameter that roughly indicates the strain amplitude at which one complete cycle on a
virgin material will cause failure, C is the fatigue ductility exponent which is the material
parameter which describes the sensitivity of the log of the total strain amplitude to the log of the
number of cycles to failure and Nf is the numbers of reversals to failure.
These reversals then linearly accumulated into a series of fatigue indexes by mean of a
cumulative damage rule, known as Miner’s rule (Palmgren 1924), which is defined as
n
ni
DI = ∑ (5)
Nfi
i=1
where Nfi is the number of reversals causing failure at a strain level of i and ni is the number of
reversals for which that strain level is applied.
The sum of these fatigue indexes indicates the degree of exclusive fatigue from 0 to 1. At a
fatigue index of 1, it is highly likely the fatigue failure has evolved in the specimen at a region
where the strain outputs had been extracted and the resulting cycle is marked as fatigue-life cycle.
For fatigue prediction, a fatigue evaluation criterion was required in tests and FE models. For
tests, the initiation of a crack in the corner region was marked as fracture occurrences (refer to
Table 5). This criterion was used in the FE models to extract strain outputs for fatigue prediction.
In FE models, however, it was unclear about the extent of the plastic hinge and the point of
integration to be used for fatigue prediction as an accurate mean of assessment. In this context, an
assessment methodology outlined by Nip et al. (2013) was used. They carried out a detailed study
on strain distribution across regions surrounded by plastic hinges, such as brace tube mid-length
section. Their study found that an area covering the corner and one-eighth of the buckling
wavelength is suitable for fatigue prediction. To predict fatigue, section positive (SPOS)
(ABAQUS 2013) point of integration of a shell element can be used for abstracting strain outputs
from the FE models as it carries the influence of local buckling, but is less sensitive to mesh size.
4. Modelling validation
Validation of the numerical models under fully reversed cyclic axial loading was carried using
measurement data from the physical laboratory tests described in Section 2. Table 5 contains the
results of the physical tests in terms of yield tensile load, Fy, ultimate tensile load, Fu, initial
buckling load, Fc, energy dissipation capacity, Wtotal and numbers of cycles to global and local
buckling including fatigue. These results are obtained by transforming the measured storey shear
Table 5 Summary of test measurements

Specimen b/t 𝜆̅ Fy Fu Fc Wtotal Number of cycles to
KN KN KN KN.m Global buckling Local buckling corner opening fracture
S40-CA-G1-1 16 1.41 118 182.4 -38.4 38.9 3 10 16 17
S40-CA-G2-2* 16 1.83 152 215.7 -21.4 52.2 3 - 28 29
S40-CB-G1-3 16 1.49 175 200.1 -41.9 36.8 3 10 21 23
S40-CB-G2-4 16 1.86 188 209.8 -32.7 61.7 3 10 28 29
S60-CA-G1-5 24 1.10 196 211.7 -113.4 18.8 4 6 10 11
S60-CA-G2-6 24 1.27 190 204.5 -90.1 20.1 4 7 10 11
*survive maximum ductility demand
Table 6 Summary of FE results

Initial geometric Ratio of FE to physical test
Specimen ID fy,g fy Number of cycles to
imperfection results
N/m N/m Global Local Global Local Fractu
Fy Fu Fc Wtotal
m2 m2 1 2 Eq.(3) buckling buckling re
S40-CA-G1-
271 351 Lb/1000 3.6%t 1.45 1.13 1.44 1.04 2 11 16
1
S40-CA-G2-
311 351 Lb /1000 3.8%t 0.99 0.95 1.09 0.87 3 28 29
2*
S40-CB-G1-
284 351 Lb /1000 3.6%t 0.98 1.06 1.24 0.80 3 17 19
3
S40-CB-G2-
302 351 Lb /1000 3.8%t 0.89 0.94 1.00 0.86 2 23 26
4
S60-CA-G1-
279 336 Lb /1000 2.2%t 1.30 1.45 1.09 1.49 3 7 9
5
S60-CA-G2-
315 336 Lb /1000 2.4%t 1.21 1.13 1.12 1.13 2 8 10
6
Mean 1.14 1.11 1.16 1.03
Cov 0.18 0.15 0.12 0.23
*survive maximum ductility demand in test, “t” is brace tube thickness and Lb is brace tube length
force and lateral frame displacement into brace axial force and elongation. For convenience, the
actual plots of storey shear force and lateral frame displacement of each specimen are presented in
Figs. 8-13. Moreover, Table 6 contains the predictions from the numerical model and
corresponding comparisons with physical test measurements. It also contains essential data used in
modelling the gusset plates and bracing members.
4.1 Yield strength, initial buckling and post buckling loads
The axial load to initially yield the specimens was approximated using the 1/10th slope method.
In this method, yield values correspond to the point where the tangent at the origin and the tangent
that has a slope of one-tenth the tangent at the origin intersect. From Table 6, while comparing the
initial yield load values of FE models to those obtained from physical tests, it was found that FE
models gave a good prediction of initial yield loads with the mean being 1.14 and Cov 0.18 for all
specimens investigated. On the other hand, the ultimate load, defined as the largest load a
(a) Physical test model (b) FE model

Fig. 8 Plots of storey shear versus lateral frame displacement of specimen S40-CA-G1-1.

Fig. 9 Plots of storey shear versus lateral frame displacement of specimen S40-CA-G2-2.

Fig. 10 Plots of storey shear versus lateral frame displacement of specimen S40-CB-G1-3.
specimen carried in tension, was found from the FE models to be on average 1.11 times the
physical test results with a Cov on average of 0.15 for all specimens studied.
The initial buckling load (Fc) is typically the highest load a specimen resists in compression. As

Fig. 11 Plots of storey shear versus lateral frame displacement of specimen S40-CB-G2-4

Fig. 12 Plots of storey shear versus lateral frame displacement of specimen S60-CA-G1-5

Fig. 13 Plots of storey shear versus lateral frame displacement of specimen S60-CA-G2-6
only Class 1 sections are considered in this study, initial buckling typically occurs in the same
cycle in which the specimen yields for stockier members, whereas it happens in earlier cycles for
more slender members due to elastic buckling. The results of the physical test model show that
Fig. 14 Plot of normalised initial-buckling loads versus normalised global slenderness ratios for all
specimens, including buckling curve from EC3 (CEN 1993)
specimens designed using a balanced design approach carried lower buckling loads than
specimens based on conventional design. This difference is attributable to factors influencing
gusset plate detailing and slenderness of the bracing members (Salawdeh et al. 2017). While
comparing FE models to test models, it is found that the FE models gave very good predictions of
initial buckling loads, with a mean ratio of 1.16 and Cov of 0.12 for all specimens investigated
(Table 6). However, greatest difference arises for the first test specimen S40-CA-G1-1, primarily
due to yielding of the supportive channel section during the physical test, which was replaced with
a strengthened channel for other tests. FE results show that braces with larger gusset plate
connections attain greater buckling load but delivered less ductile response subsequently (refer to
Table 6). In contrast, braces with thinner and more compact gusset plates attain lower buckling
load, but yield improved ductile response (refer to Table 6). These observations are also evident in
the physical test results (refer to Table 5).
Fig. 14 presents the plot of initial buckling loads, Fc normalised by brace yield strength, fy
times brace cross-section, A, against normalised global slenderness obtained in accordance with
EC3 (CEN 1993). These buckling loads are compared with the buckling curves given by EC3
(CEN 1993), as
(6)
where φ = 0.5[1 + α(λ̅ − 0.2) + λ̅2 ] and Fc is the design buckling load, α is the generalised
imperfection factor with a value of 0.49 for cold-formed sections.
As expected, in general, the EC3 (CEN 1993) design buckling curves gave a lower bound
(conservative design), although they significantly underestimated the buckling loads obtained from
both the physical tests and FE models for specimens with global slenderness ratio between 1.1 and
1.27. This conservatism is due to the generalised imperfection factor to cover various fabrication
defects, such as imperfection and residual stresses. In addition, the initial buckling loads of these
slenderness models are close to their yield capacity. This suggests that they had elastic plastic
buckling behaviour. The amount of residual stresses and imperfection is typically observed higher
in cold-formed sections than hot-rolled sections, primarily due to further post-fabrication process
that is cold-working. The larger imperfection factor for cold-formed curves is inferred to this
reason.
Tremblay (2002) proposed a model for evaluating of post-buckling load of Class 1 brace
specimens subjected to symmetrical loading. Nonlinear regression analysis of the measured post-
buckling loads for cold-formed sections resulted in the proposal of the predictive expressions
given by Eq. (7); as
−c Fc
̅
a + b1 λ = (7)
A fy
where a, b1 and c are the model’s parameters which vary in value with respect to the ductility of
interest.
Goggins (2004) proposed a similar model by using an independent set of data. The data was
also associated to the cold-formed Class 1 brace tubes and, generated experimentally through
symmetric loading protocol applied to tubular members. Unlike Tremblay, Goggins (2004) used
an upper envelope to develop the expression given in Eq. (8)
Fc
̅ −c =
b1 λ (8)
A fy
Fig. 15 presents a comparison of the post-buckling loads obtained from FE and tests models,
along with predictions of Tremblay (2002) and Goggins (2004). The values of the parameters in
the Tremblay model were 2.5 and 3.1 for b1 parameter, 3.9 and 5.2 for a and 5.9 and 3.5 for c, for
ductility 2 and 4, respectively. In the Goggins model, these values were 5.9 and 2.2 for b1
parameter and 3.6 and 2.9 for c parameter in accordance with ductility 2 and 4, respectively. A
post-buckling load relative to the first cycle of the displacement demand was used. As depicted,
Tremblay’s model consistently underestimates predictions for ductility levels of 2 and 4. In
contrast, Goggins’ model gave a relatively better prediction for ductility level of 2, although, in
general, underestimated the post-buckling capacity at both a ductility of 2 and 4. This is due to
number of possible reasons, including, (i) Tremblay (2002) employed a lower bound definition of
the post-buckling compressive strength in Eq. (7) and (ii) for identical slenderness range, the
models of Tremblay (2002) and Goggins (2004) did not effectively cover the data of the test
specimens.
In Fig. 16 (a)-(b) normalised post-buckling loads are plotted against various ductility levels for
both physically tested and numerical models. As depicted, post-buckling loading capacities
degraded as the demand of axial ductility increases for all specimens, except of S40-CA-G1-1 and
S40-CA-G2-2. A possible reason would be the influence of frame action due to the CA type
connection employed in tests 1 and 2. In tests 5 and 6, the influence of the CA connection is never
fully realised because the higher cross-section slenderness ratio (from the S60 section) is dominant
and causes fracture at earlier post buckling cycles. In general, the degradation in strength is the
result of hinge formation at brace mid-length tube section. However, degradation in the post-
buckling capacity varies with the slenderness ratios. Specimens with larger global slenderness may
not suffer identical compressive strength degradation than the specimens with lower global
slenderness ratio, provided that the local slenderness is identical.
(a) Ductility 2 (b) Ductility 4

Fig. 15 Plots of normalised post-buckling loads versus normalised slenderness ratio, including
predictions from Tremblay (2002) and Goggins (2004)
(a) Physical test models (b) FE models

Fig. 16 Post-buckling compressive strength of specimens
While comparing the post-buckling loading capacity of the physical test and FE models, it is
found that a good correlation exists between the two models, except for where the frame had
unintended deformation in the lower channel section (S40-CA-G1-1).
4.2 Lateral deformation
In the system under consideration, lateral deformation occurs as a result of out-of-plane

bending. For design considerations, it is important to be able to predict the deformation of brace
members in the normal direction of the loading. For example, it can be used to determine the
clearance required between bracing members and non-structural members like glass, cladding, etc
for safety purposes. In tests and FE simulations, all models deformed in the out-of-the-plane
direction. However, in the absence of measured data, the results of the lateral deformation from the
FE models are only presented.
Tremblay (2002) proposed a simple model for determining lateral displacements as a function
of applied displacement, c and brace tube length, Lb, as
∆= 0.7√δc Lb (9)
where ∆ is the lateral deformation, c is the applied displacement, Lb is the brace tube length and
0.7 is a factor that represents all end conditions implicitly.
Hassan et al. (2015) modified the model proposed by Tremblay by incorporating an
imperfection factor. A least square regression analysis of the analytical data resulted in the
proposal of the predictive expression given by Eq. (10).
1
∆= . . 0.7√δc Lb (10)
Lb
0.97
= (11)
β0.96
where  is a model parameter and should be obtained from Eq. (11). In Eq. (11),  is the ratio of
global imperfection magnitude, 1 and brace tube length, Lb. Eq. (10) is valid for global
imperfection magnitudes ranging from Lb /5 to Lb /2000.
Fig. 17 presents results of the FE models and those obtained from Eqs. (9) and (10). The FE
results correspond to the models of Standard Linear Clearance method are shown collectively in
Fig. 17(a). In Fig. 17(b), FE results corresponds to Elliptical Clearance method are plotted. As
depicted, the model of Hassan et al. (2015) gave a very good prediction of lateral deformations. In
contrast, the model of Tremblay (2002) overestimates FE predictions significantly, primarily due
to the fact that model yields lateral displacement relative to the mechanisms formed by the plastic
hinges at ultimate load level.
4.3 Number of cycles to global buckling and local buckling
The instant of global buckling is typically defined by the largest load attained in compression.
The identification of the cycle during which buckling occurs is easy identifiable for sections
having intermediate to lower slenderness ratios, as these sections have pronounced global buckling
following local buckling and hinge formation. It is often difficult to identify the onset of global
buckling of specimens with larger global slenderness ratios due to their uniform compressive
response. For these specimens, care is required when identifying global buckling loads.
From Table 5, it is found that slender specimens experienced global buckling in the 3rd cycle of
loading (i.e., with amplitude of 0.75δy, where δy is the expected yield displacement), except for
specimens S60-CA-G1-5 and S60-CA-G2-6. These specimens experienced global buckling in the
4th cycle of loading, which is a yield cycle. It is also found that the FE models gave a reasonable
prediction of the instance of global buckling when compared to observations from the physical
tests (Table 6).
The instant of local buckling can be quantified by two methods from strain readings; (i) by
examining strains in the longitudinal direction and (ii) by examining strain rate outputs. In the
former method, the magnitude of strains in the longitudinal direction are used to identify the
(a) G1 type gusset plate (b) G2 type gusset plate

Fig. 17 Plots of lateral deformation of braces against axial displacement magnitude for FE models in
conjunction with plots of Tremblay (2002) and Hassan et al. (2015).
Fig. 18 Location and direction of strain gauges used in the experiments
instant of local buckling. In the latter method the instant of local buckling is indicated by the time
when the highest strain rate occurs.
Strain gauges applied to the centre of the outer faces of the specimen at mid-length in the
longitudinal direction, identified as SG10 and SG11 in Fig. 18, were used to quantify the instant of
local buckling in physical tests. As depicted in Fig. 19(a), SG10 and SG11 have approximately
identical strains before the occurrences of local buckling in compression. However, at the instant
of local buckling, SG10 displays relatively higher compressive strain readings, primarily due to
being located on the outside face of the tube. In subsequent cycles, there is a shift in recordings of
SG10 and SG11 from compressive to tensile. This shift is believed to be because local buckling
occurs slightly away from the mid-length brace tube section and that the tensile strain is the result
of formation of convex rise next to concave slope formed by the local buckling.
Thus, it may not be reliable to use strain gauge recordings in physical tests for quantification of
local buckling. To overcome this issue, recordings of strain gauges and loading histories were
investigated together (Fig. 19). A sudden change in the loading rate, identified by a sharp
narrowness in the load-time plot, can be seen in the compression cycle of the loading when local
buckling occurs in test specimen. The presence of this pinching in the time-varying loading plot
from the test data is consistent with that observed in the FE models (Fig. 20). It can be concluded
that local buckling occurs near mid-length of the brace tube section. However, as stain gauges are
(a) S60-CA-G1-5 (b) S60-CA-G2-6

Fig. 19 Plots of measured time-varying longitudinal strains and brace axial force for physical test
specimens
placed in discrete points that didn’t necessarily coincide with where local buckling was present,
restricted the exact identification of the cycles causing local buckling in physical tests. For future
experiments, real-time digital image correlation (DIC) offers the potential to overcome this issue.
On the other hand, Fig. 19(b) shows a completely different picture of recordings of the two
strain gauges for specimen S60-CA-G2-6. Such that, the strain values from SG10 and SG11
deviated from each other following the occurrence of global buckling. This effect is due to the
occurrence of global buckling in the negative direction in relation to the location of SG10 strain
gauge. As a result, SG10 records positive (tensile) strains for the entire loading history. This SG10
recording indicates that specimen never regained its original position after experiencing inelastic
global buckling in compression. On the other hand, recording of SG11 shows that it has identical
pattern of the strain gauge recording obtained on SG11 of specimen S60-CA-G1-5 (Fig. 19(a)).
However, it fails in compression as a result of concave deformation next to convex rise formed by
local buckling of the upper face at mid-length of the brace tube section.
Fig. 20 presents time-varying lateral strain, strain rate and brace axial forces plots for two FE
models identified as (a)-(b) S60-CA-G1-5 and (c)-(d) S60-CA-G2-6. These strain recordings were
abstracted from shell elements located at mid-length and mid-section of the braces. As seen in Fig.
20, the compressive strain rate is the largest at the instance where there is a sudden change in the
unloading rate during the compression cycle of the loading, resulting in a sharp narrowness in the
load-time plot, which indicates that local buckling has occurred at this instance in the FE models.
Thus, observing the loading rate can successfully be used to identify the instance that local
buckling first occurs, as well as identifying the instant of global buckling. Furthermore, a shell
element has at least three integration points, in which one point records stress/strain history of the
outer integration point, while another records histories for the inner integration point. The outer
integration point experienced relatively larger compressive strains than inner integration point, due
to local buckling.
From Table 5, it is found that all test specimens buckled locally in the second and fourth
ductility level, except for the specimen identified as S40-CA-G2-2. The visual inspection of this
test specimen had indicated presence of limited local buckling (Hunt 2013). It was, however,
difficult to quantify the intensity of this limited local buckling because; (i) it had buckling in the
negative direction making it difficult to locate local buckling using available strain gauge
measurements (SG10), (ii) the working range of strain gauges was limited, (iii) the response of
(a) S60-CA-G1-5 (b) S60-CA-G2-6
(c) S60-CA-G1-5 (d) S60-CA-G2-6

Fig. 20 Plots of (a-b) time-varying longitudinal strain (primary) and strain rate (secondary) from FE
models and (c-d) plots of time-varying brace axial force (primary) and strain rate (secondary) from FE
models
specimen was highly uniform in bending, making difficult to locate local buckling using loading
history and (iv) the specimen survived maximum ductility without failure meaning it is highly
likely the failure mode of specimen is global buckling.
As expected, the occurrences of the local buckling significantly degraded the compressive
strength of specimens under axial cyclic loading. However, physical tested specimens lasted
between 4 and 19 cycles after local buckling before fracture (Table 5), whereas the FE models
predicted fracture after 1 to 5 additional cycles after local buckling (Table 6). The specimens are
classified as Class 1 sections according to EC3(CEN 1993), which means those tubes have a
tendency to form plastic hinges with the rotation capacity without reduction of the resistance
before yielding. EC8 (CEN 2004) implies that Class 1 cross-section tubes should be utilised when
a high dissipative structural response is desired, whereas Class 2 and Class 3 sections can be used
when a low dissipative structural response is suitable. For these sections, local buckling is the limit
state to quantify the member capacity. In such cases, quantification of the local buckling becomes
important crucially in the CBF design.
4.4 System behaviour
A crucial criterion for accurate prediction of fracture is reproducing the strain distribution
correctly in the FE models. The FE and physical test strain measurements over the loading history
of specimen S40-CA-G1-1 and S40-CA-G2-2 are compared and presented in Figs. 21 and 22,
respectively. The FE strain output was extracted from similar locations as the strain gauges (Fig.
18). As shown, FE models acceptably reproduced strains measured in the physical tests. The few
discrepancies are (i) due strains in the physical tests exceeding the working range of the strain
gauges and (ii) may be attributed to the framing imperfection that may arise during testing set-up
and experimentation.
While investigating strain recordings of SG5, it is found that the gusset plate of specimen S40-
CA-G1-1 remained elastic throughout the loading, while the strains in the gusset plate of specimen
S40-CA-G2-2 exceeded the elastic range (Figs. 21-22). This difference is primarily dedicated to
the design philosophies and that the balanced design approach used in designing S40-CA-G2-2
allows plastic yielding and, thus, tends to contribute to the system ductility. In contrast, no such
yielding is permitted in conventional design, which was used for specimen S40-CA-G1-1.
To examine the component behaviour of a system, similar strain recordings are used. In Fig. 21,
specimen S40-CA-G1-1 experienced positive tensile strain at the gusset plate, while it experienced
negative compressive strain over the front and top surfaces of the mid-length brace tube section
when the time is 1000 sec. A gusset plate tends to bend in the opposite direction relative to the
buckling mode of steel braces to accommodate brace end rotation at the connected ends. This is
the reason behind the positive (tensile) strain in the gusset plate. The negative strain recording at
the brace tube mid-length is the result of formation of the local buckling; however, the formation
of convex rise as a result of concave slope tends to shift the behaviour of the strain gauge
recording from compressive to tensile.
Fig. 22 shows a completely different picture in the strain gauge recordings of S40-CA-G2-2
specimen compared to those found in tests of S40-CA-G1-1 specimen. The specimen S40-CA-G2-
2 was buckled in the negative direction out-of-the-plane of the testing frame. As a result, SG10
records tensile deformation in the opposite face of the locally buckled brace tube. The strain
recording of the SG11 shows no effect of this negative buckling and it has identical pattern of the
strain recording obtained in specimen S40-CA-G1-1 (Fig. 21(a)). The strain gauge recording SG5
of the gusset plate shows positive (tensile) strains, primarily due to bi-axial inelastic plate buckling
in the opposite direction.
4.5 Fatigue prediction
On the basis of cross-section classification, the design limit of the specimens was found to be
fatigue-life. A crack usually occurred at mid-length of the brace a few cycles after the first onset of
local buckling. Phenomenologically, the occurrence of local buckling develops high compressive
strains at upper and lower corners of the brace tube section at mid-length and that upon load
reversal transform into tensile stresses, leading to corner opening of the brace tube at mid-length.
The stress transformation from compressive to tensile is not purely axial; there is highly likely that
a complex tri-axial stress phenomenon occurs causing the formation of plastic hinges. This is the
fatigue phenomenon when the scale of assessment is macro scopic or mesoscopic.
Micromechanically, a fatigue is the result of micro-cycling degradation process of a material
matrix (Rudland and Brust 1997). In the case of braces, material matrix could be assumed to be the
region surrounding the plastic hinges as it is where the crack first emerges in tension. Rudland and
Brust (1997) found that the micro-cycling degradation process has two important phenomenon (i)
crack tip sharpening and (ii) void sharpening. A crack tip sharpened in the compression period of
the loading, which upon load reversal caused a micro-crack to join the main crack produced in
(a) gusset plate (b) brace tube mid-length front face (c) brace tube mid-length top face
Fig. 21 Plots of strain recording versus time magnitude obtained from FE and test models for specimen of
S40-CA-G1-1 at various locations, as identified in Fig. 18
(a) gusset plate (b) brace tube mid-length front face (c) brace tube mid-length top face
Fig. 22 Plots of strain recording versus time magnitude obtained from FE and test models for specimen of
S40-CA-G2-2 at various locations, as identified in Fig. 18
previous tension period to the void ahead of the crack. Crack-tip sharpening was found to be a
crucial factor in the evolution of fatigue crack growth, specifically, for symmetric loading.
In Table 5, the number of cycles of loading to fracture due to fatigue is given. On the other
hand, fatigue predictions from the FE models are presented in Table 6. A good agreement is found
between the two models. The models with lower normalised global slenderness survive a lower
number of cycles compared to models with higher global slenderness, which can sustain up to
thirty cycles. This difference is due to the central role of the local buckling, which develops earlier
in specimens that have lower global slenderness. Moreover, the specimens with balanced design
attain a greater number of fatigue cycles irrespective of the connection details. Thus, it can be
concluded that a balanced design improves fatigue lives of the bracing members by incorporating
gusset plate yielding.
4.6 Ultimate ductility and energy dissipation capacity
A key input into traditional seismic design approaches is a ductility parameter associated with
the framing system based on the expected performance of the bracing members. Accordingly,
when the response of bracing members is limited to the elastic range, the ductility class is
classified as low (CEN 2004). In contrast, if the response is highly ductile (up to plastic hinge
formation, which is Class 1 cross-sections in EC3 (CEN 1993)), the ductility class is classified as
medium to high (CEN 2004); further stringent guidelines follow for high dissipative structural
response. The bracing members involved in this study are classified as Class 1 cross-section
according to EC3 (CEN 1993). Consequently, brace ductility values, μ, are determined relative to
the maximum brace elongation at failure,  normalised by the axial elongation of the brace at
yield, y. This change may be an increase in length (elongation under tension) or a reduction in
length (shortening under compression) and includes the effects of axial deformations in the brace
tube length and gusset plate strains. The brace yield displacement, δy, is obtained as the product of
the length of the unstiffened brace tube (actual length – 2 times weld length, where weld length is
100mm) and its characteristic yield strain (εy = fy/E), where fy is the brace yield strength (Table
6) and E is the modulus of elasticity assumed as 190,000 N/mm2. The observed values of μ vary
from below 6 to above 15, as illustrated in Fig. 23.
Design guidelines do not provide guidance on obtaining ductility capacities of the dissipative
elements. However, Tremblay (2002) proposed a simple model, in which the total ductility
reached at fracture, μT, is related only to the normalised slenderness parameter, λ̅, as
Cold-formed carbon steel
𝜇T = 2.4 + 8.3λ̅ (13)

On the other hand, Goggins et al. (2006) working independently on test data, proposed a
characterisation of ductility capacity, μ (the change in brace length, δ, at failure normalised by the
brace yield displacement, δy) based on the normalised slenderness parameters, λ̅ and width to
thickness ratio (b/t) as,
𝜇 = −0.68 + 26.2λ̅ (14)
𝜇 = 29.1 − 1.07(b/t) (15)

Nip et al. (2010), using similar definition of ductility capacity used by Goggins et al. (2006),
developed improved expressions for three steel materials, as
Hot-rolled carbon steel
𝜇 = 3.69 + 6.97λ̅ − 0.05(b/tε) − 0.19(λ̅)(b/tε) (16)

𝜇 = 6.45 + 2.28λ̅ − 0.11(b/tε) − 0.06(λ̅)(b/tε) (17)

Cold-formed stainless steel
𝜇 = −3.42 + 19.86λ̅ + 0.21(b/tε) − 0.64(λ̅)(b/tε) (18)

Table 7 compares the peak displacement ductility demand observed in each test in which brace
fracture occurred to the ductility capacities calculated using the equations proposed by Tremblay
(2002) and Nip et al (2010). The peak ductility demand observed in each cyclic test exceeded the
ductility capacity, μ, calculated using Eq. (17) in all cases, with the experimental ductility
demands ranging from 31% to 122% of the predicted capacity values. On the other hand, the
application of Eq. (13) results in predictions of ductility capacity, μT, that generally underestimate
the peak ductility demand observed prior to brace fracture. Specifically, the experimental ductility
Table 7 Measured and predicted ductility capacities

 T T 
Specimen y (mm)Test (mm)FE (mm) =testy =FEy FE /test
Eq. (17) (Test) (FE) Eq. (17)
S40-CA-G1-1 4.0 58.4 51.2 14.6 12.8 6.6 30.0 25.0 14.1 0.9
S40-CA-G2-2 4.3 51.3 50.4 12.0 11.8 7.1 23.3 23.3 17.6 1.0
S40-CB-G1-3 4.0 48.1 41.0 12.0 10.2 6.7 22.5 17.5 14.8 0.9
S40-CB-G2-4 4.3 63.0 60.1 14.8 14.1 7.1 27.9 22.6 17.8 1.0
S60-CA-G1-5 3.8 28.6 24.5 7.5 6.4 4.7 12.4 12.4 11.5 0.9
S60-CA-G2-6 4.0 25.6 24.1 6.4 6.0 4.9 12.3 12.3 12.9 0.9
Mean 0.91
Cov 0.06
(a) (b)
Fig. 23 Comparison of observed peak displacement ductility demand in test to predicted brace ductility
capacity in FE models, along with predictions of (a) Nip et al. (2010) and (b) Tremblay (2002). Note
=y for Nip et al. (2010) and T =(elong+shor)y for Tremblay (2002)
demand in the failure test was greater than the value predicted by Eq. (13) for experiments with
brace specimens of 40x40x2.5mm, with the predictions ranging between -32% and -112% of the
experimental value.
These data are illustrated in Fig. 23, which indicates that Eq. (17) gave mostly conservative
predictions of brace ductility capacity. This conservatism may be due to the limited test data that
were used to develop the expression. In contrast, Fig. 23(b) indicates that Eq. (13) significantly
over predicts brace ductility capacity for specimens with 40x40x2.5mm tubes. This is likely due to
scattering of the data used to develop that expression.
EC8 (CEN 2004) allows concentric steel braces up to global slenderness, 𝜆̅ between 1.3 to 2.0
and EC3 (CEN 1993) allows cross-section with limiting width to thickness (b/t) ratio less than and
equals to 72 for Class 1 braces subjected to web bending, where =√235/fy, fy is the nominal
yield strength, which is 275MPa herein. Thus, considering these limits imposed by EC8 (CEN
2004), it is possible to use Eqs. (13) and (17) to determine a range of allowable peak displacement
ductility for design conveniences. With the maximum allowable global slenderness and the
maximum allowable local slenderness, Nip et al. (2010) gave an absolute ductility of 6, while it is
Fig. 24 Ductility capacity of the physical test models and FE models
Fig. 25 Energy dissipation capacity of the physical test models and FE models
4 when the minimum allowable global slenderness (𝜆̅ =1.3) was used with a constant magnitude of
maximum allowable local slenderness. On the other hand, these values were 13 and 19 using
Tremblay’s model (Tremblay 2002) for the minimum and maximum allowable slenderness
(global), respectively.
Fig. 24 compares peak displacement ductility observed in the FE and physical test models. It is
found that the FE model gave a good prediction of the measured ductility capacity with a mean of
0.91 and Cov of 0.06.
Energy dissipation capacity is another important measure of structural capacity from the
seismic design aspect. It can be defined as an area enclosed by the hysteresis loop under axial
compression and tension loading. The effect of the energy dissipation capacity is implicitly
accounted in the behaviour q factor in EC8 (CEN 2004) to transform the elastic response spectrum
(a) Occurrence of local (b) Corner opening and (c) Broken brace (d) Gusset plate
buckling at brace tube fatigue occurrences (at tensile tube from mid- bent shape after
mid-length (at period 10th) length (after test (after test)
compressive period 6th) test)
Fig. 26 Snapshots of specimen S60-CA-G1-5 during testing programme
(d) Gusset
(a) Occurrence of local (b) Formation of plastic plate bent
(c) Damage occurrence (after
buckling (at compressive hinges at upper and lower shape after
simulation)
period 7th) compressive corners (at test (after
tensile period 9th) simulation)
Fig. 27 Snapshots of FE model S60-CA-G1-5 during and after simulation
into a design spectrum to avoid the need to conduct non-linear structural analysis. Fig. 25 presents
a comparison of the energy dissipation capacity between physical test and FE models. While
comparing predicted to test results, it is found that the FE model gave a close prediction of the
brace energy dissipation capacity. Thus, it can be concluded that the FE model provides a
reasonable prediction of the test results, validating the accuracy of the model.
4.7 Failure modes
As an example, Fig. 26 shows the experimentally obtained deformed structural behaviour of

specimen S60-CA-G1-5 under fully reversed cyclic axial loading. Occurrence of local buckling
forms plastic hinges at upper and lower compressive corners, which triggers cracking at these
locations and evolve a fatigue upon maturity. In physical experiments, however, without the use of
advanced monitoring equipment such as digital image correlation, it is difficult to inspect in detail
the life-cycle degradation process of material from initial loading to damage unless analysing it
through finite element modelling. Fig. 27 shows the deformed specimen and Von Mises stresses
for the FE model of the specimen displayed in Fig. 26. It can be observed that substantial inelastic
deformation occurred by local buckling, followed by formation of plastic hinges and finally
damage evolution in the corner regions, which is the typical ductile steel behaviour of cold-formed
structural steel square and rectangular hollow sections to failure. This generally follows the visual
development of local buckling in the experimental results.
5. Conclusions
The inelastic cyclic response of gusset plates and bracing members in CBFs is asymmetrical
and complex with periodic cyclic buckling and yielding, leading to hinge formation and fatigue.
These response features are difficult to predict under dynamic loading conditions. Previous
research and design guidelines that established the influence of brace slenderness (global and local)
on system ductility and energy dissipation capacity was principally based on the testing of steel
braces. Design codes are unable to provide clear guidelines on detailing of gusset plate connection
primarily due to lack of sufficient data on the response of gusset plates and bracing members with
practical structural details and loading conditions. Moreover, finite element models for simulating
the cyclic response of braced frames have not always been fully validated due to lack of sufficient
data on tests of full-scale models with realistic detailing of brace gusset-plate specimens in a plane
CBF structure.
To provide this essential data on the response of CBFs with practical end details and realistic
gusset plates, the results of six full scale tests on a plane CBF structure have been examined.
Physical measurements and observations from tests made an important contribution by allowing
straight forward assessment of gusset plate details, including assessment of comparative base shear
against lateral frame displacement and examination of bracing members’ fatigue lives under fully
reversed cyclic axial loading. The test specimens and testing set-up were more realistic than most
of the previous tests published in the literature, which are mainly either static monotonic tests or in
other cases, quasi-static tests on steel braces with idealised stiffened connection end conditions.
The CBF frame was subjected to uniaxial loading in accordance with loading protocol
recommended for the testing of structural steel specimens exposed to fatigue. The symmetric
loading regime employed in the tests simulates braces with identical levels of expected peak
deformation demand in tension and compression. The brace-gusset plate test specimens
consistently displayed alternating buckling and yielding producing hysteresis loops at culminating
in fatigue at brace tube mid length. The recorded responses of gusset plate and steel braces were
consistent with those that have been recorded in quasi-static cyclic testing of steel braces.
Particularly, the final failure was confined to the thin-walled tubular section themselves, even in
tests with yielding of the gusset plate, validating conventional and balanced design methods used
to dimension and detail the gusset plates connecting the steel braces with the beams and columns
of the CBF. Gusset plate failure was not observed, despite allowing slightly higher proportional
yielding in one test specimen. Moreover, the balance factors of the test specimens made an
important contribution by examining new threshold limit values for the conventional and balanced
design approaches. These threshold limits ranged between 0.38 and 0.52 for conventional design
and between 0.75 and 1.03 for balanced design. However, further testing is recommended to
investigate the transition between these two design limit ranges.
A finite element modelling developed to predict cyclic response effects in the inelastic response
of CBFs has been presented comprehensively in this paper. The finite element models employed in
this study made an important contribution by allowing clear and direct access to salient response
features of brace gusset-plate specimens, direct comparative assessment of real time strain
measurements and observation of life-cycle degradation process under fully reversed cyclic axial
loading. Complimentary analysis and assumptions suitable for CBF design were made to quantify
the complexity associated with the geometry of steel braces and gusset plates. A linear buckling
analysis has been presented to determine effective length factor for models composed of gusset
plate and bracing members. The direct calculation was not possible for complex geometries such
as gusset plate and steel braces. The results of the analysis indicated substantial differences
attributable to gusset plate detailing, even if the length of the brace tube is identical. Thick and
large gusset plates restricted brace buckling while more compact and thinner gusset plates may
enhance overall effective length, leading to reduce buckling strength. The actual effective length of
gusset plate and bracing members lies between the two extremes of the brace models with
idealized pinned and rigid connections end conditions. A good comparison between the hysteresis
loops of the FE models and test models supports the effectiveness and accuracy attributable to the
finite element framework that was developed explicitly in this study. Particularly, comparing
predicted to test values, the mean ratio and Cov on average were 1.14 and 0.18 for yield load, 1.11
and 0.15 for ultimate load, 1.14 and 0.15 for initial buckling load, 1.03 and 0.23 for energy
dissipation capacity and 0.91 and 0.06 for ultimate ductility capacity, respectively.
The FE models confirmed the experimental findings and validated the connection design
methods by examining local strain measurements, lateral displacement outputs, cycles to global
and local buckling, including fatigue-life. The FE model and test results endorse the balanced
design method for delivering consistently greater energy dissipation capacity, including fatigue
lives with improvement. While comparing models of Tremblay (2002) and Goggins (2004) for
predicting post-buckling strength, the predictions mainly were below the results obtained from
tests and FE models, specifically at ductility of 4. Moreover, models of predicting lateral
deformation were compared with results of FE models and found that the model of Tremblay
(2002) gave overestimated prediction, while the model of Hassan et al. (2015) gave relatively
better predictions. The comparison of the observed and predicted peak ductility capacities with
models of Nip et al. (2010) and Tremblay (2002) indicated that the former model is conservative
over the range of slenderness investigated, while the latter model gave relatively close prediction
for specimens with lower slenderness, but overestimated predictions for larger slenderness.
The current study contributes to the need for information on detailing and modelling gusset
plates and bracing members by engineers who look for guidance on realistic structural details
involving economy. The FE and test results presented in this paper allows further investigation of
these observations, including the study of various parameters influencing response of CBF
structure under cyclic axial loading, such as during earthquakes.
Acknowledgments
The first author would like to acknowledge the College of Engineering and Informatics (COEI)
of the National University of Ireland Galway for providing a Ph.D. studentship. This material is
also in part based upon works supported by the Science Foundation Ireland Marine and Renewable
Energy Ireland (MaREI) research centre under Grant No. 12/RC/2302. The last author would like
to acknowledge the support of Science Foundation Ireland through the Career Development
Award programme (Grant No. 13/CDA/2200).
References
ABAQUS (2013), Analysis User’s Manual I-V, Version 6.13, ABAQUS, Inc., Dassault Systémes, U.S.A.
Bjorhovde, R. and Chakrabarti, S. (1985), “Tests of full-size gusset plate connections”, J. Struct. Eng.,
111(3), 667-684.
Broderick, B.M., Goggins, J., Beg, D., Elghazouli, A.Y., Mongabure, P., Le Maoult, A., Hunt, A., Salawdeh,
S., Moze, P. and O’Reilly, G. (2015), “Assessment of the seismic response of concentrically-braced steel
frames”, Exp. Res. Earthq. Eng., Springer, Berlin, Germany.
Brown, V.L.S. (1989), “Stability of gusseted connections in steel structures”, Ph.D. Dissertation, University
of Delaware, U.S.A.
Eurocode (1993), Eurocode 3: Design of steel structures - Part 1-1: General rules and rules for buildings,
European Committee for Standardization (CEN); Brussels, Belgium.
Eurocode (2004), Eurocode 8 : Design of Structures for earthquake resistance Part 1: General rules, seismic
actions and rules for buildings, European Committee for Standardization (CEN); Brussels, Belgium.
Chaboche, J.L. (1986), “Time independent constitutive theories for cyclic plasticity”, J. Plasticity, 2(2), 149-
188.
Cheng, J.J.R., Yam, M.C.H. and Hu, S.Z. (1994), “Elastic buckling strength of gusset plate connections”, J.
Struct. Eng., 120(2), 538-559.
Coffin, L. (1954), “A study of the effect of cyclic thermal stresses on a ductile metal”, Trans. ASME, 76,
931-950.
ECCS (1986), “Recommended testing procedure for assessing the behaviour of structural steel elements
under cyclic loads”, ECCS Technical Committee 1, Structural Safety and Loadings, Technical Working
Group 1.3, Seismic Design, ECCS, Brussels, Belgium.
English, J. and Goggins, J. (2012), “Nonlinear seismic response of concentrically braced frames using
frames using finite element models”, 15th World Conference on Earthquake Engineering, Lisbon,
Portugal, September.
Fell, B.V. (2008), “Large-scale testing and simulation of earthquake-induced ultra low cycle fatigue in
bracing members subjected to cyclic inelastic buckling”, University of California, Davis, MA, U.S.A.
Gardner, L., Saari, N. and Wang, F. (2010), “Comparative experimental study of hot-rolled and cold –
formed square and rectangular hollow sections”, Thin-Walled Struct., 48(7), 495-507.
Goggins, J. (2004), “Earthquake resistance hollow and filled steel braces”, Ph.D. Dissertation, Trinity
College, University of Dublin, Ireland.
Goggins, J., Broderick, B.M., Elghazouli, A.Y., Salawdeh, S., Hunt, A., Mongabure, P. and English, J.
(2018), “Shake table testing of concentrically braced steel structures with realistic connection details
subjected to earthquakes”, Structures, 13, 102-118.
Goggins, J. and Salawdeh, S. (2013), “Validation of nonlinear time history analysis models for single-storey
concentrically braced frames using full-scale shake table tests”, Earthq. Eng. Struct. Dynam., 42(8), 1151-
1170.
Goggins, J.M., Broderick, B.M., Elghazouli, A.Y. and Lucas, A.S. (2006), “Behaviour of tubular steel
members under cyclic axial loading”, J. Construct. Steel Res., 62(1-2), 121-131.
Haddad, M. (2015), “Concentric tubular steel braces subjected to seismic loading: Finite element modelling”,
J. Construct. Steel Res., 104, 155-166.
Haddad, M., Brown, T. and Shrive, N. (2004), “Fintie element modelling of concentriccally braced frames
for earthquakes”, 13th World Conference on Earthquake Engineering, Vancouver, Canada, August.
Hassan, M.S., Goggins, J. and Salawdeh, S. (2015), “Charactersing the effect of global and local geometric
imperfections on the numerical performance of a brace member”, J. Physics: Conference Series, 628(1),
012063.
Hassan, M.S., Salawdeh, S. and Goggins, J. (2018), “Determination of geometrical imperfection models in
finite element analysis of structural steel hollow sections under cyclic axial loading”, J. Construct. Steel
Res., 141, 189-203.
Hu, S. and Cheng, J.J.R. (1987), “Compressive behavior of gusset plate connections”, Structural
Engineering Rep. No. 153; University of Alberta, Canada.

Hunt, A.D. (2013), “Design and analysis of concentrically braced steel frames under seismic loading”. Ph.D.
Dissertation, Trinity College Dublin, Dublin, Ireland.
Kavinde, A.M. and Deierlein, G.G. (2004), “Micromechanical simulation of earthquake-induced fracture in
steel structures”, Report no.145; Stanford University, San Francisco, California.
Lehman, D.E., Roeder, C.W., Herman, D., Johnson, S. and Kotulka, B. (2008), “Improved seismic
performance of gusset plate connections”, J. Struct. Eng., 134(6), 890-901.
Manson, S.S. (1953), “Behaviour of materials under conditions of thermal stress”, Heat Transfer Sympo-
sium, University of Michigan Engineering Research Institute, Michigan, U.S.A.
Nip, K.H., Gardner, L., Davies, C.M. and Elghazouli, A.Y. (2010), “Extremely low cycle fatigue tests on
structural carbon steel and stainless steel”, J. Construct. Steel Res., 66(1), 96-110.
Nip, K.H., Gardner, L. and Elghazouli, A.Y. (2010), “Cyclic testing and numerical modelling of carbon steel
and stainless steel tubular bracing members”, Eng. Struct., 32(2), 424-441.
Nip, K.H., Gardner, L. and Elghazouli, A.Y. (2013), “Ultimate behaviour of steel braces under cyclic
loading”, Proc. Inst. Civil Eng. - Struct. Build., 166, 219-234.
Palmgren, A. (1924), “Die Lebensdauer von Kugellagern (The life of roller bearings)”, Zeitschrift des
Vereines Deutscher Ingenieure, 68(14), 339-341.
Rudland, D.L. and Brust, F. (1997), “The effect of cyclic loading during ductile tearing on the fracture
resistance of nuclear pipe steels”, Fatigue Fracture Mech, 27, 406-426.
Ryan, T., Broderick, B.M., Hunt, A., Goggins, J. and Salawdeh, S. (2017), “Recommendations for numerical
modelling of concentrically braced steel frames with gusset plate connections subjected to earthquake
ground motion”, J. Struct. Integrity Maintenance, 2(3), 168-180.
Salawdeh, S., English, J., Goggins, J., Elghazouli, A.Y., Hunt, A. and Broderick, B.M. (2017), “Shake table
assessment of gusset plate connection behaviour in concentrically braced frames”, J. Construct. Steel Res.,
138, 432-448.
Salawdeh, S. and Goggins, J. (2013), “Numerical simulation for steel braces incorporating a fatigue model”,
Eng. Struct., 46, 332-349.
Tremblay, R. (2002), “Inelastic seismic response of steel bracing members”, J. Construct. Steel Res., 58(5-8),
665-701.
TK
Appendix
Table A1 Symbols and nomenclature

Lower case
a Parameter of a nonlinear regression equation
b Brace width
b1 Parameter of a nonlinear regression equation
bw Whitmore width
Hardening parameter that defines the rate at which the size of the yield surface changes as
biso
plastic strain increases
c Parameter of a nonlinear regression equation
fy Yield strength of brace material
fy,g Yield strength of gusset plate
k Effective buckling length factor
lh Width of gusset plate
lv Depth of gusset plate
ni Number of reversals for which the strain level is applied
t Thickness of brace section
tp Thickness of gusset plate
Upper case
A Cross-sectional area of brace section
Anet,brace Net brace area for gusset plate connection
C Fatigue ductility exponent
Ckin Kinematic hardening parameter
Cov Coefficient of variation
DI Damage index
E Modulus of elasticity
Fc Initial buckling capacity of brace member
Fy Initial yield capacity of brace member
Fu Ultimate load capacity of brace member
Fy,gusset Yield capacity of gusset plate
Lb Length of brace
Ncr Elastic critical buckling load
Nfi Number of reversals causing failure at a specific strain level
Q∞ Maximum change in the size of the yield surface
R y,brace Ratio of actual to nominal yield strength of brace
R y,gusset Ratio of actual to nominal yield strength of gusset plate
SPOS Section positive (point of integration of shell element)
Wtotal Total energy dissipated
Table A1 (continued)
Greek upper case
β Ratio of global imperfection magnitude, 1 and brace tube length, Lb
βww Balanced factor for balanced design
γ Parameter that determines the rate at which the back stresses vary as the plastic strain increases
∆ Lateral deformation
ε′f Fatigue ductility coefficient
εp
Plastic strain amplitude
2
εy Yield strain of material
𝜆̅ Global normalised slenderness
σcr Plate critical buckling stress
σ|o Yield stress at zero equivalent plastic strain
Greek Lower case

α Imperfection factor
δc Applied displacement magnitude
δy Yield displacement
ω1 Global imperfection amplitude
ω2 Local imperfection amplitude
 Ultimate displacement

A Study On Detailing Gusset Plate and Bracing Members in Concentrically Braced Frame Structures

Uploaded by

Copyright:

Available Formats

A Study On Detailing Gusset Plate and Bracing Members in Concentrically Braced Frame Structures

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Study On Detailing Gusset Plate and Bracing Members in Concentrically Braced Frame Structures

Uploaded by

Copyright:

Available Formats

Advances in Computational Design, Vol. 3, No.

A study on detailing gusset plate and bracing members in

Corresponding author, Ph.D. E-mail: [email protected]

Copyright © 2018 Techno-Press, Ltd.

Table 1 Comparison of two gusset plate design philosophies

2. Physical laboratory test

Table 2 Summary of test models (Hunt 2013)

(a) S40-CA-G1 (b) S40-CA-G2 (c) S40-CB-G1

(d) S40-CB-G2 (e) S60-CA-G1 (f) S60-CA-G2

Fig. 2 Test set-up, including frame details

2.1 Loading protocol

2.2 Hysteretic behaviour

3. Finite element modelling framework

modelling, two assumptions were made:

3.1 Material model

(b) FE model incorporated (c) FE model incorporated

Table 3 Material model for steel braces and gusset plates

3.2 Element and mesh size

3.3 Linear buckling analysis

3.4 Geometrical imperfection model

ω2 /t = 0.034 (fy /σcr )0.5 (3)

3.5 Finite element solver

3.6 Fatigue-life model

Table 5 Summary of test measurements

Table 6 Summary of FE results

4.1 Yield strength, initial buckling and post buckling loads

(a) Physical test model (b) FE model

(a) Physical test model (b) FE model

(a) Physical test model (b) FE model

(a) Physical test model (b) FE model

(a) Physical test model (b) FE model

(a) Physical test model (b) FE model

(a) Ductility 2 (b) Ductility 4

(a) Physical test models (b) FE models

4.2 Lateral deformation

In the system under consideration, lateral deformation occurs as a result of out-of-plane

(a) G1 type gusset plate (b) G2 type gusset plate

Fig. 18 Location and direction of strain gauges used in the experiments

(a) S60-CA-G1-5 (b) S60-CA-G2-6

(a) S60-CA-G1-5 (b) S60-CA-G2-6

(c) S60-CA-G1-5 (d) S60-CA-G2-6

4.4 System behaviour

4.5 Fatigue prediction

4.6 Ultimate ductility and energy dissipation capacity

𝜇T = 2.4 + 8.3λ̅ (13)

𝜇 = −0.68 + 26.2λ̅ (14)

𝜇 = 29.1 − 1.07(b/t) (15)

𝜇 = 3.69 + 6.97λ̅ − 0.05(b/tε) − 0.19(λ̅)(b/tε) (16)

𝜇 = 6.45 + 2.28λ̅ − 0.11(b/tε) − 0.06(λ̅)(b/tε) (17)

𝜇 = −3.42 + 19.86λ̅ + 0.21(b/tε) − 0.64(λ̅)(b/tε) (18)

Table 7 Measured and predicted ductility capacities

Fig. 24 Ductility capacity of the physical test models and FE models

4.7 Failure modes

As an example, Fig. 26 shows the experimentally obtained deformed structural behaviour of

Engineering Rep. No. 153; University of Alberta, Canada.

Table A1 Symbols and nomenclature

Greek Lower case

You might also like