You are on page 1of 18

Experimental Tests of Steel Frames with Different

Beam–Column Connections under Falling Debris Impact


Hao Wang 1; Kang Hai Tan 2; and Bo Yang 3
Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Falling debris may initiate or promote the progressive collapse of a structure. To mitigate progressive collapse under falling
debris impact, steel frames with five types of beam–column connections were tested. Both single and multiple impacts were applied to
each specimen to obtain dynamic responses, load-resisting mechanisms, and impact resistance. The results showed that a typical impact
process was divided into three stages. The falling debris impact was mainly resisted by the inertia effect in the initial stage and by deformation
in subsequent stages. The majority of the external work applied to the system was absorbed by bending deformation, especially by the plastic
rotation at midspan of the beam. All the specimens had the same global deformation shape under the midspan impact scenario. The evolution
of catenary action in specimens with different beam–column connections was similar. Catenary action was shown to significantly improve the
load-carrying capacity and energy absorption in specimens with high levels of rotational ductility. Greater load-carrying capacity did not
offer greater impact resistance. Load-carrying capacity and ductility both played a significant role in the energy absorption capability of the
specimens. DOI: 10.1061/(ASCE)ST.1943-541X.0002469. © 2019 American Society of Civil Engineers.
Author keywords: Debris impact; Progressive collapse; Disproportionate collapse; Steel frames; Beam–column connections; Impact
resistance; Energy absorption.

Introduction (Fujikake et al. 2009; Chen et al. 2017; Ribeiro et al. 2016; Lew
et al. 2013; Sadek et al. 2013) focused more on this scenario, which
Researchers began to pay attention to progressive collapse inci- investigated the loading redistribution in the remaining structures
dents after the partial collapse of the Ronan Point apartment build- after the column loss. In these studies, if the remaining structures
ing in 1968. Many progressive collapse events have occurred in could not redistribute the vertical load, and the structural compo-
the following decades, such as the World Trade Center disaster nents reached failure criteria, progressive collapse or disproportion-
on September 11, 2001. To mitigate the progressive collapse of ate collapse occurred. However, a falling-debris-impact scenario
building structures, the alternate load path method, an important may also cause progressive collapse of structures, regardless of
design approach, was recommended by many design codes, includ- accident loading. This scenario assumes that structural and non-
ing the GSA (2013) and DoD (2009), as it can conveniently and structural elements drop down and hit the remaining structures
effectively address the issue of structures subjected to abnormal below due to abnormal loads. Fig. 1 shows the collapse process of
loads. the Ronan Point apartment building. Falling debris caused the final
To date, most studies on progressive collapse have been based collapse. The entire collapse process could be divided into two
on column-removal scenarios. In this method, regardless of the stages. In the first stage, the gas blast on the 18th floor destroyed
accident loading effect, it is assumed that the column is removed the load-bearing walls. The upper structures could not reach a new
suddenly. If the remaining structures can redistribute the vertical stable status. The collapse occurred on the upper floors due to the
load sustained by the column removed and reach a new stable removal of vertical load-bearing elements. In the second stage,
status, progressive collapse can be prevented. Previous studies elements from the upper floors dropped down to hit the structures
below. The lower structures did not have sufficient impact resis-
1
tance, and collapse occurred on the lower structures. A new dispro-
Ph.D. Candidate, Key Laboratory of New Technology for Construction portionate collapse standard will be proposed by ASCE to mitigate
of Cities in Mountain Areas, Chongqing Univ., Ministry of Education,
the disproportionate collapse of building structures based on GSA
83 Shabeijie, Chongqing 400045, China; Ph.D. Candidate, School of Civil
Engineering, Chongqing Univ., 83 Shabeijie, Chongqing 400045, China.
(2013)/DoD (2009) guidelines, published research papers, and re-
ORCID: https://orcid.org/0000-0001-6552-7821. Email: wanghao13689@ ports. The new standard may allow for partial collapse where a
cqu.edu.cn falling debris impact that may directly result in a disproportionate
2
Professor, School of Civil and Environmental Engineering, Nanyang collapse. However, it is difficult to predict the dynamic behaviors
Technological Univ., 50 Nanyang Ave. Singapore 639798. Email: ckhtan@ and failures of structures subjected to falling debris impacts. Stud-
ntu.edu.sg ies need to be carried out to mitigate the progressive collapse of
3
Professor, Key Laboratory of New Technology for Construction of structures under impact load.
Cities in Mountain Areas, Chongqing Univ., Ministry of Education, In recent years, more attention has been paid to building struc-
83 Shabeijie, Chongqing 400045, China; Professor, School of Civil Engi- tures subject to falling impact. Kaewkulchai and Williamson
neering, Chongqing Univ., 83 Shabeijie, Chongqing 400045, China (cor-
(2006) developed a modeling strategy to demonstrate the impact
responding author). Email: yang0206@cqu.edu.cn
Note. This manuscript was submitted on August 7, 2018; approved on of failed members against other structural components. The im-
May 10, 2019; published online on November 12, 2019. Discussion per- portance of demonstrating the effect of impact on predicting pro-
iod open until April 12, 2020; separate discussions must be submitted for gressive collapse was given by an example illustration. Vlassis et al.
individual papers. This paper is part of the Journal of Structural Engi- (2009) proposed an energy-based design-oriented methodology
neering, © ASCE, ISSN 0733-9445. for building structures subject to impact from an upper floor.

© ASCE 04019183-1 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Collapse process of Ronan Point apartment.

Wang et al. (2016) simulated the dynamic behavior of steel beams This connection could be designed as either a single connection
with fin plate connections under impact loads. The effects of or a semirigid connection. It was conveniently used to connect
parameters, i.e., impact energy, impact position, imposed load, bolt beams and different column sections, such as steel hollow sections,
hole distance, and fin plate thickness, were studied by validated concrete-filled tubular columns, and universal columns.
finite-element models. Olmati et al. (2017) presented a framework Progressive collapse is a dynamic process. To consider the
for simplified reliability analysis for computing the probability of dynamic effect, sudden-column-removal tests were carried out by
punching failure due to a slab falling from above. The reliability Liu et al. 2013, 2015, 2016) to investigate the dynamic behaviors
method proposed was applied to a case study of an existing build- of web cleat connections, top-and-seat with web angle connec-
ing, showing that a column removal situation is not always critical, tions, and flush end plate connections. Both force-based and
whereas a slab falling from above is much more detrimental. Based displacement-based dynamic increase factors were measured from
on the energy-transfer methodology proposed by Vlassis et al. the tests. Another test method was proposed by Grimsmo et al.
(2009), Jawdhari et al. (2018) assessed a withstanding progressive (2015, 2016) and Al-Rifaie et al. (2017), where a drop hammer
collapse under falling debris for building structures. A theoretical machine was employed to apply impact loads to specimens. The
calculation was performed to investigate a reinforced concrete flat tests were conducted under a low strain-rate effect to compare the
plate structure and composite steel-concrete floor structure under a difference in terms of loading rate. It should be noted that these
fully plastic-impact scenario and a fully rigid impact. Haas et al. studies focused on the dynamic behaviors of structures and com-
(2018) presented an overview of analysis methods, i.e., conserva- ponents above the failed columns.
tion of energy method, single degree-of-freedom (SDOF) method, Under a falling debris impact, whether a structure will survive
and multiple degree-of-freedom (MDOF) method, to predict struc- depends on both the impact resistance of the structure and the
tural element performance under debris impact. A design example impact load. If the structure can absorb the energy of the impact
showed that all three analysis methods offered similar predictions. without failure, the structure can survive. Otherwise, progressive
The conservation of the energy method is the simplest approach to collapse may occur. In history, there are examples of structures that
predict peak displacement. However, if the entire displacement survived or did not survive under impact. In 1967, the Ronan Point
time-history is needed, the SDOF or MDOF methods should be building was not robust enough to resist the falling debris impact
chosen. from upper stories, as shown in Fig. 1. Finally, partial collapse oc-
In general, beam–column joints play an important role in the curred. In 1995, the sixth story of an eight-story building was dam-
stability of structures, because they are the weaker components in aged in the Great Hanshin Earthquake. The upper stories dropped
a structural system, especially for steel and composite structures. down and struck the lower story, but collapse did not occur, as
The beam–column connections usually control the extent of cat- shown in Fig. 2. Thus, it is necessary to conduct studies to ensure
enary action because of the limited resistance and rotation capacity that the structures can absorb impact energy without failure. In this
of the connections. Some studies have been conducted to improve paper, the experimental test assumed that the upper structures could
the progressive collapse resistance of beam–column connections not bear the redistributed load due to the failure of columns. The
under column-removal scenarios. Computational simulations were debris from the upper floor impacted the structure below. If the
conducted by Khandelwal and EI-Tawil (2007) to study catenary residual structure cannot resist the impact from the upper story,
action in moment-resisting steel frames. Welded joints with and the debris from the two levels will fall and impact the lower story
without reduced beam sections were employed in the structures. together, and progressive collapse will occur. This is a typical
Many key design parameters that affect the formation of catenary domino effect. However, it is possible that the collapse process
action were investigated. Yang and Tan (2012, 2013a) carried out a stopped if the lower story can absorb more energy than the testing
series of experimental tests and numerical analyses to investigate story to resist the impact load from both the upper stories and test-
catenary action and the progressive collapse resistance of differ- ing story.
ent types of bolted beam–column joints under a central-column- This paper focuses on the impact resistance of steel-framed
removal scenario. According to the results of tests and numerical structures. A series of experimental tests were carried out to study
analyses, component-based models were proposed by Yang and the dynamic behavior of steel frames with different types of beam–
Tan (2013b) to predict the behavior of bolted-angle beam–column column connections subject to impact loads. Five types of beam–
joints. Wang and Xue (2013) proposed a new type of beam–column column connections, i.e., welded unreinforced flange-bolted web
connection named a reverse channel connection. A series of ex- connections (WUF-B), reduced beam section connection (RBS),
perimental tests were carried out to investigate the effects of fin plate connection (FP), reverse channel connection with flush
parameters on moment-rotation characteristics and failure modes. end plate (RC-FEP), and reverse channel connection with extended

© ASCE 04019183-2 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Building survives under falling impact.

end plate (RC-EEP), were used in the test program. The midspan the weights could be installed on top of the hammer body, which
of the beam was chosen as the loading position for all five tests was gripped by a crossbeam above. An electric control system was
under the greatest flexural damage. The resistance of individual used to operate the testing machine, such as moving the crossbeam
connections was demonstrated. The energy absorption capacity up or down, gripping the hammer hook, or releasing the drop ham-
was proposed to evaluate the resistance of structures instead of mer. The drop hammer slid along two vertical rods. The drop ham-
the load-carrying capacity. mer was first lifted to a certain height, and then the hammer hook
was released. Immediately, the drop hammer freely dropped to im-
pact the specimen below. The drop hammer was centered on the
Test Program axis of the steel beam.

Test Setup
Specimens and Material Properties
A prototype structure was used to test the dynamic behaviors of
steel frames subject to a falling debris impact, as shown in Fig. 3. Full-scale specimens were not suitable for experimental tests due
A one-span steel frame was extracted from a steel-framed building to limited headroom in the laboratory. Therefore, half-scale spec-
to study the dynamic behavior, deformation, and impact resistance imens were designed and fabricated for impact tests. A total of five
of steel frames under falling debris impact. The impact tests as- specimens were tested, and the details of the tested specimens are
sumed that part of a structure in the story above the specimen was shown in Table 1. All specimens had the same span of 3,160 mm
severely damaged by abnormal loads so that it dropped down and between the middle of the columns. British universal beams (UB)
hit the testing beam at midspan. Fig. 4 shows the front view and 203 × 133 × 30 and British universal columns (UC) 203 × 133 × 71
three-dimensional view of the test setup. Horizontal restraint was were used as the sections of steel beams and steel columns, respec-
provided by a steel A-frame on each side to simulate the restraint tively. At the midspan of the beam, steel plates and stiffeners were
from undamaged surrounding structural elements. Columns were welded to limit local deformations and out-of-plane distortions.
connected to A-frames by horizontal load cells. The bottom of the A total of five types of beam–column connections, including
columns was connected to the strong floor by pinned supports. WUF-B, RBS, FP, RC-FEP, and RC-EEP, were considered in the
Impact loads were applied by a ZCJ9034 drop weight impact experiments. The WUF-B, RBS, and FP were commonly applied in
testing machine (MTS Industrial Systems, Shenzhen, China). steel frames, viz. WUF-B and RBS for rigid or semirigid joints and
The drop hammer consisted of the hammer body, hammerhead, FP for conventionally pinned joints. Reverse channel connection
and weights. The basic weight of the drop hammer was 530 kg, was a new type of connection proposed by Wang and Xue (2013).
including a load cell. The weight of one block was 30 kg, and A 6-mm-thick fin plate was used in the WUF-B, RBS and FP.
A parallel flange channel (PFC) 150 × 75 × 18 and an end plate
with 8-mm thickness were used in RC-FEP and RC-EEP. In all
specimens, S355 steel was used for the beams and columns, and
S275 was used for fin plates, end plates, and PFCs. Grade 10.9 M20
bolts were used for all specimens. The details of the five specimens
are shown in Fig. 5.
The falling-debris-impact scenario is a process in which failed
elements drop down and strike structures below. In the experiment,
the mass of the drop hammer was 830 kg, which was approximately
equal to the mass of the beam and a 1-m-wide slab in the story
above. The dropping height of the hammer was 3 m, which is a
typical story height in a building. Because both the mass and the
dropping height of the hammer approximated that of actual falling
debris, the experimental test was representative of the impacts an
actual structure would experience in a collapse scenario.
Tensile coupon tests were carried out to obtain the mechanical
properties of the different steel components. The coupons were
Fig. 3. Prototype steel frame.
taken from the flange and web of the beam, the column and

© ASCE 04019183-3 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

(a)

(b)

Fig. 4. Test setup: (a) front view; and (b) three-dimensional view.

Table 1. Summary of specimens


Specimen Mass of drop Dropping Fin plate/ Parallel flange
identification hammer (kg) height (m) Beam section Column section end plate (mm) channel (PFC) section
WUF-B 830 3.0 UB 203 × 133 × 30 UC 203 × 203 × 71 150 × 70 × 6 —
RBS 830 3.0 UB 203 × 133 × 30 UC 203 × 203 × 71 150 × 70 × 6 —
FP 590 3.0 UB 203 × 133 × 30 UC 203 × 203 × 71 150 × 70 × 6 —
RC-FEP 830 3.0 UB 203 × 133 × 30 UC 203 × 203 × 71 220 × 150 × 8 PFC 150 × 75 × 18
RC-EEP 830 3.0 UB 203 × 133 × 30 UC 203 × 203 × 71 290 × 150 × 8 PFC 150 × 75 × 18
Note: FP = fin plate; WUF-B = welded unreinforced flanges and bolted web; RBS = reduced beam section; RC-FEP = reversed channel connection with flush
end plate; and RC-EEP = reversed channel connection with extended end plate.

channel, the fin plate, the end plate, and the stiffener. Table 2 sum- were placed at four sections symmetrically, viz. Section 2 and Sec-
marizes the measured yield stress, ultimate strength, and fracture tion 3 at 700 mm away from the midspan and Section 1 and Section
strain. 4 at 900 mm away from the midspan, respectively. Four strain
gauges were placed in each cross-section, as shown in Fig. 6.
All strain gauges were in the elastic state during the impact process.
Instrumentation The bending moment MðtÞ at each section and the axial force NðtÞ
Measurements of the impact force and vertical displacement were calculated by Eqs. (1) and (2). At the left half of the beam,
were based on the load cell and laser sensor, as shown in according to the bending moments in Sections 1 and 2, the bending
Fig. 4. A Type 9393A load cell (Kistler Instrumente, Winterthur, moments at midspan M mid;L ðtÞ and connection M con;L ðtÞ were cal-
Switzerland) with 1,000 kN capacity was installed on the drop culated by Eqs. (3) and (4). At the right half of the beam, similarly,
hammer to measure directly the impact force. To capture the rap- the bending moments at midspan and connection were calculated
idly changing vertical displacement at the impact location, a by the bending moments at Sections 3 and 4 as follows:
Type HG-C1400 laser distance sensor (Panasonic, Osaka, Japan)
with a measuring range of 200–600 mm was placed below the 1
MðtÞ ¼ W Z σm ðtÞ ¼ W Z E½ε3 ðtÞ þ ε4 ðtÞ − ε1 ðtÞ − ε2 ðtÞ ð1Þ
specimen at midspan. To obtain the internal forces, strain gauges 2

© ASCE 04019183-4 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

(a)

(b) (c)

(d)

(e) (f)

Fig. 5. Details of specimens and beam–column connections: (a) overview of specimens; (b) FP; (c) WUF-B; (d) RBS; (e) RC-FEP; and
(f) RC-EEP.

1 X4 where σm ðtÞ = bending stress at the edge of the cross-section;


NðtÞ ¼ AS σn ðtÞ ¼ EAS εi ðtÞ ð2Þ σn ðtÞ = axial stress of the cross section; εi ðtÞ = strain measured by
4 i¼1 different strain gauges at the same cross section; AS = area of the
cross section; and MS1 and M S2 = bending moments at Sections 1
900 and 2, respectively.
M mid;L ðtÞ ¼ ½M ðtÞ − M S1 ðtÞ þ M S1 ðtÞ ð3Þ
200 S2 Two sets of data acquisition systems, SIRIUS STG (Dewesoft,
Trbovlje, Slovenia) and TML (Tokyo Measuring Instruments Lab-
772 oratory, Tokyo, Japan) multirecorders, were employed simultane-
M con;L ðtÞ ¼ ½M ðtÞ − MS2 ðtÞ þ M S2 ðtÞ ð4Þ
200 S1 ously, because each system only had 16 channels. In both systems,

© ASCE 04019183-5 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Table 2. Material properties of steel impact forces of RC-FEP and RC-EEP were 254 and 223 kN,
Yield strength Ultimate strength Fracture and the corresponding maximum displacements were 102.9 and
Component (MPa) (MPa) straina 92.7 mm, respectively. A comparison of steady impact force and
maximum displacement showed that greater steady impact force
Beam flange 370 536 0.33
Beam web 386 540 0.32
reduced the deformation of specimens to resist the same impact
Column flange 375 556 0.29 load. Because WUF-B, RBS, RC-FEP, and RC-EEP used the iden-
Column web 412 574 0.29 tical drop hammer (830 kg, 7.67 m=s), the impact energy of the
Channel flange 402 555 0.27 four specimens were the same. In Stage 1, according to the test
Channel web 406 552 0.33 results, the energy absorbed by the four specimens was 3.7, 3.8,
Stiffener 395 553 0.29 4.2, and 4.0 kN, respectively, and the displacement of the four spec-
Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

Fin plate 349 485 0.36 imens was 18, 17, 17, and 18 mm, respectively. In Stage 1, the four
End plate 311 460 0.36 specimens absorbed approximatively an identical amount of en-
pffiffiffiffiffi
a
Fracture strain is based on proportional coupon gauge length of 5.56 S0 , ergy and appeared with approximatively identical displacement.
where S0 is original cross-sectional area of coupons. Because the four specimens had identical energy absorption in
Stage 1 and identical impact energy, the four specimens needed to
absorb an identical amount of energy in Stage 2. According to the
relationship between force, displacement, and energy [Eq. (5)], the
the test data were recorded at a sampling rate of 100 kHz with specimen with greater steady impact force absorbed an identical
low-pass filters at 1 kHz to eliminate high-frequency ambient amount of energy under less displacement. Thus, the greater impact
noise. force reduced the deformation of specimens under identical impact
energy. The FP had the smallest steady impact force of 190 kN, but
its maximum displacement was 88.2 mm, which was in the middle
Test Results and Discussions under Single Impact of the other four tests. This outcome is because the drop hammer
employed in FP was 590 kg, 71% of the mass of the drop ham-
Impact force-time and vertical displacement-time histories were mer used in other tests. Under the impact with the same drop
directly measured by the load cell and laser sensor, respectively. hammer mass, the steady impact force affected the duration of the
Internal forces were calculated by Eqs. (1)–(4) based on strain impact process. The specimen with greater steady impact force fin-
gauges in Sections 1–4. Fig. 7 shows time histories of impact force, ished in a shorter time.
vertical displacement, axial force, bending moment at midspan, and
connections. Table 3 summarizes the test results, including maxi-
mum impact force, steady impact force, duration of impact process, Axial Force and Bending Moment
maximum displacement, and the steady bending moment at mid-
The internal forces obtained from the strain gauges of all specimens
span and connections.
are compared in Figs. 7(c–e). At the midspan, the beam section,
the impact plate, and the stiffeners of all five specimens were
Impact Force and Displacement the same. Thus, the bending moment at midspan of all specimens
The test results are summarized in Table 3. The steady impact force developed similarly during the impact process, and the steady
was defined as follows: bending moment at midspan was approximately 130 kN · m, as
R t2 shown in Fig. 7(c). Fig. 7(d) shows the comparison of bending
E t PðtÞdδðtÞ moment-time histories at connections. As a typical rigid/semirigid
Fs ¼ 2 ¼ 1 ð5Þ beam–column connection, WUF-B had a greater bending moment
δ2 δ2
of −105 kN · m in the steady stage. The steady bending moment
where Fs = steady force; E2 = energy absorption during Stage 2; t1 at the connection of RBS was less than that of WUF-B due to the
and t2 = start and end times of Stage 2, respectively; PðtÞ = applied reduced beam section. Typically used as a simple connection, the
force; δðtÞ = vertical displacement; and δ 2 = vertical displacement FP connection provided nearly no bending moment during the im-
during Stage 2. Figs. 7(a and b) show impact force-time and pact process. The capacity of the two reverse channel connections
displacement-time histories. It was obvious that all five specimens was between the typical rigid/semirigid connections and the typical
had different steady impact forces, maximum displacements, and simple connection. The steady bending moments at the connection
durations of the impact process. The WUF-B and the RBS had of RC-FEP and RC-EEP were −29 and −44 kN · m, respectively.
greater steady impact forces of 324 and 310 kN and smaller maxi- Compared with RC-FEP, the bending moment at the connection
mum displacements of 75.4 and 75.2 mm, respectively. The steady of RC-EEP increased by 52% due to the extended end plate.

Fig. 6. Layout of strain gauges.

© ASCE 04019183-6 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

(c) (d)

(e) (f)

Fig. 7. Time history curves: (a) impact force; (b) vertical displacement; (c) bending moment at midspan; (d) bending moment at connection; (e) axial
force; and (f) typical impact process (WUF-B).

Table 3. Summary of test results under single impact


Maximum Steady bending Steady bending
Specimen Peak impact Steady impact Duration of impact displacement moment at midspan moment at connection
identification force (kN) force (kN) process (ms) (mm) (kN · m) (kN · m)
WUF-B 607 324 30.7 75.4 139 −105
RBS 592 310 31.3 75.2 127 −75
FP 564 190 36.9 88.2 127 −9
RC-FEP 587 223 41.2 102.9 131 −29
RC-EEP 603 254 37.3 92.7 137 −44
Note: FP = fin plate; WUF-B = welded unreinforced flanges and bolted web; RBS = reduced beam section; RC-FEP = reversed channel connection with flush
end plate; and RC-EEP = reversed channel connection with extended end plate.

In Fig. 7(e), axial forces that developed in the beams of all spec- Impact Process
imens are compared. Under the single impact load, the maximum
displacement of all specimens was less than 102.9 mm, which was A typical impact process was divided into three stages based on
only half of the height of the beam section. The catenary action did the impact force-time and displacement-time histories, as shown
not significantly develop due to the small displacement. in Fig. 7(f). The first stage was the contact stage from the time

© ASCE 04019183-7 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


(a) (b)
Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Simplified model of beam subjected to point mass impact: (a) system of beam under impact; and (b) SDOF system.

the drop hammer first contacted the specimens to approximately md vd0 þ αmb vb0 ¼ md vdt þ αmb vbt ð6Þ
2.5 ms. In this stage, the impact force changed rapidly from
0 kN to the maximum impact force and then decreased to nearly where vdt and vbt = velocities of the drop hammer and specimen
0 kN. The maximum impact force was 564 kN in FP and approx- after contact, respectively; and αmb = equivalent mass of the beam.
imately 600 kN in the other tests. The 590 kg drop hammer was When the drop hammer and the contact region of the specimen
employed in FP, and the 830 kg drop hammer was employed move together, the velocity vt can be calculated by Eq. (7)
in the other tests, which means that the connection types slightly md vd0
vt ¼ ; for vb0 ¼ 0 m=s ð7Þ
affected the maximum impact force, whereas the maximum im- ðmd þ αmb Þ
pact force and the mass of the drop hammer had a positive corre-
lation. The vertical displacement increased linearly. As shown in Based on the impact force-time curves [Fig. 7(a)], the shape
Fig. 7(c–e), the axial force and bending moment at midspan and of the impact impulse I was assumed to be triangular and equaled
connection developed rapidly in this stage. From approximately the change of the momentum of the drop hammer and the specimen
2.5 milliseconds to the time the displacement reached maximum as Eq. (8)
value, the impact process was in the second stage, i.e., the plateau Z
1
stage. In this stage, the oscillation of the impact force asymp- I ¼ FI ðtÞdt ¼ FP Δt ¼ md ðvd0 − vt Þ ¼ αmb vt ð8Þ
totically decreased to a steady value. The vertical displacement in- 2
creased nonlinearly. The internal force also remained steady until where Δt = duration of the process. Combining Eqs. (7) and (8), the
the vertical displacement reached maximum value. The connection peak impact force can be calculated by Eq. (9)
types significantly affected the impact force in this stage. Greater
steady impact force appeared in specimens with connections that 2αmb vt 2αmb md vd0
FP ¼ ¼ ð9Þ
had greater bending moment. Under the same impact load, spec- Δt ðαmb þ md ÞΔt
imens with greater bending moment at beam–column connection
absorbed the impact energy in a smaller deformation. After the dis- For the simplified SDOF system, α is equal to 0.5 for a simply
placement reached maximum value, the impact process entered supported beam and approximately 0.375 for a fixed-fixed beam
the third stage, i.e., the bouncing stage. After reaching maximum (Kaewkulchai and Williamson 2006). The duration Δt was approx-
displacement, specimens started to rebound. The direction of mo- imately 1 ms, according to the test results. The peak impact force
tion of specimens and drop hammer changed from downward to should have been 420–641 kN for WUF-B, RBS, RC-FEP and
upward. All parameters were noticeably descending until the drop RC-EEP and 497–628 kN for FP.
Fig. 7(f) shows that Stage 2 had a relatively long time duration
hammer was separated from the specimens.
versus Stage 3. The reason can be explained as follows.
The load applied by a mass was its self-weight in a static test.
To simplify the analyses, Stage 1 was ignored because the du-
However, in the impact test, the maximum impact force was ap-
ration was short. According to the momentum theorem, Stages 2
proximately 600 kN, but the self-weight of the drop hammer was
and 3 can be expressed as follows:
only 8.1 kN (5.8 kN for FP). The maximum applied force was Z
approximately 74 (103 for FP) times the self-weight of the drop m v
md Δvs2 ¼ Fs2 ðtÞdt ¼ FS t2 ⇒ t2 ¼ d d0 ð10Þ
hammer. Even in the plateau stage, the steady impact force was FS
190–324 kN for different specimens, which were 28–40 times the
Z
self-weight of the drop hammer. The results showed an obvious 1 2md v3
difference between the static test and the impact test under gravity. md Δvs3 ¼ Fs3 ðtÞdt ¼ FS t3 ⇒ t3 ¼ ð11Þ
2 FS
For frame-type structures, the impact of dropping weight on
the beam is a collision that generates a flexural response with struc- where md = mass of drop hammer; Δvs2 and Δvs3 = change of
tural deformation far from the contact region (Kaewkulchai and velocities in Stages 2 and 3, respectively; Fs2 ðtÞ and Fs3 ðtÞ = load
Williamson 2006). A point mass striking transversely against a curves in Stages 2 and 3, respectively; Fs = steady force during
beam, referred to as a beam–mass system [Fig. 8(a)], was used as Stage 2; t2 and t3 = durations of Stages 2 and 3, respectively;
the basis for analysis. An SDOF model was used to approximate vd0 = initial velocity of drop hammer; and v3 = velocity of the drop
the response at the point of contact in the beam–mass system, as hammer and specimen at the end of Stage 3.
shown in Fig. 8(b). The time of contact between the drop hammer In Stage 2, the energy absorbed by specimens, i.e., elastic de-
and specimens was assumed to be sudden so that the conservation formation energy Ee and plastic deformation energy Ep, was equal
of momentum is expressed by Eq. (6) to the initial kinetic energy of drop hammer Ek. In Stage 3, the

© ASCE 04019183-8 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


kinetic energy Ek3 was equal to the elastic deformation energy by the inertial effect, whereas the second and the third items show
released by the specimen the load resisted by catenary and flexural actions, respectively
sffiffiffiffiffiffiffiffi
1 2Ek M L;mid ðtÞ
Ek ¼ md v2d0 ¼ Ep þ Ee ⇒ vd0 ¼ ð12Þ V L;mid ðtÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð16Þ
2 md 14722 þ δ 2 ðtÞ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi −M L;con ðtÞ
1 2Ee V L;con ðtÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð17Þ
Ek3 ¼ ðmd þ αmb Þv23 ¼ Ee ⇒ v3 ¼ ð13Þ 14722 þ δ 2 ðtÞ
2 md þ αmb
ML;mid ðtÞ − M L;con ðtÞ
Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

where αmb = equivalent mass of the beam. The coefficient α is V L ðtÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð18Þ
equal to 0.375–0.5 for different boundary conditions, according 14722 þ δ2 ðtÞ
to the study from Kaewkulchai and Williamson (2006). Combining
Eqs. (10)–(13) M mid;L ðtÞ − M con;L ðtÞ
RV;L ðtÞ ¼ N L ðtÞ sin α1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos α1 ð19Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 14722 þ δ 2 ðtÞ
t2 md vd0 FS 1 vd0 1 2Ek ðmd þ αmb Þ
¼ ¼ ¼
t3 FS 2md v3 2 v3 2 md 2Ee PðtÞ ¼ FI ðtÞ þ RV;L ðtÞ þ RV;R ðtÞ ð20Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 Ek ðmd þ αmb Þ PðtÞ ¼ FI ðtÞ þ N L ðtÞ sin α1 þ N R ðtÞ sin α2
¼ ð14Þ |ffl{zffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
2 Ee md Inertial effect Catenary action

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ½M mid;L ðtÞ − Mcon;L ðtÞ ½M ðtÞ − M con;R ðtÞ


For this study, the ðmd þ αmb Þ=md was set as 1, because md þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos α1 þ mid;R pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
was significantly larger than αmb . Then, the ratio of t2 and t3 can be 14722 þ δ 2 ðtÞ 14722 þ δ2 ðtÞ
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
simply expressed as Flexural action
sffiffiffiffiffiffi ð21Þ
t2 1 E k
¼ ð15Þ where M L;mid ðtÞ and M L;con ðtÞ = bending moments at the left mid-
t3 2 E e
span and left connection based on the strain gauges at four sections,
According to this equation, if Ek is greater than four times Ee , as shown in Fig. 6, respectively; V L;mid ðtÞ and V L;con ðtÞ = shear
the duration of Stage 2 should be longer than that of Stage 3. Based forces on the left side based on the bending moments of the left
on the experimental results, the Ee values of WUF-B, RBS, midspan and left connection, respectively; V L ðtÞ = shear force
RC-EEP, and RC-FEP were 2.8, 2.7, 2.2, and 2.3 kJ, respectively, on the left side; δðtÞ = vertical displacement at midspan; α1 ðtÞ
and the Ee value of FP was 1.5 kJ. The Ek value of WUF-B, RBS, and α2 ðtÞ = angles between the moving beam axis and the origi-
RC-FEP, and RC-EEP was 24.4 kJ and that of FP was 17.3 kJ, nal horizontal beam axis on the left and right sides, respectively;
which is greater than four times that of the Ee . Thus, Stage 2 had RV;L ðtÞ and RV;R ðtÞ = resistance provided by the deformation on the
a longer time duration than Stage 3. left and right halves of the specimens, respectively; FI ðtÞ = inertia
force; and PðtÞ = applied force. On the right side, the expressions
for the shear force V R ðtÞ, the shear force provided by right midspan
Load-Resisting Mechanism V R;mid ðtÞ and right connection V R;con ðtÞ, and the resistance pro-
vided by deformation RV;R ðtÞ were not presented because they have
Fig. 9(a) shows the residual deflection profiles of specimens after the same form as the left side due to symmetry.
impact. All specimens had the same global deformation shape.
During the impact process, the midspan of the beam moved down-
Load-Resisting Histories
ward. The angle (α1 ) between the moving beam axis and its origi-
nal horizontal location continued to increase, as shown in Fig. 9(b). Based on the load-resisting mechanism, the applied force-time his-
Only the left side of the specimens is shown in Fig. 9(b) due to tory of all five specimens is shown in Fig. 10 with various parts of
symmetry. Two parts, viz. inertial effect and deformation, including resistance. All specimens had a similar impact process, which could
catenary action and flexural action, resisted the impact load applied be divided into three stages, as mentioned previously. In the first
to specimens. Fig. 9(c) shows the force equilibrium of specimens stage, the impact force reached maximum value within a very short
under impact load. The axial force NðtÞ and bending moment time. The flexural action developed in this stage. The drop hammer
at midspan M mid ðtÞ and connection Mcon ðtÞ were calculated by had obtained considerable acceleration when it contacted the speci-
Eqs. (1)–(4) based on the strain gauges of the four sections in Fig. 6. men. When the impact load reached peak value, approximately
On the left side of the specimen, the shear force provided by mid- 83%–94% of the applied load was resisted by inertia force. The
span V L;mid ðtÞ and connection V L;con ðtÞ was calculated by Eqs. (16) peak value of the impact load was primarily resisted by inertia
and (17). The flexural action was represented by shear force V L ðtÞ effect. During the second stage, the oscillation of inertial effect
in Eq. (18). The vertical resistance provided by deformation on the decreased to near zero, and flexural action reached a steady value.
left side could be obtained by Eq. (19). On the right side, the ver- The impact load was primarily controlled by flexural action. The
tical resistance provided by deformation RV;R ðtÞ, i.e., the shear resistance provided by the flexural action of five specimens was
force V R ðtÞ, the shear force provided by midspan V R;mid ðtÞ and more than 97% of the applied load. According to Eqs. (16)–(18),
connection V R;con ðtÞ could be acquired in the same way as the left flexural action depended on the bending moment at midspan and
side. Substituting axial and shear forces into Eq. (20), which was connections. The five specimens had the same beam section but
obtained from the force equilibrium shown in Fig. 9(c), the load- different beam–column connections, which means that connection
resisting mechanism of steel frames under impact load was ob- types primarily controlled the impact load. After reaching maxi-
tained as Eq. (21). In Eq. (21), the first item shows the load resisted mum displacement, specimens entered the third stage. The velocity

© ASCE 04019183-9 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

(a)

(b)

(c)

Fig. 9. Load-resisting mechanism under impact: (a) residual deflection profiles of specimens; (b) deformation geometry of left side; and (c) force
equilibrium of specimens.

of the drop hammer was zero at maximum displacement. Impact Deformation Modes
energy was completely transformed into other forms, such as the
Fig. 11 shows the deformation of each specimen after impact.
deformation of specimens. Deformation included two parts, viz. Deformation included axial and bending deformations at mid-
elastic and plastic deformation. Specimens rebounded due to elastic span and connections. According to Figs. 7(c and d), the sagging
deformation after maximum displacement. Similar to the second moment appeared at the midspan, and the hogging moment ap-
stage, the impact load was primarily contributed by flexural action peared at the connections. As all specimens had the same beam
in this stage. The catenary action provided almost no contribution section, only the midspan of WUF-B was demonstrated, as shown
to resisting the impact load during the entire impact process, even if in Fig. 11(a). Under the sagging moment, the upper and lower
obvious axial force appeared in all five specimens, as shown in flanges were under compressive force and tensile force, respec-
Fig. 7(e). This is because maximum displacement was too small tively. An arch appeared at the upper flange near the midspan at
to provide effective resistance. For instance, maximum displace- each side. All connections were under the hogging moment dur-
ment of the RC-EEP specimen was 102 mm, and the corresponding ing the impact process. The upper and lower flanges were under
axial force was 28 kN. According to Fig. 9(c) and Eq. (19), sin α tensile force and compressive force, respectively. For WUF-B and
was 0.073, and the resistance provided by catenary action was RBS, beam flanges and columns were welded. The arches ap-
4 kN. For WUF-B, the maximum value of axial force was 194 kN peared at the lower flange, as shown in Figs. 11(b and c). For
with a corresponding displacement of 17 mm. The calculated sin α FP, the beam web and the fin plate were connected by bolts. Origi-
was 0.012, and the resistance provided by axial force was 5 kN. In nal gaps appeared between beam and columns. After impact, the
summary, the applied load was primarily resisted by inertial effect gap was opened at the upper flange and closed at the lower flange.
in the first stage. From the beginning of the second stage, the pri- In the fin plate, a hump appeared at the right of the upper bolt.
mary resistance changed from inertial effect to flexural action. The obvious bending deformation appeared on the connection, as
After that, flexural action played the key role in resisting impact shown in Fig. 11(d). Fig. 11(e) shows the deformation of RC-FEP.
until the end of the impact process. A gap was opened between the channel web and the end plate

© ASCE 04019183-10 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

(c) (d)

(e)

Fig. 10. Load-resisting time histories: (a) WUF-B; (b) RBS; (c) FP; (d) RC-FEP; and (e) RC-EEP.

near the upper flange. However, no arch was found at the lower connection completely broke in FP, RC-FEP, and RC-EEP. For
beam flange under compressive force because the deformation WUF-B and RBS, out-of-plane deformation appeared at the impact
was concentrated on the channel. At the top of the channel, a con- location after several moments of impact. Although the connections
vex form appeared under tensile force along the beam axial, as in these two specimens did not break, there were fractures in the
shown in Fig. 11(f). A concave form under compressive force large deformation area. For safety reasons, no more impact was
appeared at the bottom of the channel, as shown in Fig. 11(g). applied to these two specimens.
The deformation of RC-EEP was similar to that of RC-FEP, as
shown in Fig. 11(h). All specimens showed typical bending de-
formation, because flexural action played a key role in resisting WUF-B
impact load, as shown in Fig. 10. A force-displacement curve could be obtained by combining force-
time and displacement-time histories. The load versus displacement
curves of WUF-B is shown in Fig. 12(a), with three parts resisted
Test Results under Multiple Impacts by the two actions provided by deformation, i.e., flexural action
provided by midspan (FAM), flexural action provided by connec-
After the first impact, no failure was found in any of the five tions (FAC), and catenary action (CA). The specimen subject to
specimens. To investigate the ultimate capacity of specimens impact load experienced an energy conversion process. Large de-
under impact as far as possible, multiple impacts were applied formation of the specimen was the primary method to absorb im-
to each specimen. Among the five specimens, one beam–column pact energy. The energy absorption of each part, i.e., the bending

© ASCE 04019183-11 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Deformation after first impact: (a) midspan (WUF-B); (b) WUF-B; (c) RBS; (d) FP; (e) RC-FEP; (f) Top of channel (RC-FEP); (g) Bottom of
channel (RC-FEP); and (h) RC-EEP.

(a) (b)

(c) (d)

Fig. 12. Comprehensive results of WUF-B: (a) force-displacement curves; (b) energy absorption curves; (c) fracture at upper flange (top view); and
(d) fracture at upper flange (bottom view).

deformation at midspan, the bending deformation at connections, WUF-B is a typical semirigid connection that could provide an
and the axial deformation, was obtained by integrating the force- effective bending moment. A total of five impacts were applied
displacement curves of FAM, FAC, and CA, respectively, as shown to the WUF-B specimen. In Fig. 12(a), the load applied on the
in Fig. 12(b). specimen rapidly increased at the initial stage. After that, the FAM

© ASCE 04019183-12 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


entered the steady stage, and the FAC still slowly increased. Both the length of the beam near the ends of the beam span to satisfy the
the strength and stiffness of the connections were less than those ductility requirements in seismic design. Figs. 13(a and b) show the
at midspan. After 89 mm, both FAM and FAC were in the steady force-displacement and energy absorption curves of RBS, respec-
stage, and CA was mobilized due to large deformation, which con- tively. In Fig. 13(a), FAM and FAC rapidly increased at the begin-
tinued to increase until the end. Under midspan loading, connec- ning and then entered the steady stage. The catenary action was
tions were combined with the axial force and the hogging moment. mobilized after 75 mm. The load rapidly increased in the initial
The maximum tensile force appeared at the upper flange at the stage and then steadily increased to 168 mm. Before 168 mm, the
beam end. After the fifth impact, the upper flange at the right beam development of load and resistance was similar to WUF-B. The
end first cracked, as shown in Figs. 12(c and d). Because the frac- load was less than that of WUF-B because the bending moment
ture only appeared in less half of the flange and occurred at the at the connection was less than WUF-B due to the reduced beam
Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

end stage of the fifth impact, the FAC did not obviously decrease. section, as shown in Fig. 7(d). At 168 mm, the left reduced beam
In the end, the load reached 479 kN at 221 mm. The corresponding section cracked from the upper flange to the web due to tensile
resistances of FAM, FAC, and CA were 217 kN (45%), 151 kN force, which was produced by the combination of axial force and
(32%), and 111 kN (23%), respectively. In Fig. 12(b), the energy hogging moment, as shown in Figs. 13(c and d). Therefore, the
absorbed by bending deformation at midspan and connections was lever arm of the bending moment at the left connection decreased
58% and 41%, respectively, whereas the axial deformation was 1% so that FAC started decreasing. The load dropped suddenly due to
at the beginning of the catenary action. After that, FAM and FAC the fracture of the left reduced section. The FAC was primarily pro-
maintained a steady value, and CA increased with increasing dis- vided by the right connection. After 168 mm, the FAM and FAC
placement. The ratio of energy absorbed by axial deformation con- remained steady, whereas the load increased with increasing CA.
tinuously increased and reached 8% at 221 mm. The ratio of energy At 240 mm, although the load resisted by CA (130 kN) was more
absorbed by axial deformation was significantly less than that of than two times that at 168 mm (52 kN), the load (352 kN) was still
the load resisted by catenary action. The bending deformation ab- less than the maximum load (401 kN at 168 mm) due to decreasing
sorbed the most energy (52% at midspan and 40% at connections). flexural action. The energy absorption curve of RBS is shown in
After the fifth impact, serious out-of-plane distortion appeared Fig. 13(b). Before the mobilization of catenary action, almost all
at midspan, and the drop hammer did not align with the center of energy was absorbed by bending deformation at midspan and con-
the upper flange. Although the right connection did not break com- nections. With the development of catenary action after 75 mm, the
pletely, no more impact was applied to WUF-B for safety reasons. ratio of energy absorbed by axial deformation increased. In the
end, compared to WUF-B, the ratio of energy absorbed by axial
deformation (10%) increased, and the ratio of energy absorbed by
RBS
bending deformation at connections (37%) decreased, because the
Based on WUF-B, the RBS utilized circular radius cuts in both left connection could not effectively absorb energy after cracking.
upper and lower flanges of the beam to reduce the flange area over Although the final load resisted by CA was more than two times

(a) (b)

(c) (d)

Fig. 13. Comprehensive results of RBS: (a) force-displacement curves; (b) energy absorption curves; (c) fracture at left reduced section (front view);
and (d) fracture at left reduced section (back view).

© ASCE 04019183-13 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

(c) (d)

Fig. 14. Comprehensive results of FP: (a) force-displacement curves; (b) energy absorption curves; (c) left connection after Impact 3 (front view); and
(d) left connection after Impact 4 (front view).

that of FAC, bending deformation at connection absorbed almost in Fig. 15(a). Because RC-FEP could not provide an effective bend-
four times the energy than axial deformation. ing moment, FAC stayed at a low level until the end. During the
second impact, a large deformation appeared at the channel web
FP and the end plate corresponding to the upper beam flange, as shown
in Fig. 15(c). After 121 mm, the catenary action significantly in-
FP is a typical single connection. Compared to WUF-B, beam creased with increasing vertical displacement. The flexural action
flanges were not welded to columns, whereas only the beam web decreased with the development of plastic deformation at the con-
was connected to the column by a fin plate. The force-displacement nections. At 230 mm, CA provided the same resistance as FAM.
and energy absorption curves are shown in Figs. 14(a and b). The gap between the end plate and channel web at the location of
In Fig. 14(a), FAM provided primary resistance until the end. After the upper beam flange developed while a new gap at the lower
68 mm, catenary action was mobilized and provided partial resis- flange formed due to significant tensile axial force, as shown in
tance. At 157 mm, the load reached the maximum value of 230 kN. Fig. 15(d). The load reached a maximum value of 339 kN at
After the third impact, a large deformation appeared at the fin plate. 265 mm. At 284 mm, failure occurred at the left connection. The
The first fracture was found at the left fin plate between the upper final failure mode was the fracture of the reverse channel at the web
and middle bolt holes, as shown in Fig. 14(c). After 157 mm, the near the flange/web junctions under tensile axial force, as shown
load remained steady until final failure at 207 mm. Finally, the de-
in Fig. 15(e). Fig. 15(b) shows the energy absorption curves of
formation was concentrated on the fin plate, whereas no obvious
RC-FEP. The final ratios of energy absorption by midspan, connec-
deformation was found on the beam section at the connection. The
tions, and axial deformation were 65%, 9%, and 26%, respectively.
fracture on the left fin plate showed a reverse L shape under the
Because RC-FEP had better deformation capacity, catenary action
combination of bending moment and axial force, as shown in
developed better in RC-FEP than in FP.
Fig. 14(d). In Fig. 14(b), bending deformation at midspan absorbed
the most energy (83%). The ratio of energy absorbed by connection
was 6% because FAM maintained a very low level from begin- RC-EEP
ning to end. Axial deformation absorbed more energy (11%) than
connections. Compared to RC-FEP, RC-EEP employed an extended instead of
a flush end plate. The extended part increased the lever arm of the
bending moment and enhanced the moment capacity of the con-
RC-FEP nection. The flexural action of RC-EEP was greater than that of
Reverse channel connection was a new type of beam–column RC-FEP. In Fig. 16(a), the catenary action began at 102 mm and
connection proposed by Wang and Xue (2013). This connection increased to 330 mm. The flexural action remained stable. The load
could be designed as either a single or a semirigid connection. increased with increasing catenary action and reached the maxi-
The development of the load and resistance of RC-FEP is shown mum value (455 kN) at 330 mm. After 330 mm, a fracture appeared

© ASCE 04019183-14 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

(c) (d)

(e)

Fig. 15. Comprehensive results of RC-FEP: (a) force-displacement curves; (b) energy absorption curves; (c) left connection after Impact 2;
(d) left connection after Impact 3; and (e) left connection after Impact 4.

and developed at the channel web, as shown in Fig. 16(c). Finally, increasing displacement. The load resistance was provided by the
the upper half of the web separated from the reverse channel, and combination of flexural action and catenary action. Because the five
the lower bolts were pulled out of their bolt holes, as shown in specimens had the same section at midspan, flexural resistance de-
Fig. 16(d). Fig. 16(b) demonstrates the energy absorption curves pended on the strength of the beam–column connections. The spec-
of RC-EEP. Compared to RC-FEP, RC-EEP had greater FAC. imen with the beam–column connection of the highest (WUF-B)
CA fully developed in RC-EEP due to good deformation capacity. and the lowest (FP) bending moments had the highest and the low-
Thus, the energy absorbed by axial deformation reached 26%, est flexural resistance, respectively, as shown in Fig. 17(a). It was
which was the highest energy absorption ratio in all five specimens obvious that catenary resistance depended on deformation rather
as that of RC-FEP. than connection types. Before failure occurred, in Fig. 17(b), the
catenary resistance of different specimens had a growing trend. The
greatest catenary resistance appeared on the specimen with the best
Comparison and Discussion ductility (RC-EEP). The load-carrying capacity with and without
The load resisting-displacement and the energy absorption- catenary action of the five specimens is demonstrated in Table 4,
displacement curves of the specimens are compared in Figs. 17 which clearly shows that WUF-B had the highest load-carrying
and 18, respectively. As Fig. 17 shows, the specimens in these five capacity due to its greatest flexural strength. RBS had the second
tests went through two stages. During the small deformation stage, greatest flexural resistance among the five specimens, but its load-
the specimens sustained applied loads by flexural action. After carrying capacity was lower than that of RC-EEP, which had
the small deformation stage, catenary action formed rapidly with the best catenary resistance due to its best ductility. Although the

© ASCE 04019183-15 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

(c) (d)

Fig. 16. Comprehensive results of RC-EEP: (a) force-displacement curves; (b) energy absorption curves; (c) right connection after Impact 6; and
(d) right connection after Impact 7.

(a) (b)

Fig. 17. Comparison of load resistance of different specimens: (a) flexural resistance-displacement curves; and (b) catenary resistance-displacement
curves.

load-carrying capacity of RC-FEP was improved by 35% by cat- resistance, energy absorption by specimens was also divided into
enary action, it was still lower than that of RBS. FP had the lowest two parts, i.e., energy absorption of bending and axial deformation.
load-carrying capacity due to the weakest beam–column connec- In Fig. 18(a), the energy absorption curves approximated a straight
tion and poor ductility. line, and the slope was positively correlated with flexural strength,
Structures subject to impact load for this research experienced as shown in Fig. 17(a). Because catenary action was related to
an energy-absorption process. During the impact process, impact deformation, the energy absorption curves of axial deformation
energy was absorbed by the deformation of specimens. The capac- overlapped. Energy absorbed by axial deformation depended on
ity for energy absorption is an important index to evaluate the the final displacement, as shown in Fig. 18(b). Total energy absorp-
impact resistance of structures, and it is equal to the virtual work tion with and without axial deformation is shown in Table 4. The
done by the resistance of each part. The energy absorption capac- energy absorption of specimens with larger deformation (RC-FEP
ity of each specimen was obtained from the integral of the force- and RC-EEP) was greatly improved by axial deformation. More-
displacement curve, as shown in Fig. 18. Corresponding to load over, although WUF-B had the greatest load-carrying capacity,

© ASCE 04019183-16 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

Fig. 18. Comparison of energy absorption of different specimens: (a) energy absorption of flexural deformation; and (b) energy absorption of axial
deformation.

Table 4. Comparison with and without catenary action


Ultimate Load capacity Load capacity Load capacity Energy Energy Energy
vertical with without increase caused absorption absorption absorption
Specimen displacement catenary catenary by catenary with axial without axial increase caused by
identification (mm) action (kN) action (kN) action (%) deformation (kJ) deformation (kJ) axial deformation (%)
WUF-B 227 479 414 16 82.5 76.2 8
RBS 241 401 364 10 75.4 67.9 11
FP 235 230 215 7 43.2 38.6 12
RC-FEP 302 339 252 35 71.2 52.8 35
RC-EEP 394 455 311 46 121.6 89.7 36

RC-EEP had the best capacity for energy absorption, which means steady impact force finished in a shorter time. Under the single
that load-carrying capacity cannot independently confirm impact impact load, the maximum displacement of all specimens was
resistance. The ductility of the specimen was also a factor influenc- less than half of the height of the beam section. Although ob-
ing the capacity of energy absorption. vious axial force occurred in the specimens, it did not signifi-
In addition, the total impact energies of WUF-B, RBS, FP, cantly affect the impact force. The greater steady impact force
RC-FEP, and RC-EEP were 122.0, 109.8, 69.4, 97.6, and 170.8 kJ, appeared on the specimen with beam–column connections with
respectively. The corresponding energies absorbed by deformation higher bending moment resistance.
were 82.5, 75.4, 43.2, 71.2, and 121.6 kJ, respectively. There was a 2. Under a single impact, all specimens had the same global
difference between total impact energy and total energy absorbed deformation shape. The load-resisting mechanism was in-
by deformation. The difference can be attributed to the energy dis- vestigated. The applied load was resisted by inertial effect and
sipated by the test setup; the energy loss due to the rebound of the deformation, including catenary action and flexural action.
drop hammer; and the sound, light, and heat dissipation during Flexural action was provided both at midspan and by connec-
the impact processes. The impact energy was always larger than tions. During the first stage, the global deformation was limited.
the energy absorbed by deformation. When the impact load reached peak value, approximately 83%–
94% of the applied load was resisted by inertia force. The peak
value of the impact load was primarily resisted by inertia effect.
Conclusions During the following stages, flexural action played the key role.
More than 97% of the applied load was resisted by flexural ac-
This paper focused on steel-framed structures subjected to impact tion. Catenary action did not fully mobilize due to limited de-
loads. A series of experimental tests of steel frames with different formation. The resistance provided by catenary action was less
beam–column connections under both single and multiple impacts than 5 kN, which was less than 2% of the applied load. In short,
were carried out. The results in this paper were obtained from flexu- the applied load was primarily resisted by inertial effect at the
ral mode under the midspan-impact scenario. Further studies are first stage and by flexural action in following stages.
needed to obtain the impact resistance of specimens under other 3. No failure occurred on any of the specimens after first impact.
scenarios. Conclusions of this work can be drawn as follows: Multiple impacts were applied to each specimen to investigate
1. The process of specimens under single impact can be divided ultimate impact resistance. For each specimen, the load-carrying
into three stages. The peak impact force appeared at the first capacity was discussed. Load was resisted primarily by flexural
stage. In the second stage, the impact force maintained a steady action at the small deformation stage. Specimens sustained ap-
value and decreased to zero at the third stage. The comparison of plied load by a combination of flexural action and catenary ac-
steady impact force and maximum displacement showed that tion at the large deformation stage. The specimen with a higher
greater steady impact force reduced the deformation of speci- bending moment resistance connection and better deformation
mens to resist the same impact load. The specimen with greater capacity had greater load-carrying capacity.

© ASCE 04019183-17 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183


4. Energy absorption of a specimen was more important than its GSA (General Services Administration). 2013. Alternate path analysis
load-carrying capacity to evaluate impact resistance because and design guidelines for progressive collapse resistance. Washington,
the structures subject to impact load experienced an energy- DC: GSA.
absorption process. Although WUF-B had the greatest load- Haas, K., M. Bychkowski, and S. Marjanishvili. 2018. “A review of analy-
sis methods for falling debris resulting from structural collapse.” In
carrying capacity, RC-EEP absorbed more impact energy due to
Proc., Structures Congress 2018. Reston, VA: ASCE.
its development of large deformation. Both strength and ducti- Jawdhari, A., S. Orton, B. Izzuddin, and D. Cormie. 2018. “Dynamic debris
lity were important for the capacity of energy absorption. FP had loading in flat plate and steel framed buildings.” In Proc., Structures
the worst impact resistance due to its low load-carrying capacity Congress 2018. Reston, VA: ASCE.
and poor ductility. In addition, impact energy is always larger Kaewkulchai, G., and E. B. Williamson. 2006. “Modeling the impact of
than energy absorbed by deformation. failed members for progressive collapse analysis of frame structures.”
Downloaded from ascelibrary.org by Addis Ababa Science and Technology University on 12/26/19. Copyright ASCE. For personal use only; all rights reserved.

5. The failure mode of FP was a reverse L-shape fracture at the left J. Perform. Constr. Facil. 20 (4): 375–383. https://doi.org/10.1061
connection. The fractures of both WUF-B and RBS appeared /(ASCE)0887-3828(2006)20:4(375).
on the upper flange of the beam section. The first fracture of Khandelwal, K., and S. EI-Tawil. 2007. “Collapse behavior of steel special
WUF-B occurred at the beam end, while that of RBS occurred moment resisting frame connections.” J. Struct. Eng. 133 (5): 646–655.
at the reduced beam section. The extended end plate changed https://doi.org/10.1061/(ASCE)0733-9445(2007)133:5(646).
Lew, H. S., J. A. Main, S. D. Robert, F. Sadek, and V. P. Chiarito. 2013.
the failure mode of the reverse channel connection. The failure
“Performance of steel moment connections under a column removal
mode of RC-FEP was the laceration on the whole web of the scenario. I: Experiments.” J. Struct. Eng. 139 (1): 98–107. https://doi
parallel flange channel, while that of RC-EEP was the laceration .org/10.1061/(ASCE)ST.1943-541X.0000618.
on the upper half of the web of the parallel flange channel and Liu, C., T. C. Fung, and K. H. Tan. 2016. “Dynamic performance of
the lower bolts pulling out of their bolt holes. flush end-plate beam-column connections and design applications in
6. The development of catenary action depended on ductility in- progressive collapse.” J. Struct. Eng. 142 (1): 04015074. https://doi.org
stead of strength of connections. If the specimen had good duc- /10.1061/(ASCE)ST.1943-541X.0001329.
tility, catenary action significantly improved both load-carrying Liu, C., K. H. Tan, and T. C. Fung. 2013. “Dynamic behaviour of web cleat
capacity and energy absorption. connections subjected to sudden column removal scenario.” J. Constr.
Steel Res. 86 (Jul): 92–106. https://doi.org/10.1016/j.jcsr.2013.03.020.
Liu, C., K. H. Tan, and T. C. Fung. 2015. “Investigations of nonlinear
Acknowledgments dynamic performance of top-and-seat with web angle connections sub-
jected to sudden column removal.” Eng. Struct. 99 (Sep): 449–461.
The authors gratefully acknowledge the financial support provided https://doi.org/10.1016/j.engstruct.2015.05.010.
by the Fundamental Research Funds for the Central Universities in Olmati, P., J. Sagaseta, D. Cormie, and A. E. K. Jones. 2017. “Simplified
reliability analysis of punching in reinforced concrete flat slab buildings
China (No. 2019CDQYTM037), the National Natural Science
under accidental actions.” Eng. Struct. 130 (Jan): 83–98. https://doi.org
Foundation of China (51778086), the program of China Scholar- /10.1016/j.engstruct.2016.09.061.
ships Council (No. 201506050087), and the Ministry of Home Af- Ribeiro, J., A. Santiago, C. Rigueiro, P. Barata, and M. Veljkovic. 2016.
fairs in Singapore. The test specimens were supported by the Prof “Numerical assessment of T-stub component subjected to impact
Tan Kang Hai’s MHA project The Structural Resilience Study: A loading.” Eng. Struct. 106 (Jan): 450–460. https://doi.org/10.1016/j
Study of Concrete Pre-cast and Composite Steel Joints subject to .engstruct.2015.10.047.
Missing Column Scenario. All the tests were conducted by Wang Sadek, F., J. A. Main, H. S. Lew, and S. EI-Tawil. 2013. “Performance
Hao while he was at Nanyang Technological University, School of of steel moment connections under a column removal scenario. II:
Civil and Environmental Engineering. Analysis.” J. Struct. Eng. 139 (1): 108–119. https://doi.org/10.1061
/(ASCE)ST.1943-541X.0000617.
Vlassis, A. G., B. A. Izzuddin, A. Y. Elghazouli, and D. A. Nethercot. 2009.
References “Progressive collapse of multi-storey buildings due to failed floor
impact.” Eng. Struct. 31 (7): 1522–1534. https://doi.org/10.1016/j
Al-Rifaie, A., Z. W. Guan, S. W. Jones, and Q. Wang. 2017. “Lateral .engstruct.2009.02.009.
impact response of end-plate beam–column connections.” Eng. Struct. Wang, H., B. Yang, X. H. Zhou, and S. B. Kang. 2016. “Numerical analy-
151 (Nov): 221–234. https://doi.org/10.1016/j.engstruct.2017.08.026. ses on steel beams with fin-plate connections subjected to impact
Chen, J., W. Shu, and H. Huang. 2017. “Rate-dependent progressive col- loads.” J. Constr. Steel Res. 124 (Sep): 101–112. https://doi.org/10
lapse resistance of beam-to-column connections with different seismic .1016/j.jcsr.2016.05.016.
details.” J. Perform. Constr. Facil. 31 (2): 04016086. https://doi.org/10 Wang, Y. C., and L. Xue. 2013. “Experimental study of moment-rotation
.1061/(ASCE)CF.1943-5509.0000922. characteristics of reverse channel connections to tubular columns.”
DoD (Department of Defense). 2009. Design of buildings to resist J. Constr. Steel Res. 85 (Jun): 92–104. https://doi.org/10.1016/j.jcsr
progressive collapse. Washington, DC: Dept. of Defense. .2013.03.006.
Fujikake, K., B. Li, and S. Soeun. 2009. “Impact response of reinforced Yang, B., and K. H. Tan. 2012. “Numerical analyses of steel beam–column
concrete beam and its analytical evaluation.” J. Struct. Eng. 135 (8): joints subjected to catenary action.” J. Constr. Steel Res. 70 (Mar):
938–950. https://doi.org/10.1061/(ASCE)ST.1943-541X.0000039. 1–11. https://doi.org/10.1016/j.jcsr.2011.10.007.
Grimsmo, E. L., A. H. Clausen, A. Aalberg, and M. Langseth. 2016. Yang, B., and K. H. Tan. 2013a. “Experimental tests of different types
“A numerical study of beam-to-column joints subjected to impact.” Eng. of bolted steel beam–column joints under a central-column-removal
Struct. 120 (Aug): 103–115. https://doi.org/10.1016/j.engstruct.2016 scenario.” Eng. Struct. 54 (Sep): 112–130. https://doi.org/10.1016/j
.04.031. .engstruct.2013.03.037.
Grimsmo, E. L., A. H. Clausen, M. Langseth, and A. Aalberg. 2015. “An Yang, B., and K. H. Tan. 2013b. “Robustness of bolted-angle connections
experimental study of static and dynamic behaviour of bolted end-plate against progressive collapse: Experimental tests of beam-column joints
joints of steel.” Int. J. Impact Eng. 85 (Nov): 132–145. https://doi.org and development of component-based models.” J. Struct. Eng. 139 (9):
/10.1016/j.ijimpeng.2015.07.001. 1498–1514. https://doi.org/10.1061/(ASCE)ST.1943-541X.0000749.

© ASCE 04019183-18 J. Struct. Eng.

J. Struct. Eng., 2020, 146(1): 04019183

You might also like