Engineering Structures 151 (2017) 221–234

Contents lists available at ScienceDirect

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

Lateral impact response of end-plate beam-column connections

A. Al-Rifaie a, Z.W. Guan a,b,⇑, S.W. Jones a, Q. Wang b
a
School of Engineering, University of Liverpool, Liverpool L69 3GQ, UK
b
School of Mechanical Engineering, Chengdu University, Shiling Town, Chengdu City, Sichuan Province, PR China

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

Article history: The behaviour of different steel beam to column connections has been studied intensively against static
Received 14 April 2017 and seismic loading regimes. However, there is a lack of knowledge on the response of such connections
Revised 3 July 2017 against impact and blast. In order to close this gap, the most common connections with partially depth
Accepted 14 August 2017
end plate (PDEPCs), as a simple connection, and flush plate (FPCs), as a moment resisting connection,
were investigated under both quasi-static and impact loads. Here, eight specimens were tested under
those loading conditions with different locations. 3 D finite element models were then developed and val-
Keywords:
idated against the corresponding experimental results. Full range analyses of the connection responses
Steel connection
Lateral impact
under both loading regimes are then carried out using the validated FE models to examine the internal
Finite element forces of the connections. Finally, the results of full analyses under both loading regimes were compared
Dynamic increase factor and dynamic increase factors (DIF) were proposed to assist predicting the impact response of these types
of connections using the static analysis. The results showed that failure modes under both loading
regimes were similar, but with the larger fracture on the PDEPC under quasi-static load than that under
lateral impact. The DIFs were found to be between 1.02 and 1.21, 1.03 and 1.36 and 1.22 and 1.45 based
on the bolt tensile strength, axial resistance and bending resistance of the connections, respectively.
However, if based on the energy approach, the range of DIFs was recorded between 1.25 and 1.38 using
the experimental results and between 1.19 and 1.34 using the finite element analysis results.
Ó 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://
creativecommons.org/licenses/by/4.0/).

1. Introduction be beneficial to investigate the dynamic behaviour of this critical

part on the structural frame particularly for connections with a
In the past four decades, the structural engineers have given low moment resistance. In steel framed structures having simple
considerable attention on investigating the response of structural or semi-rigid beam-to-column connections, the connections are
members subjected to accidental loads such as impact and blast. likely to be weaker than the columns and beams. However, in this
These loads may be resulted from faulty practice, terrorist attack case, any local failure developed in the connection due to the acci-
or vehicle impact, etc. The collapse of Ronan point in 1968 alerted dental loading may likely be followed by a partially or entirely pro-
the structural designers to the problem of progressive collapse at gressive collapse of the steel frame. Hence, connection response
which local failure of primary structural elements led to the col- should be investigated prior to other steel frame components to
lapse of the connected members [1] which resulted in a dispropor- prevent or reduce the possibility of progressive collapse
tionate collapse. SCI publication P391 [2] that presents the occurrence.
structural robustness of steel framed buildings in accordance with Generally the impact loading could be transferred to any struc-
the Eurocode and UK National Annexes states that ‘‘ In essence, the tural beam-to-column connection by either striking the beam or
objective is to ensure that buildings do not suffer disproportionate the column connected. However, columns are more likely to
collapse under accidental loading. Largely, this is assured in steel expose to such forces than beams such as vehicle impact, flying
framed buildings by designing connections appropriately”. Also, debris or internal explosion, as shown in Fig. 1. Consequently,
after the WTC collapse, it was reported that the connection intensive studies were carried out to investigate the response of
response against impact and fire needs to be understood and quan- different types of columns under such loads (Yu and Jones [4],
tified as critical components of structural frame [3]. Hence, it will Mannan et al. [5], Bambach et al. [6], Zeinoddini et al. [7,8], Al-
Thairy [9], and Shakir et al. [10]). In those studies, axially and
⇑ Corresponding author at: School of Engineering, University of Liverpool, non-axially loaded columns were investigated experimentally
Liverpool L69 3GQ, UK. and numerically under lateral impact loads. Nevertheless, the
E-mail address: zguan@Liverpool.ac.uk (Z.W. Guan). structural aspects of steel frames require columns to be connected

http://dx.doi.org/10.1016/j.engstruct.2017.08.026
0141-0296/Ó 2017 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
222 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234

Fig. 1. The possible cause of a lateral impact.

to beams using suitable connections. Then, studying the columns impact loads applied laterally on the column. The experimental
response with ignoring the connection contribution would lead work contained testing eight L-beam to column connections, four
to incomplete understanding of the overall steel frame behaviour. of them under impact loads and the others under quasi-static load.
Izzuddin et al. [11] realized this fact and concluded that progres- The test set-up was designed to provide moment and axial tensile
sive collapse failure of buildings is largely dominated by the max- force at the connections. Moreover, finite elements models were
imum deformation allowed on the connections in relation to their developed and verified by the experimental results, which were
built-in ductility. further used to predict the internal forces and energies dissipated
The lack of knowledge on the dynamic capacities of steel con- under static and dynamic loads. In order to present a relation
nections indicates that limited studies were conducted. Recently, between the static and dynamic behaviour of the connections,
an experimental and numerical study on fin-plate connections Dynamic Increase Factor (DIF) which is preferred by the structural
under static and dynamic conditions was undertaken with a load- engineers, was suggested to assist predicting internal forces gener-
ing time to failure less than 32 ms. The study verified the ability of ated on the connection due to impact loading based on forces and
the modified component method to predict the connection energy.
response under high strain rate loading [12]. Wang et al. [13] also
investigated numerically the response of a fin-plate connection
due to falling floor impact loads. The main finding was the total 2. Experimental study
displacement could be reduced using high strength steel. Angle
cleat connections were investigated as another type of connection 2.1. Reaction frame fabrication
under high loading rate by Rahbari et al. [14] with the results indi-
cating that such connections are relatively insensitive to the strain The specimens to be tested require a stiff reaction frame to sup-
rate. A numerical study was presented by Kang et al. [15] as an port them under both the quasi-static and impact loads. This frame
attempt to investigate the response of steel frame with moment should be stiff enough to minimize any movement during the test
resisting connections against vehicle impact. The results showed that may affect the results. The frame was designed and fabricated
that the frame remained stable under 40 km/h (11.1 m/s) car hit- at the University of Liverpool and some trial tests were carried out
ting speed, while the frame was severely damaged in a progressive to examine its suitability.
manner when the car speed reached more than 80 km/h. Grimsmo Fig. 2 shows the schematic diagram of the test setup containing
et al. [16] conducted an experimental study at which extended end the details of the frame, in which the frame contains three parts,
plate connections were tested under quasi-static and impact load i.e. floor mounted rails, moveable sub-assembly and bracers. The
hitting the column axially (i.e. shear and bending moment pro- rails provided a fixed location for the drop hammer operator. Also,
duced on the connection). The results showed that the connections holes in the rails were provided to allow the movable sub-
tested behaved in a preferable manner and became more ductile assembly for variable lengths. Two vertically mounted supports
under impact loads. However, Tyas et al. [17] showed that the fabricated were bolted to the rails to provide a rigid base. The cross
PDEPC connection became less ductile under dynamic test com- members with a detachable clamping setup provided a method to
pared to that under quasi-static one. This contradiction in the rigidly clamp the samples. Three rigid bracers were employed to
results from Tyas et al. [17] and Grimsmo et al. [16] indicates that connect the rails to both ends of the sub-assembly frame and the
more research need to be carried out on both connection types to detachable clamping setup in order to minimize the rotational
improve the knowledge on this issue. It should be mentioned that movement of the sub-assembly frame which supports the
the studies conducted by Rahbari et al. [14], Kang et al. [15] and specimens.
Tyas et al. [17] were under lateral dynamic load while the others The rigidity of the reaction frame was examined prior to test the
were under gravitational dynamic loads. The experimental study specimens. Hence, three additional trial specimens having a con-
that carried out by Grimsmo et al. aforementioned at Ref. [16] nection stronger than all of those to be investigated in this study
was followed by a numerical study conducted by the same authors were tested under impact load. The translational and rotational
using the finite element modelling [18]. The main findings were movements of the reaction frame at the detachable clamping
that the energy dissipated by the connection was significantly where the specimen connected to the stiff frame were recorded
increased by reducing the end plate thickness, while marginal using high speed camera. The maximum rotational angle and the
effect on the response of the connection was found by applying dif- maximum downward translation of the detachable clamp for all
ferent axial forces on beam. trials measured were 0.61° and 1.7 mm, respectively. However, it
In this paper, simple and semi-rigid end plate connections were is expected that this error is to be minimized using weaker joints
investigated experimentally and numerically against static and as proposed in this study.
 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234 223

Fig. 2. A schematic and an image of the test setup.

2.2. Specimen preparation corresponding engineering stress-strain curves were obtained for
all constituent materials under quasi-static load. True stress–loga-
Figs. 3(a) and (b) display the test setup considering two loca- rithmic plastic strain curves were also obtained in order to con-
tions of the applied load and two types of connections. The size sider a large displacement in the modelling as it will be
of beam and column were 305
127
37 and 152
152
37, discussed in Section 3.2.1. The material properties of both washers
respectively. An 8 mm thick end plate was welded to the bottom and nuts were assumed to be similar to the bolt properties. Fig. 4
of the beam by 4 mm fillet weld then connected to the column shows the stress strain curves of the steel sections used in this
by four M16 bolts with the grade 8.8. All welding work was carried study. The effect of strain rate was also taken into account which
out by a licensed welder on a private company (WREN Industrial & is to be discussed in Section 4.2.1 to obtain dynamic material prop-
Marine Fabrication Ltd., Sandon Industrial Estate-Liverpool-L5 erties and then to be used as input data in the FE modelling.
9YN) and special care was taken to avoid plate distortion due to
high welding temperature. In the real situation, most structural 2.4. Structural testing
frames are provided with lateral bracing to prevent the sway of
the frame. This was considered in the test setup by providing a lat- The connections prepared were tested under static load using
eral restraint to the beam. Thus, each beam flange was connected hydraulic actuator and lateral impact load using a drop hammer.
to the stiff frame by six M16 bolts with the grade 8.8. Roller sup- A schematic of the test setup under impact load is shown in
port was attached to the end of column to prevent any horizontal Fig. 2. The same setup was used for the static tests by replacing
sway of the column as in a practical situation and to enhance the the drop hammer by a quasi-static loading device. Also the same
rotational stiffness of the frame. stiff frame was used to clamp the specimens under both load
Fig. 3(c) also indicates the two types of connections investigated regimes.
in this study, i.e. the PDEPC (as a simple connection) and the flush For the impact test, the beam of each specimen was connected
end plate (as moment resisting connection). In the former, the first to the stiff frame by 12 M16 bolts with the grade 8.8. Then a
plate was welded to the beam web only while it was welded to flat projectile with a mass of 107.5 kg and a contact area of
the beam web and flange in the latter. Table 1 shows the test spec- 100
100 mm was released from a height of 2.9 m to hit the free
imens and matrix of parameters used in the experimental work end of the column with an initial measured velocity of
under static and impact loads. 7.5 ± 0.05 m/s, which generate a dynamic tensile force and bending
moment on the connection. A laser Doppler Velocimeter (LDV) sys-
2.3. Material testing tem was used to acquire the velocity time history of the projectile
during the short period of the impact loading. The system was con-
Standard tensile coupon tests were performed for all steel sec- nected to a computer via an interface card and fiber optic cable to
tions involved in the specimens, in addition to the high strength manage the operating software. The processing of the LDV signals
bolts. The coupons were also taken from the web and flange of both was performed to minimize the noise using the digital filter soft-
the beam and column to be tested separately as the properties of ware imPRESSion 6 with a cut-off frequency of 1000 Hz [19]. The
them may be varied. The stress-strain curves of the web and flange acceleration was evaluated from the differentiation of the mea-
of the aforementioned sections showed a slight difference. Hence, sured velocity time history. The impact force was obtained by mul-
the results from both the webs were selected to represent the tiplying the impactor mass by the acceleration.
stress-strain curves of these parts. All the tests were repeated three In order to capture the displacement time history at the free
times and the replicated tests showed excellent repeatability. The end of the specimens, a high speed camera (HSC) was used with
224 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234

(a) Elevation view of L-beam to column connection specimen loaded at a distance

(L1) from the centre line of the connection

(b) Elevation view of L-beam to column connection specimen loaded at a distance

(L2) from the centre line of the connection and the locations of strain gauges

(c) Types of connection being investigated in this study (Section a-a in Fig. 3a)
Fig. 3. Dimensions, loading, strain gauge locations and boundary conditions of the two types of connections investigated with two different loading locations (all dimensions
in mm).
 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234 225

Table 1 dynamic tensile force and moment under lateral impact. The finite
Test specimens. element analysis (FEA) was employed to obtain those internal
Specimen reference Loading type Connection type Load location forces as they cannot be measured experimentally. Also, the vali-
PFI8 Impact PDEPC L1a dated FEA model could be used to investigate the effects of various
PNI8 Impact PDEPC L2b parameters on the connection response by intensive parametric
FFI8 Impact FPC L1 studies.
FNI8 Impact FPC L2
PFI8S Quasi-static PDEPC L1
PNI8S Quasi-static PDEPC L2
3.1. Description of the finite element modelling
FFI8S Quasi-static FPC L1
FNI8S Quasi-static FPC L2 Finite element analysis was performed using ABAQUS/Explicit
a [20] which is suitable for both of dynamic and quasi-static loading.
L1 load located 1224 mm from the center line of the connection.
b
L2 load located 374 mm from the center line of the connection. Also, it is a preferable approach for dealing with complicated con-
tact problems rather than implicit or general static procedure. A
special attention should be taken for using ABAQUS Explicit to
900 model quasi-static loading by ensuring that the inertial forces
Bolt (fy=645,fu=798) remain insignificant. This was checked by ensuring that the ratio
800
Column (fy=336,fu=442) of the kinetic energy to the internal energy (ALLKE/ALLIE) should
700
Engineering Stress (MPa)

be always less than 5% [20]. Different loading rates and mass scal-
Beam (fy=337, fu=448)
ing were studied to save the analysis time and the optimum values
600 plate (fy=253,fu=374)
were found out to be 0.67 mm/s and 106, respectively, with ALLKE/
500 ALLIE less than 2% for all models.
400 Fig. 5 shows the FE model developed to simulate the impact
test. Eight-noded solid element with reduced integration (C3D8R)
300 was used to model all parts. Four elements through the thickness
200 of each part were generated to avoid element hourglass. Due to
the large deformations that connection components (bolts and
100 plate) may experience, finer mesh was generated at and near those
0 components. The surface of the column flange with possible con-
0 0.05 0.1 0.15 0.2 0.25 0.3 tact with the projectile was also refined to obtain reasonable
Engineering Strain results of velocity and impact force time histories. Also, to make
use of the symmetry nature, a half model was employed to save
Fig. 4. Engineering stress-strain curves obtained from uniaxial tension tests.
the CPU of the analysis. The connection between the stiff frame
and the beam flanges was modelled as a fixed end due to negligible
one high voltage light to increase the clarity of the target to be hit movement, which was observed during the trial tests using the
by the projectile. ProAnalyist motion analysis software was high speed camera as mentioned in Section 2.2. The projectile
employed to convert the frames captured by the HSC to displace- was modelled as a flat rigid body with dimensions of
ment time history curve. In addition to the LDV and HSC, four mul- 100
100
100 mm and a mass of 53.75 kg (half of the total
tipurpose strain gauges were attached to each specimen at mass) assigned with a downward velocity of 7.5 m/s. The bolt
different locations to capture the strain time history. Two of them was modelled without threads but their effect was considered by
were placed on the end plate, while the other two were located on a reduction of the area of the threaded region. Hence, a hole is cen-
the top flange of the column where the maximum moment was trally located through the threaded region only with an area equal
expected, and on the web of the beam as shown in Fig. 3b. The to the gross area of the bolt minus the net tensile area. A represen-
strain gauges were wired and connected to a conditioning unit tative hole with a diameter of 7.9 mm was applied to make the
which completes a full bridge and amplifies the bridge output. equivalent area where the hole is located equal to 152 mm2.
The output of the signal conditioner was connected to four chan- Modelling of contact between various surfaces is one of the
nels Tektronix TDS2024C oscilloscope of 2 Gs/s sampling rate. A most critical processes. Hence, considerable attention was made
calibration equation of 5 V = 10,000 me was used to convert the to select the proper master and slave surfaces and to assign appro-
voltage time history of the oscilloscope to strain time history. priate interaction properties to model the interaction. Tie con-
Hence, after 10,000 me the system is unable to capture strain. strains and surface to surface contact formulations were used to
In the static test, a hydraulic actuator was used to apply a quasi- model contact surfaces. The former was employed to connect weld
static load on the specimen as mentioned in Section 2.1. Also, the to the beam and plate, nuts to bottom washers, and nuts to bolts as
specimens were attached with strain gauges at the same locations no visible deformation was observed on these parts during the
for impact specimens and the same system used in impact test was experimental tests. However, the surface to surface contact was
used to obtain strain measurements with a longer time scale. One employed between bolt heads and top washers, washers and end
LVDT was placed at the free end of the column to capture the plate, bolt shank and holes of plate and column, plate and column,
deflection with each load increment. roller and column, in addition to projectile and column. Penalty
friction formulation with a coefficient of friction of 0.2 between
contact surfaces was selected to simulate the tangential behaviour
3. Finite element analysis of the contact, while the normal hard contact allowing separation
was assumed for all interaction surfaces. The time histories for
In order to investigate the connection response under static and impact force and displacement underneath the projectile in addi-
impact loads, the internal forces produced in the connection com- tion to the strain in the selected four locations where strain gauges
ponents need to be determined. The main parameters that affect were placed were requested in the model for the validation
the connection behaviour under lateral impact load were specified purpose.
to be tensile strength of bolts, axial resistance and moment resis- The FE model developed above for modelling the impact
tance of the connection. This is because the connection experiences response was modified for quasi-static models. It was initially
226 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234

Fig. 5. The FE model of the impact test.

started to model the quasi-static analysis using similar mesh size point (Fig. 4). These curves then were modified using the dynamic
to that used for impact analysis. However, the CPU time of the increase factor included in Johnson-Cook (DIFJC) model as follows.
quasi-static analysis was found to be significantly larger than that
DIF JC ¼ 1 þ C ln e_  ð4Þ
of the impact case. Therefore, mesh sensitivity study was con-
ducted to select the optimum mesh size at which the results would Here, C is the strain rate constant; e_  ¼ e_ =e_ 0 is the strain rate ratio,
not show noticeable dynamic effect. It was found that the coarser where e_ and e_ 0 are the current strain rate and the reference quasi
mesh could be used for quasi-static analysis comparing with that static strain rate (e_ 0 = 0.001 s1), respectively. The constant C was
for the impact one. assumed to be 0.039 for the end plate and weld, and 0.0072 for high
strength bolt, washers and nuts as adopted by Ribeiro et al. [21].
3.2. Modelling of materials behaviour The beam and column were not considered as strain-rate sensitive
materials in the modelling of the low velocity impact. Also, it should
3.2.1. Modelling the Elasto-plastic behaviour be mentioned that to simplify the simulation, the material proper-
The Elasto-plastic behaviour of steel materials was modelled ties of the heat-affected zone were assumed to be the same as the
using ABAQUS/Explicit based on the engineering stress-strain base material.
curves obtained experimentally as shown in Fig. 4. Hence, modulus
of elasticity (E) and yield stress (ry) were taken from the curves for 3.2.2. Modelling the damage
each material to be used as input data to define the elastic stage up After the strain hardening stage, onset of damage and damage
to the yielding point. In order to obtain reasonable material prop- evolution need to be modelled. Fig. 6 shows a typical stress strain
erties in the plastic stage, the engineering stress-strain curves were curve with progressive damage degradation of an isotropic mate-
modified to obtain the true stress-strain curves using the following rial. The solid line after onset of damage (Damage parameter (D)
equations. = 0) represents the initiation of damage, while the dashed curve
rtrue ¼ reng ð1 þ eeng Þ ð1Þ refers to the material response without damage. Hooputra et al.
[22] proposed a procedure at which ductile damage and shear
etrue ¼ lnð1 þ eeng Þ ð2Þ

where reng and eeng are the engineering stress and the correspond-
ing engineering strain, respectively. These equations were used up
to the onset of necking (at the point of ultimate stress (Fu)). Beyond
necking, a tri-axial strain starts to develop which makes the beha-
viour more complicated and the equations above are not valid.
Therefore, a simplified methodology was used to draw the true
stress strain curve beyond necking by assuming that the material
experiences considerable strain corresponding to a constant stress
level. The ultimate stress (Fu) was used as the constant stress value,
while the fracture strain was calculated using Bridgman strain
equation.
A
etruef ¼ ln ð3Þ
Af

where A is the original cross sectional area of the tensile specimen

tested and Af is the cross sectional area after fracture.
Plastic strain hardening behaviour up to the peak stress was
modelled based on the stress-strain relationships after the yielding Fig. 6. Stress-strain curve with progressive damage degradation [20].
 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234 227

damage could be predicted for aluminium alloys. This approach It can be seen that the maximum peak force produced in a time less
was adopted to predict damage for isotropic ductile material. In than 1 ms followed by drastically descended multi-peaks due to
this paper, ductile damage under impact load was modelled using the intermittent impact. Then, the impact force begins to be more
the equivalent plastic strain-triaxial stress state envelope devel- stable at a force level of about 50 kN. As the maximum displace-
oped by Ribeiro et al. [21]. It was used to estimate the damage ini- ment reached, the projectile start to rebound and a noticeable
tiation of T-stub connection under impact load which was related degradation in force value can be seen. On the other hand, the dis-
to this study. Also, it was developed for both of bolts and plate placement time history indicates that the maximum displacement
which have more or less the same material properties of those used of the specimen corresponds to the stabilised force that was pro-
in this study. Shear damage initiation modelling in ABAQUS is duced after multi peak forces.
described as a function of shear stress ratio, strain rate and the Figs. 8–11 show the impact force-displacement curves for all
equivalent plastic strain (epl
o ). In order to define these parameters,
specimens. It can be seen that the impact force (F) versus displace-
the same model was employed using isotropic metal plasticity ment can be generally divided into three stages which are demon-
constitutive model and both of shear stress ratio and strain rate strated in Fig. 8, i.e.
were requested as output in a group of elements where the shear
failure occurred close the weld toe on the plate. Then, the fracture (1) The peak stage: at which the impact force reaches its maxi-
strain is obtained from true stress-strain relationship under uniax- mum value with less than 3 mm of displacement;
ial tension. Once damage initiation is detected at any element, the (2) The plateau stage: in this stage, the connection begins to
damage evolution stage starts to lead to the progressive degrada- deform plastically with a relatively stable impact force after
tion of the element stiffness until the ultimate failure. In this study, the first peak;
an effective plastic displacement assuming a linear relationship (3) The bounce stage: in this stage, the curve descends from the
between effective plastic displacement (upl) and the damage vari- total displacement to the separation point as the projectile
able (D) was adopted to model the damage evolution. It is consid- starts to separate from the struck column.
ered also that the effective plastic displacement to be an input to
the model is a function of the mesh size and the equivalent plastic Changing the location of impact load from L1 = 1.224 m to L2 =
strain as follows. 0.374 m leads to increasing the peak force by 8% and 24% for the
PDEPCs and the FPCs, respectively. Also, the peak forces of the FPCs
upl ¼ Lc epl
o ð5Þ are greater than those of the PDEPC by 25% and 10% when the load
is located at L2 and L1 from the connection, respectively. This is due
where Lc is the element size. (epl o ) can be determined in Fig. 6, to the higher stiffness of the FPC, at which the velocity of the pro-
depending on the stress-strain curve of the materials tested. The jectile at the onset of contact with the stuck column decelerated
modelling of ductile and shear failure under quasi-static was per- faster than the PDEPC specimens. After the first peak stage partic-
formed using ABAQUS keyword option. Hence, the ultimate tensile
stress of the plate and bolt were used as input data to model the
ductile damage. Also, shear damage of the end plate was modelled 450
in the same way of ductile damage and a value of the equivalent Stage 1 Stage 2
400
plastic strain at the onset of shear damage (epl
s ) of 0.2 was complied
350
with the experimental results for all models.
300 FE
Force (kN)

250 Exp
4. Results and discussion
200 Stage 3
4.1. Experimental results under impact load and validation of the FE 150
model
100

4.1.1. Impact force-displacement relationships 50

The impact force-displacement curves were obtained by com- 0
0 5 10 15 20 25
bining both the impact force-time history acquired from the LDV
and the displacement-time history acquired from HSC. Fig. 7 shows Displacement (mm)
the time histories of load and displacement for the specimen FFI8.
Fig. 8. The impact force-displacement curve of the specimen FNI8.

350 70
350
300 60
300 FE
Displacement (mm)

250 50 Exp
250
Force (kN)

Force (kN)

200 40 200
Force-FE
150 Force-Exp 30 150
Displacement-FE
100 20 100
Displacement-Exp
50 10 50

0 0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0 10 20 30 40 50 60 70
Time (s) Displacement (mm)

Fig. 7. The force time history and displacement time history of the specimen FFI8. Fig. 9. The impact force-displacement curve of the specimen FFI8.
228 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234

350 The validation process contained the validation of both the

load-time history and displacement-time history first for a speci-
300 FE men. Then, both experimental curves were combined to obtain
250 Exp the experimental load-displacement curve as mentioned above.
Finally, the former curves were compared with those obtained
Force (kN)

200 from the FE model. Fig. 7 shows the displacement time history
150 and load time history of the specimen FFI8, together with the
numerical modelling output. Good correlation is obtained. Figs. 8–
100 11 also show the validation of the FE models against experimental
results. A good agreement is obtained, which indicates the model is
50
capable of predicting the three stages of impact response. How-
0 ever, the peak force in the simulation seems increasing faster than
0 5 10 15 20 25 30 35
that in the test, which may be attributed to the assumption of the
Displacement (mm) rigid projectile that has the much higher contact stiffness than the
Fig. 10. The impact force-displacement curve of the specimen PNI8.
deformable steel projectile used in the test. Also, the model is able
to predict the total displacement at the free end of the column up
to the separation point.

350
4.1.2. Deformation modes
300 FE
The results from all impact tests demonstrate three different
250 Exp modes of failure, i.e. end plate bending, first bolt pair bending
and fracture of the end plate close to the weld toe. Figs. 12–14
Force (kN)

200
show those failure modes of the connection components under
150 impact load. It can be seen that all specimens experienced large
deformations at the end plate, while fracture of the weld toe
100 occurred only in the PDEPC. The bend of the first pair of bolts is
50 noticed as well which is likely attributed to the bearing stresses
between the end plate and the bolt shank. This may lead to another
0 type of failure representing either by combined shear-bending of
0 20 40 60 80 100
the bolt or bearing failure of the plate. Owens and Moore [23]
Displacement (mm)
showed that the bearing failure of plate occurred under pure quasi
Fig. 11. The impact force-displacement curve of the specimen PFI8. static axial tensile load on the joint. Hence, both of these failure
modes need to be covered using the validated FE model under
impact load. Moreover, the failure mode of both joints was not
ularly when the load is far away from the joint (L1), a clear inter- affected significantly by changing the load location.
mittent impact can be seen. This is attributed to missing contact In general, very good correlation between the observed and pre-
between the projectile and the struck column where a clear gap dicted failure modes was obtained, as indicated in Figs. 12–14. It
was noticed between them by the high speed camera. Besides, can be seen that the FE models produce accurate predictions of
the peak forces induced in these specimens descends gradually the three failure modes. In addition to the failed shape, the strain
until reaching the mean stage as the difference in velocities time histories in the four critical locations of each specimen were
between the projectile and column was smaller. Grimsmo et al. also examined and verified against the measurements, which con-
[16] also observed multiple hits leading to intermittent impact firms the validity of the FE models developed. Fig. 15 shows the
and gradual descending of peak forces in joints with a flexible validation of strain time histories in these locations (Fig. 3b) for
end-plate under impact load. the specimen PNI8. A good agreement is obtained for all the gauges
Regarding the plateau stage, it is clear that the specimens except strain gauge 2 which saturated at 10,000 me that is the max-
loaded near the connection (L2) experienced a higher plateau force. imum strain to be captured by the oscilloscope as mentioned in
Also, around 84% and 60% of the total energy were absorbed in this Section 2.4.
stage for the specimens loaded at L2 and L1 from the joint, respec-
tively. It is expected that these percentages will be higher if a col-
umn is supported from the free end as that in the practical case.
However, this can be investigated by using the validated model (Pa)
to perform parametric studies to study different cases of geometry
and energies in addition to the boundary conditions. Therefore, the
plateau stage can be classified as the critical stage between other
stages at which most of the applied energy is absorbed.
As the velocity of both projectile and struck column attains to
zero, the third stage starts and the displacement approaches to
its maximum value. The area under the curve at this stage repre-
sents the recovered energy by the specimen after the impact
energy is dissipated. The recovered energy of both connections
was not affected for the load being located near the connection
(L2), i.e. approximately same area under curve, as shown in Figs. 8
and 10. While, the FPC demonstrated a recovered energy of 285% of
that of the PDEPC if the load was located far away from the connec-
tion (L1) as shown in Figs. 8 and 10. Fig. 12. The failure mode of the first bolt under impact load.
 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234 229

(Pa)

Fig. 13. The failure mode of the specimen FNI8 under impact load.

(Pa)

Fig. 14. The failure mode of the specimen PNI8 under impact load.

0.8 18
0.7 SG1-FE
16
0.6 SG1-Exp
14
0.5
Strain (με) x103

Strain (με) x103

SG4-FE
0.4 12
SG4-Exp
0.3 10
0.2 SG3-FE
8
SG3-Exp
0.1 6 SG2-FE
0
4 SG2-Exp
-0.1
-0.2 2
-0.3 0
0 0.005 0.01 0.015 0 0.005 0.01 0.015
Time (s) Time (s)
Fig. 15. Strain-time histories of the specimen PNI8 under impact load.

4.2. Experimental results under Quasi-static load and validation of the 17 also show the verification of the quasi static model against
FE models experimental results. Good agreements are obtained for the bilin-
ear load-displacement curves in spite of the slight oscillations pro-
4.2.1. Force-displacement relationships duced in theses curves particularly after the onset of fracture.
Figs. 16 and 17 show the load displacement curves of all spec- These oscillations can be controlled using lower loading rate which
imens tested under quasi-static load. It can be seen that all the con- in turn leads to increase the time of analysis. Hence, this was
nections demonstrated bilinear behaviour with noticeable avoided in this study because very slight effect was found on the
degradation in the stiffness up to failure. The FPC showed higher internal forces using lower loading rate.
stiffness than the PDEPC in both loading locations, as expected,
since the former is a moment resistance connection. Also, the 4.2.2. Deformation modes
PDEPC showed a considerable ductility up to failure though it is Generally, the deformation modes under quasi-static load were
classified as a simple connection. similar to that under impact load for all specimens but with a
As mentioned in Section 3.1, ABAQUS/explicit was found to be higher damage for the PDEPC specimens close to the weld toe, as
the appropriate approach for simulating a connection with com- shown in Fig. 18. The strain rate where the plate tearing occurred
plex contact conditions under quasi-static loading. Figs. 16 and seems to be higher than other locations which increase the princi-
230 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234

160 and both of them were delayed due to the strain rate effect. Hence,
the PDEPC responded relatively more ductile under impact load
140 than quasi-static load.
120
4.3. Full-range analyses of the structural response
100
Force(kN)

The validated FE models were used to analyze the full-range

80 response of end plate connections under quasi-static and impact
load, including the internal forces in bolts, the axial and moment
60
resistances of the connections investigated. These outputs repre-
FNI8S-FE sent the key factors that affect the joint response under lateral
40
FNI8S-Exp loads. Also, this analysis helps with proposing a DIF for each key
20 PNI8S-FE factor separately and then to propose a preliminary DIF at which
PNI8S-Exp the impact response can be predicted using static analysis that is
0
0 5 10 15 20 25 30 35 a preferable procedure for structural engineers. In addition to the
Displacement (mm) internal force based DIF, an energy based DIF was also proposed
by comparing both energies dissipated by the system under
Fig. 16. Force-displacement curves of the specimen FNI8S and PNI8S under quasi- quasi-static and impact loading. The validated model is also
static load. employed to estimate the strain-rate distribution in the critical
parts of the connection.

60
4.3.1. Internal forces on bolts
The tensile forces on the bolt were requested in the FE model as
50 the contact force between the bolt head and the top washer. Figs. 19
and 20 show the force time histories of the first and second bolt
under impact load. Generally, continuous force flow with a few
40
spikes can be seen for all specimens in spite of intermittent impact
Force(kN)

observed in the impact force-displacement relationships. This is

30 because the struck column was kept accelerated and displaced
despite the contact separation of the impactor. The numerical
FFI8S-FE results showed that the first bolt for all specimens experienced a
20
FFI8S-Exp faster change in force than the second bolt after the onset of loading
PFI8S-FE (for example in specimen FFI8, the first bolt needs less than 1.5 ms
10 PFI8S-Exp
to reach 60 kN while the second bolt reaches the same load by more
than 6 ms). Also, it can be seen that after the maximum displace-
ment of the free end of the column, the forces on both bolts begin
0
0 20 40 60 80 100 to degrade rapidly due to the fast deceleration of the applied force
in the bounce stage. Moreover, specimens loaded at L2 from the
Displacement (mm)
joint center experienced a faster bolt force degradation to reach
Fig. 17. Force-displacement curves of the specimen FFI8S and PFI8S under quasi- the separation point. Besides, it can be seen that the first bolt for
static load. all specimens loaded dynamically gains a considerable amount of
its peak force in a short period, which indicates that the bolt expe-
rienced a high strain rate. This may lead to bolt thread stripping fail-
ple stresses of the plate and then to delay the tearing failure in that ure which is a type of failure to be avoided in connection. Mouritz
location. In other words, two major components contributed to [24] showed that failure load of mild steel bolt thread under impact
produce this failure, i.e. excessive bending and shear on the plate tension load decreased with increasing strain rate. However, this

(Pa)

Fig. 18. Failure mode of the specimen PNI8S under quasi-static load.
 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234 231

140 150
FNI8 FNI8
120
FFI8 100 FFI8
PNI8 PNI8
100

Axial force (kN)

PFI8 PFI8
Force (kN)

50
80

60 0

40
-50
20

0 -100
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0 0.01 0.02 0.03 0.04
Time (s) Time (s)

Fig. 19. Internal tensile force time histories of the first bolt row under impact load. Fig. 21. Internal axial resistance time histories under impact load.

90 robustness of building frames subjected to terrorist explosion

FNI8 attack in the US [25]. However, using simple connections may raise
80
FFI8 the possibility of local failure followed by progressive collapse.
70 PNI8 Therefore, it would be beneficial to investigate the internal
60 PFI8 moment capacity under impact and static loads to have a deep
Force (kN)

insight into the structural behaviour. The ‘‘Free body cut” com-
50 mand was used again in the same section that was used to inves-
40 tigate the internal axial force to examine the internal moment
resistance of the connection under both load regimes. Fig. 22
30
shows the internal moment versus time curves of the four speci-
20 mens tested under lateral impact load. As expected, PDEPC demon-
10 strated a lower moment resistance than the FPC in both loading
points. For a fair comparison between the moment resistances of
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 the connections, the maximum displacement at the free end of
Time (s) the FPC was used here as a bench mark. The comparison showed
that specimens FFI8 and PFI8 have an internal moment resistance
Fig. 20. Internal tensile force time histories of the second bolt row under impact of 39.27 kN m and 15.79 kN m, respectively at a displacement of
load.
60.2 mm (at time = 0.0179 s). In other words, the internal moment
resistance of the PDEPC was about 40% of that of the FPC when
type of failure was not observed in all specimens tested here, which
loaded at L1, while this ratio decreased to 25% for specimens loaded
may need to be examined under higher impact energies.
at L2. Moreover, it can be seen that both the connection types reach
the peak moment faster as the loading point is moving towards
4.3.2. Internal axial resistance joint. Therefore, the specimens loaded close to the joint required
The axial capacity of the joint plays a very important role in the less than 1 ms to reach its peak moment resistance, while more
progressive collapse failure which is referred in Eurocode as a tying than 5 ms was needed for others loaded far away from the
force [2]. Hence, a section was made in the beam near the joint connection.
using ‘‘free body cut” command available in ABAQUS and axial
force versus time was requested to represent the axial resistance
4.3.4. Strain-rate distributions at the critical locations
of the connection. Fig. 21 shows the axial resistance versus time
Strain rate is defined as the change in strain of a material with
curves for specimens tested under impact load. It can be seen that
respect a time. To obtain more knowledge on the strain rate distri-
the connection axial force loaded at L1 started its peak with a con-
siderable negative value of about 50 kN. In this short period of
50
time, the column tried to push the beam upward, creating down-
ward (negative) reaction which is opposite to the situation under FNI8
40
quasi-static load. In order to prove this, a beam translation was FFI8
PNI8
tracked in y-axis using FE model with a large deformation scale 30
Moment (kN.m)

PFI8
factor and the upward translation of the beam was observed dur-
ing that period. Moreover, disregarding the negative peaks of those 20
specimens shows that FPC specimens need a shorter time to reach
10
their peak force than the PDEPC specimens. The specimens loaded
far away from the joint seem having more vibration than others, as 0
can be seen in Fig. 21, which may be considered as a precaution of
the nut loosening that is an unfavourable connection failure. -10

-20
4.3.3. Internal moment capacity 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Perhaps, there is no concern about the integrity and robustness Time (s)
of a structural frame having moment resistance connections.
Hence, it was suggested to use them to improve the structural Fig. 22. Internal moment resistance time histories under impact load.
232 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234

35 methods to predict the strength of a structural member under

FNI8 dynamic load is to enhance its static strength multiplying by an
30 appropriate DIF. This approach was used by Wang et al. [26] to
FFI8
estimate the dynamic resistance capacity of concrete filled double
PNI8 steel tubular members under lateral impact. Also, it was used by
25
PFI8 Liu et al. [27] to address the appropriate DIF for the design of struc-
Strain rate (s-1)

tures against progressive collapse. Generally, two types of DIF were

20
commonly used: force-based DIF and displacement-based DIF. The
former is the ratio of the dynamic load to the static load of the
15 structural system under the same displacement, while the latter
is the ratio of the dynamic displacement to the static displacement
10 under the same load. In this study, the former was found to be
more appropriate so that the impact force-displacement relation-
5 ships were recorded first. Then, the corresponding specimens were
loaded under quasi-static loads up to the same maximum displace-
0 ment acquired under impact load. However, DIFs were proposed in
P1 P2 P3 P4 terms of the internal force and the energy dissipation in this study
Critical selected places to evaluate their consistence.

Fig. 23. The averaged strain rate distributions at the selected critical locations. 5.1. DIFs based on the internal force

bution in the system, the strain time histories were requested at Table 2 shows comparisons of the extreme values of the main
the elements in the connection that experienced large deforma- components in the connection that affect the connection response
tions, such as those in the end plate and the first bolt. The scalar under quasi static and impact loads. Clearly, all the static results
variable SDEG available in ABAQUS was used to specify the ele- demonstrate the lower extremes than the impact results in differ-
ments that initiate the damage degradation. The locations selected ent levels. The first bolt row in all specimens shows a small differ-
to investigate the strain rate distribution in this study are shown in ence in DIF, with a range of 1.17–1.21 being recorded. On the other
Fig. 5, as P1, P2, P3 and P4. Fig. 23 shows the average strain rate in hand, the second bolt row demonstrates the lower dynamic effect
those points. To compare the strain rate on the plate of both types than the first row in a range of 1.02–1.16. The DIFs predicted for
of connections, P1 could be compared with P2 and P3. Hence, it the axial connection resistance were identified for the FPC to be
could be noticed that the strain rate at P1 of the specimen PNI8 1.03 and 1.12 for specimens loaded at L2 and L1, respectively. How-
was 32.39 s1, while it was 15.31 s1 at P2 and 18.15 s1 at P3 in ever, these factors increased from 1.15 to 1.36 in the PDEPC spec-
the specimen FNI8. In other words, the PDEPC experienced the imens. Hence, the axial capacity of moment resistance connections
twice strain rate of the FPC, if the load is located near the connec- loaded close to the connection seems to be less affected dynami-
tion. Approximately, the same trend could be identified if the spec- cally than those loaded far away. The PDEPCs exhibited the similar
imens loaded far away from the connection. This could be trend of the DIF for both loading locations but slightly higher than
attributed to the higher deformation that PDEPC exhibited at a those of the moment resistance connection by about 20%. Regard-
specific time if compared with FPC as the latter is stiffer than the ing the internal moment resistance, the PDEPC demonstrated
former. P4 could be used to compare the strain rate at the critical higher DIFs than those of FPC particularly if the load applied far
point on the first bolt. It is clear that specimens loaded near the away from the connection. Then a considerable reduction from
connection experienced the higher strain rate than those loaded 1.45 to 1.22 (about 16%) was produced if the FPC replaced by
far away from the connection. For instance, the first bolt of the PDEPC for specimen loaded far away the connection. However, a
specimen FNI8 experienced a strain rate of 12.54 s1, while it slight difference of about 2% could be noticed if the load was
was only 3.92 s1 in the specimen FFI8. This is because the first bolt applied near the connection for both types of connections. This
of the specimen loaded near the connection needs less time than could be attributed to the higher moment resistance of the FPCs
those loaded far away to reach its maximum force, as shown in than those of PDENCs which are not designed to resist bending
Fig. 19. moment. Here, the PDEPCs showed a good impact moment resis-
tance in spite of the fact that they are not designed to resist bend-
5. Dynamic increase factor (DIF) ing moment.

Dynamic increase factor can be defined as the ratio of dynamic 5.2. DIFs based on the energy principle
strength to its static counterpart of a material or structural mem-
ber. Response of a structural member at high strain rate is more The energy principle was used to estimate the experimental and
complicated than its static response. Therefore, one of the straight numerical DIF. Hence, dissipated energy which represents the area

Table 2
Summary of the extreme internal forces of the connections under impact and quasi-static loads and the related DIFs.

Specimen Static Impact DIF- Static Impact DIF - Static Impact DIF-axial Static Impact DIF –
tensile force tensile force 1st tensile force tensile force 2nd axial axial resistance moment moment moment
in the 1st in the 1st bolt in the 2nd in the 2nd bolt resistance resistance resistance resistance resistance
bolt row bolt row row bolt row bolt row row (kN) (kN) (kN m) (kN m)
(kN) (kN) (kN) (kN)
FNI8 109.40 132.84 1.21 68.05 79.21 1.16 128.76 133.08 1.03 23.94 33.35 1.39
FFI8 109.10 130.03 1.19 63.32 69.65 1.10 50.26 58.03 1.15 34.87 42.86 1.22
PNI8 82.47 96.64 1.17 56.60 62.12 1.10 95.27 107.50 1.12 9.49 13.51 1.42
PFI8 81.03 97.85 1.21 46.12 47.13 1.02 39.12 53.01 1.36 13.79 20.05 1.45
 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234 233

Table 3
Summary of energies dissipated under quasi-static and impact loads and the related DIFs.

Specimen Experimental results DIFExp FE results DIFFE

Es (J) Ei (J) Es (J) Ei (J)
FNI8 1871 2575 1.38 1981 2495 1.26
FFI8 1862 2330 1.25 1883 2248 1.19
PNI8 2007 2760 1.38 2057 2498 1.21
PFI8 2030 2802 1.38 1932 2580 1.34

under the load-displacement curve was determined using digital Both connection types investigated showed large local plastic
filter software imPRESSion 6 [19] for specimens tested under both deformations on the end plate and the first bolt row under
load regimes. Then the DIFs were calculated using the following quasi-static and impact loads. Generally, similar failure modes
equation. was observed under both load regimes but larger tearing close
the weld toe was observed on the PDEPC under quasi-static load.
Ei This confirms the increase on the connection stiffness under
DIF ¼ ð6Þ
Es impact load due to the strain rate effect at the crack tip.
Simulation of bolted connections under impact load required
where Ei is the energy dissipated under impact load and Es is the special attention due to the complexity of geometry and contact.
energy dissipated under quasi-static load. The static dissipated In addition to its efficiency to model the dynamic events, ABAQUS/-
energy was calculated by assuming that the specimen has the same explicit was found to be an effective method to simulate bolted
initial stiffness after releasing the load. Table 3 shows comparisons connections under quasi-static loading but with ensuring the
of the Ei and Es obtained based on the experimental results and kinetic energy to internal energy ratio for each part of the model
numerical analyses for specimens tested, in addition to the corre- to be less than 5%. Also, using loading rate of 0.67 mm/s with 106
sponding DIFs. mass scaling factor can help the analysis with reducing computa-
Clearly, all the static results demonstrate the lower energy dis- tional costs without a significant dynamic effect on numerical
sipations than the impact counterparts. Based on the experimental results.
results, the DIF varies from 1.25 to 1.38 for all specimens tested. Finally, two approaches were used to predict the dynamic effect
However, it varies from 1.19 to 1.34 based on the FE models. of such connections under lateral impact load. In the internal force
Hence, the difference between them is less than 5%, which indi- based approach, the DIFs identified are between 1.03 and 1.45,
cates good correlation. Moreover, the specimen PFI8 demonstrates while in the energy based approach the maximum DIF was identi-
the higher DIF both experimentally and numerically, which is coin- fied experimentally and numerically to be 1.38 and 1.34,
cided with that using the internal force approach, as the maximum respectively.
DIF was identified in this specimen against bending moment. It
could be said that the proposed DIFs based on the two approaches
obtained good insight into the dynamic effect on the end plate con- Acknowledgements
nections under lateral impact load.
It should be mentioned that the proposed DIF needs to be inves- The work presented in this paper was supported by the Min-
tigated further with different circumstances such as connection istry of Higher Education and Scientific Research in Iraq, which is
geometries, impact locations, boundary conditions and applied gratefully appreciated. The authors would like to thank Mr A. Al-
energies to verify the possibility of using these DIFs. This will be Husainy for his help with the data processing. The authors also
studied separately. wish to thank Mr. M. Bratley and Mr. D. Neary for their assistance
in the experimental work.

6. Conclusions
References
Experiments work and the finite element analyses have been
successfully undertaken to study the structural response of end [1] Adminstration U.S. G.S. GSA. Progressive collapse analysis and design
guidelines for new federal office buildings and major modernization
plate connections under both lateral quasi-static and impact loads. projects; 2003.
Here, the dynamic increase factor (DIF) has been proposed by eval- [2] Way AG. Structural robustness of steel framed buildings; 2011 [SCI. 55].
uating the dynamic response of the connections in relation to their [3] McAllister T, Corley G. World trade center building performance study: data
collection, preliminary observations, and recommendations, Fedral Emergency
quasi-static counterparts to indicate the dynamic effect on the con- Management Agency; 2002.
nection systems. The finite element models developed have also [4] Yu J, Jones N. Numerical simulation of a clamped beam under impact loading.
been rigorously validated against the corresponding experimental Comput Struct 1989;32(2):281–93.
[5] Mannan MN, Ansari R, Abbas H. Failure of aluminium beams under low
results in terms of load-displacement traces, deformation modes, velocity impact. Int J Impact Eng 2008;35(11):1201–12.
strain-time histories. The validated models are further used to [6] Bambach MR, Jama H, Zhao XL, Grzebieta RH. Hollow and concrete filled steel
obtain the internal forces of the connection components which hollow sections under transverse impact loads. Eng Struct 2008;30
(10):2859–70.
cannot be measured experimentally. Based on the above work, [7] Zeinoddini M, Parke GAR, Harding JE. Axially pre-loaded steel tubes subjected
the following conclusions can be drawn as follows. to lateral impacts- an experimental study. Int J Impact Eng 2002;27:669–90.
Test results showed that the lateral impact response of the end [8] Zeinoddini M, Harding JE, Parke GAR. Axially pre-loaded steel tubes subjected
to lateral impacts (a numerical simulation). Int J Impact Eng 2008;35
plate connection can be classified as three stages: (i) the peak
(11):1267–79.
stage, (ii) The plateau stage and (iii) the bounce stage. The plateau [9] Al-Thairy H. Behaviouand design of steel column subjected to vehicle impact.
stage can be considered as the most critical stage because more School of Mechanical, Aerospace and Civil Engineering, PhD Thesis, University
than 60% of the impact energy (area under the load-displacement of Manchester; 2012.
[10] Shakir AS, Guan ZW, Jones SW. Lateral impact response of the concrete filled
curves) was dissipated, along with the development of plastic steel tube columns with and without CFRP strengthening. Eng Struct
deformation. 2016;116:148–62.
234 A. Al-Rifaie et al. / Engineering Structures 151 (2017) 221–234

[11] Izzuddin BA, Vlassis AG, Elghazouli AY, Nethercot Obe DA. Assessment of [19] ImPREESSion 6. NecoletTecnonlgies. Germany: Weisang GmbH & Co. KG;
progressive collapse in multi-storey buildings. In: Proceedings of the 2002.
Institution of Civil Engineers – Structures and Buildings 2007;160(4):197–205. [20] Abaqus documontations. SIMULIA; 2012.
[12] Stoddart EP, Byfield MP, Davison JB, Tyas A. Strain rate dependent component [21] Ribeiro J, Santiago A, Rigueiro C, Barata P, Veljkovic M. Numerical assessment
based connection modelling for use in non-linear dynamic progressive of T-stub component subjected to impact loading. Eng Struct
collapse analysis. Eng Struct 2013;55:35–43. 2016;106:450–60.
[13] Wang H, Yang B, Zhou XH, Kang SB. Numerical analyses on steel beams with [22] Hooputra H, Gese H, Dell H, Werner H. A comprehensive failure model for
fin-plate connections subjected to impact loads. J Constr Steel Res crashworthiness simulation of aluminium extrusions. Int J Crashworthiness
2016;124:101–12. 2004;9(5):449–64.
[14] Rahbari R, Tyas A, Buick Davison J, Stoddart EP. Web shear failure of angle- [23] Owens GW, Moore DB. The robustness of simple connections. Struct Eng
cleat connections loaded at high rates. J Constr Steel Res 2014;103:37–48. 1992;70(3):37–53.
[15] Kang H, Shin J, Kim J. Analysis of steel moment frames subjected to vehicle [24] Mouritz AP. Failure mechanisms of mild steel bolts under different tensile
impact. In: APCOM & ISCM, Singapore; 2013. loading rates. Int J Impact Eng 1994;15(3):311–24.
[16] Grimsmo EL, Clausen AH, Langseth M, Aalberg A. An experimental study of [25] The Oklahoma City bombing: improving building performance though multi-
static and dynamic behaviour of bolted end-plate joints of steel. Int J Impact hazard mitigation. Washington DC; 1996.
Eng 2015;85:132–45. [26] Wang R, Han L-H, Zhao X-L, Rasmussen KJR. Analytical behavior of concrete
[17] Tyas A, Warren JA, Stoddart EP, Davison JB, Tait SJ, Huang Y. A methodology for filled double steel tubular (CFDST) members under lateral impact. Thin-
combined rotation-extension testing of simple steel beam to column joints at Walled Struct 2016;101:129–40.
high rates of loading. Exp Mech 2012;52(8):1097–109. [27] Liu C, Fung TC, Tan KH. Dynamic performance of flush end-plate beam-column
[18] Grimsmo EL, Clausen AH, Aalberg A, Langseth M. A numerical study of beam- connections and design applications in progressive collapse. J Struct Eng
to-column joints subjected to impact. Eng Struct 2016;120:103–15. 2016;142(1):04015074.