You are on page 1of 16

Engineering Structures 182 (2019) 13–28

Contents lists available at ScienceDirect

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

Review article

Calculation of the resistance of an unequal span steel substructure against T


progressive collapse based on the component method
Bao Meng, Weihui Zhong , Jiping Hao, Xiaoyan Song, Zheng Tan

School of Civil Engineering, Xi’an University of Architecture and Technology, Xi’an 710055, China

ARTICLE INFO ABSTRACT

Keywords: The rapid quantitative assessment of a structure’s resistance is critical to the structural design process in order to
Steel frame prevent the occurrence of progressive collapse. In this study, a three-bar component model of a top-seat angle
Unequal span steel substructure with a double web angle connection was constructed based on the component method by dividing and sim-
Top-seat angle with double web angle plifying its geometry. The proposed component model was then applied to a model of an unequal span sub-
connection
structure. An explicit calculation method and a numerical simulation method were proposed to describe the
Progressive collapse
nonlinear static responses of the unequal span substructure under a scenario of interior column failure. These
Component method
two methods were carefully validated against previously collected experimental data and were determined to
accurately reflect the main responses of the structure, including the load–displacement relationship, develop-
ment of internal forces, and deformation behaviour. The explicit calculation procedure, conducted using a step-
by-step process in an Excel spreadsheet, was successfully applied to rapidly analyse and quantitatively evaluate
the resistance of the unequal span steel substructure, which will lay the foundation for the resistance prediction
and performance analysis of the whole structure against progressive collapse.

1. Introduction APM is an important and widely applied method for the analysis
and design of building structures to prevent progressive collapse, which
Since the well-known collapse of the Ronan Point apartment has the advantage of easy acceptance and quantitative analysis [5]. In
building in 1968, the prevention of progressive collapse has been the APM, a main vertical bearing member (column) is artificially removed,
focus of engineers and researchers. Progressive collapse is usually de- and then, the robustness of the structural system is evaluated in terms
fined as a disproportionate collapse of an entire structure or its part, of whether the remaining structure has sufficient resistance to prevent
due to the local failure of one or more of its structural members. progressive collapse. Based on the concept of APM, Izzuddin et al. [10]
Because of the large deformations involved, material nonlinearity and developed a basic framework that allows the simplification of the level
contact nonlinearity are significant during progressive collapse, of structural representation used in analysis, in which the responses at
meaning that the traditional elastic design method is no longer ap- higher levels can be assembled from the responses at lower levels ac-
plicable, and a nonlinear collapse analysis is necessary. In principle, a cording to a simple multi-level approach. In addition, by the principle
complicated dynamic process dependent on the nature of different of energy balance, the dynamic effects are considered approximately by
causes of failure should be considered in the design of the progressive nonlinear static responses [10]. For example, the maximum dynamic
collapse of a building structure. Conducting a nonlinear dynamic Finite displacement corresponding to the dynamic load can be determined
Element analysis of the structure can produce more representative re- from the static load–displacement curve. Therefore, it is critical to ob-
sults [1–4]. While such analyses can identify and simulate the collapse tain the nonlinear static responses of the structure.
damage in an actual situation, they may require huge computing re- When the ductility of the beam-column connection is good and a
sources and specialized skills to perform them. Hence, such studies can reliable mechanism for the transfer of tensile force between beam and
be used as research tools or approaches to apply in certain situations column is provided, the remaining structure can withstand and redis-
and may be not suitable for general designers not used to engaging such tribute the external load by flexural action, arching action, flexural-
a specialized task [5]. For this reason, design specifications and catenary mixed action, and catenary action [14]. Hence, beam-column
guidelines [6–9] advocate for relatively simple approaches, one of joints are key components in a steel frame structure designed to resist
which is the alternative load path method (APM). progressive collapse. In this case, the component method becomes a


Corresponding author.
E-mail addresses: zhongweihui1980@163.com, zhongwhgm@xauat.edu.cn (W. Zhong).

https://doi.org/10.1016/j.engstruct.2018.12.053
Received 20 July 2018; Received in revised form 15 December 2018; Accepted 15 December 2018
Available online 24 December 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
B. Meng et al. Engineering Structures 182 (2019) 13–28

Nomenclature ML bending moment of left beam end


MR bending moment of right beam end
A1, A2, A3, A4 calculation parameters of Eq. (22) NL axial force of left beam
A5, A6, A7 calculation parameters of Eq. (24) NL axial force of right beam
AbL cross-sectional area of the left beam P concentrated load of the failure column
AbR cross-sectional area of the right beam Pc force of equivalent compression spring
B1, B2, B3 calculation parameters of Eq. (25) Pe force of unloading point of equivalent spring
bc height of the column section P0 force of equivalent spring of bottom row
C1, C2, C3 calculation parameters of Eq. (26) Pm force of unloading point of equivalent spring
D1, D2 calculation parameters of Eq. (30) Pti force of i row of equivalent tensile spring
DL horizontal displacement of the left beam PL concentrated load assigned to left beam
D horizontal displacement of the right beam PR concentrated load assigned to right beam
d position of tensile rigid link PT total tensile force of connection
E1, E2, E3 calculation parameters of Eq. (32) qL uniformly distributed load of left beam
Eb elastic modulus of the beam qR uniformly distributed load of right beam
Ec elastic modulus of the column RC total compressive force of connection
e translation displacement of connection v horizontal displacement of connection
F1, F2, F3 calculation parameters of Eq. (34) vc deformation of equivalent compression spring
G1, G2, G3 calculation parameters of Eq. (40) vm deformation of unloading point of equivalent spring
h distance from centre of compression to intersection of Bar vp residual deformation of equivalent spring
1 and Bar 0. vL axial displacement of the left beam-column connection
hi distance from centre of compression to each equivalent vR axial displacement of the right beam-column connection
spring vti horizontal displacement of equivalent spring of i row
Ic cross-sectional moment of inertia of column v La axial deformation of the left beam
i identifying number of equivalent spring vRa axial deformation of the right beam
iL linear stiffness of the left beam v Lb axial deformation of axial constraint of the left beam end
iR linear stiffness of the right beam vRb axial deformation of axial constraint of the right beam end
k1 first slop of simplified bi-linear curve z position of the load application line.
k2 second slop of simplified bi-linear curve z1 position of the geometric centreline of connection
kc stiffness of equivalent compression spring δ displacement of the failure column
kti stiffness of each equivalent tensile spring θ1 relative rotation between Bar 2 and Bar 0
kac axial constraint stiffness provided by side column θ2 relative rotation between Bar 1 and Bar 2
kas equivalent axial stiffness of the beam end θL connection rotation of beam end
krs rotational stiffness of connection θR connection rotation of right end
l1 net span of the left beam θ total rotation of connection
l2 net span of the right beam αL, βL, γL, λL calculation parameters of Eq. (44)
lc length of the side column αR, βR, γR, λR calculation parameters of Eq. (45)
M bending moment of the beam end α, β, γ, λ calculation parameters of Eq. (46)
N axial force of the beam

useful tool for joint design and the study of joint behavior. In general, top-seat angle with a double web angle (TSDWA) connection based on
any joint can be simplified into three different zones: tension, com- the Eurocode 3 (EC3) component method [18] is first detailed by
pression, and shear [14]. In each region, the response of the joint is
composed of several deformation sources. In theory, if reasonable Failure area
physical characteristics are provided for each individual joint compo-
nent, the arbitrary joint form under an arbitrary load condition can be
Inflection Inflection
simulated by the component method [14]. During the past decade, the
point point
component method has primarily been used to analyse the seismic
performance of joints under bending and shear force interaction.
Inflection Inflection
However, with increasing research on steel frame joints, new challenges point point
for the component method have been identified, particularly how to
consider the influence of axial force and its interaction with bending
and shear in a collapse analysis, especially in the case of semi-rigid
joints with complex configurations.
It has also been found that the unequal lengths of two beams con-
nected to a failed column, as shown in Fig. 1, may have a great impact P
on the collapse resistance mechanism of a structure. Most current stu-
Lc / 2 Lc / 2

dies [12,15–17] have focused on cases with equal lengths of connected


beams, and the analysis of unequal span structures remains rare. For Failure column
some structures such as anisotropic structures and transforming struc- L1 L2
tures, unequal spans are common. The failure of a column in a region of Unequal span substructure
unequal spans can lead to serious progressive collapse, so it remains
necessary to evaluate the resistance of unequal span structures.
In this paper, the construction of a three-bar component model of a Fig. 1. Unequal span steel frame structure with a failed (removed) column.

14
B. Meng et al. Engineering Structures 182 (2019) 13–28

simplifying the geometrical composition of the connection. Then, this springs.


component model is applied to an unequal span substructure consisting In Fig. 2, the TSDWA spring model consists of an immovable support
of three columns and two beams [19], isolated from the rest of a steel Bar 0, two rotating rigid bars (Bar 1 and Bar 2), a rigid tension link, a
structure by assuming the inflection point of the side column to be rigid compression link, and a series of springs. The position of the im-
approximately in the middle of the storey height, as shown in Fig. 1. movable support Bar 0 corresponds to the centreline of the column, the
Through theoretical derivation of an explicit calculation method and position of rigid Bar 1 corresponds to the beam end, and rigid Bar 2 is
numerical analysis, the nonlinear static responses of the internal forces located between the column flange and top angle. According to the
(bending moment and axial force) are discussed in detail and the equivalent rule of EC3, the top/seat angle was converted into a T-stub
change in effective resistance of the unequal span substructure with the and the web angles were simplified into multiple T-stubs. The spring
increase in the displacement of the failed column is presented. The system contains the following individual component forces: bolt tension
results of the analysis are then verified using existing test results. The (bt), top/seat angle bending (tab/sab), web angle bending (wab), bolt
explicit calculation and numerical analysis methods can be rapidly and shear (bs), top/seat angle bearing (tabb/sabb), web angle bearing
effectively used to assess the behaviour of this type of connection and (wabb), beam flange bearing (bfbb), beam web bearing (bwbb), and
predict the resistance of steel frames with unequal spans in a pro- bolt slip (slip). Based on the geometry of the TSDWA connection, all
gressive collapse scenario. individual components of the first and bottom rows were aligned with
the bolt axes [23,24], a series of individual components at the centre-
2. Construction of the TSDWA connection component model line of the web bolts were simplified as a series spring group in tension,
and all individual components at the centreline of the bottom angle leg
The seismic behaviour of joints in steel frame structures has been were simplified as a series spring group in compression that transmits
studied previously and their progressive collapse performance is being all the compression force. In the end, the subject TSDWA connection
investigated gradually. The three-bar model, with its explicit physical was simplified as four rows of tension springs and a row of compression
meaning, has often been used to analyse the mechanical behaviour of a springs.
semi-rigid connection under the simultaneous action of variable In the TSDWA spring model, the tension zone is located between
bending moment and constant axial force [20]. However, the axial rigid Bar 1 and rigid Bar 2, and the compression zone is located be-
force in the beams is likely to develop disproportionately relative to the tween rigid Bar 2 and support Bar 0. The tension zone is composed of a
bending moment in the connection in the case of a progressive collapse. series of springs and the total tensile force of the spring model is
To this end, Stylianidis et al. [11] studied the internal relationship of a transmitted by a rigid tension link (kRT ) that acts only in tension.
frame structure with an end plate connection, but the mechanical Similarly, the internal compressive force of the spring model is trans-
performance of different connection forms can be different under con- mitted by a rigid compression link (kRC ), located at the centreline of all
ditions of progressive collapse. Therefore, in this section, the TSDWA compression forces, that acts only in compression. Because stiffeners
connection is evaluated using a three-bar model capturing clear me- are arranged on both sides of the column web, the deformation of the
chanical behaviour. This model is proposed based on the component column web and column flange would be relatively small. Therefore,
method, which is able to reflect the resistance transformation me- the deformation of the column flange and column web was not con-
chanism of a structure against vertical loads and the entire deformation sidered when creating the simplified spring model in this study.
development process after the failure of an internal column.

2.1. Spring model of the TSDWA connection 2.2. Mechanical behaviour of each individual component of the TSDWA
connection
The composition of a TSDWA connection can be simplified as the
spring model shown in Fig. 2. In this model, it is assumed that (1) the The force-displacement responses (constitutive relationship), in-
centre of compression of the connection, Point O, is located between the cluding the failure points, of each individual component in the TSDWA
centreline of the seat angle and the flange of the column [18,21] (see connection were calculated using the geometry and material properties
Fig. 1); and (2) the leg of the top angle connected to the column is an of the connection in conjunction with models proposed in previous
elastic member, while the leg of the seat angle connected to the beam is studies [24–29], and are listed in Table 1. In this study, the constitutive
a rigid member [22]. This spring model reflects the internal force and relationship of the bolt in tension was simplified as a bilinear model
deformation responses of this connection under large deformation and based on its yield point and ultimate point determined from material
can accurately simulate the combined effect of the different rows of test results.

At the end of the beam

tab/sab Between the column flange and top angle


The center line of r1
bt r0 2 slip Ba
wab
the column Ba r
Ba bt tab bs tabb bfbb
Rigid tension link
h1 k TR
slip
bt wab bswabbbwbb
tabb,wabb,sabb d bt wab bswabbbwbb slip M N
h2
Rigid compressionC link
h3 bs sabb bfbb slip kR
bfbb,bwbb
h0 o bt sab bs tabb bfbb slip
bs

slip

Fig. 2. Spring model of the TSDWA connection.

15
B. Meng et al. Engineering Structures 182 (2019) 13–28

Table 1
Mechanical behaviour of all components.
Component/rigid link Scheme of component P-v relationship References

T-stub tension (tab, sab, wab) [24,25]

Plate bearing (tabb, wabb, sabb, bfbb, bwbb) [26]

Bolt shear (bs) [27]

Bolt slip (slip) [28]

Bolt tension (bt) –

Tensile rigid link [29]

Compressive rigid link [29]

2.3. Equivalent component model of the TSDWA connection Therefore, according to the elastic stiffness of the equivalent springs in
the tension zone, d can be approximately calculated as follows:
To make subsequent calculation and analysis more convenient, the
Pti h i k ti h i2
individual components of each row were combined in series to form an d= =
equivalent spring (such as kt1, kt2, kt3, kt0, and kc, shown in Fig. 3). It Pti k ti h i (1)
should be noted that the individual component with the smallest ca- where, kti is the elastic stiffness of the corresponding equivalent spring
pacity controls the capacity of the equivalent spring. herein, and hi is the distance from the centre of compression to each
The lever arm d represents the tensile rigid link position which equivalent spring. It is important to note here that at the initial stage,
means the distance from the centre of compression to the rigid tension the equivalent spring (kt0) is in the compressive state and is not acti-
link. The lever arm d has a strong influence on the initial stiffness of the vated. Therefore, the equivalent spring (kt0) is not included in the de-
component model [20]. When the deformation of the connection is termination of d.
larger, the compression zone is not activated, and the rigid Bar 2 is
combined with the support Bar 0, i.e. the position of d will have little 3. Explicit calculation method
influence on behaviour of the TSDWA component model. In this con-
figuration, rigid Bar 2 can be considered equivalent to a cantilever [29]. To study the resistance of an unequal span steel frame structure

16
B. Meng et al. Engineering Structures 182 (2019) 13–28

r0 r1
Ba Bar 2 k t1 Ba

Rigid tension link


k TR k t2
h1
Equivalent k t3
h2 d
h3
kc k RC
h0 o k t0

Rigid compression link

Fig. 3. Equivalent component model of the TSDWA connection.

against progressive collapse, a two-span substructure was extracted In actual application, the equivalent tension or compression springs
from the failure area in Fig. 1 and was made the object of this study. may remain loaded, be unloaded, or be reloaded after unloading, as
The deformation of the substructure with a TSDWA connection in an shown in Fig. 6. It is assumed in this calculation that unloading is
interior column failure scenario is shown in Fig. 4. In this section, using conducted along the slope of k1 and the plastic deformation after un-
the equivalent component model of the TSDWA connection developed loading is vp. In this case, the deformation of the equivalent spring at
in Section 2, an explicit calculation method for determining the pro- any time can be calculated as follows:
gressive collapse resistance capacity of an unequal span steel frame is P Pe Pe
derived. v= + + vp
k k1 (2)
Please note that when P < Pe, k equals k1, if the unloading point is
3.1. Simplified model for theoretical analysis
(vm, Pm), the residual deformation (vp) and the corresponding force (Pm)
of the unloading point can be calculated using Eqs. (3) and (4), re-
The simplified model for the theoretical analysis of the TSDWA
spectively. In the nomenclature of this paper, the first subscript of a
component model is shown in Fig. 5. In the equivalent component
variable indicates whether it is in tension (t) or compression (c), and the
model, springs kt1, kt2, kt3, and kt0 in the tension zone are only activated
second subscript corresponds to the equation in which the variable is
when subjected to a tensile force. Similarly, spring kc in the compres-
used.
sion zone is only activated when subjected to a compressive force.
Furthermore, the elastic–plastic curves of these equivalent springs are vp = vm
Pm
asymmetric, as shown in Fig. 6, and the force–displacement responses k1 (3)
of these equivalent springs are nonlinear curves, i.e. their stiffness is not
Pe
constant. To simplify the theoretical calculations, the force-displace- Pm = Pe + k2 (vm )
k1 (4)
ment responses of these equivalent springs are approximated as a bi-
linear curve (shown in Fig. 6), in which the slope of the first line is Fig. 5(c) depicts the force diagram of the rigid bars in the TSDWA
defined as k1 and the slope of the second line is defined as k2. The two component model. Defining the vertical distance between the point of
slopes can be obtained by fitting the actual nonlinear curve of the action of the load (the bending moment (M) and/or axial force (N)) and
equivalent spring, by which the boundary point coordinates of the bi- the compression centre of the component model as z, the external forces
linear curves can also be obtained. on the component model will be balanced by the tensile force (PT) and

Inflection point P Inflection point


qL qR

Inflection point Inflection point


Failure column
bc l1 bc l2 bc

Fig. 4. Substructure with TSDWA connections in an interior column failure scenario.

17
B. Meng et al. Engineering Structures 182 (2019) 13–28

r0 r1 ar
1 r0 2 r1
Ba Bar 2 k t1 Ba r2
Ba Pt1 B Ba Bar Ba

Pt P
T
k TR Pt2
h1 k t2 θ
Pt3 v
k t3 M N
h2
h3 d θ1 h

z1
z
kc k RC PC Pc RC
h0 o k t0 Pt0

(a) Connection joint (b) Component model (c) Mechanical model (d) Deformation mode
Fig. 5. Simplified model for the theoretical analysis of the TSDWA component model.

compressive force (RC) in the rigid links. obtained for the TSDWA component model.
Fig. 5(d) depicts a typical deformation diagram of the component Based on the force equilibrium of rigid Bar 1
model. Here, θ1 represents the relative rotation of rigid Bar 2 relative to
N = Pt0 + Pti RC (5)
support Bar 0. Similarly, θ2 represents the relative rotation between
rigid Bar 1 and rigid Bar 2. The sum of θ1 and θ2 is defined as the total
M= Pti h i Nz Pt0 h 0 (6)
angle of rotation of the component model. The position of the rigid
tension link is at the centre of rotation of rigid Bar 2 and the position of Similarly, the force equilibrium of rigid Bar 2 yields
the rigid compression link is at the centre of rotation of rigid Bar 1. The
intersection of rigid Bar 1 and immovable support Bar 0 is defined as Pc = PT + RC Pti Pt0 (7)
the centre of rotation of the entire connection, and its distance from the
centre of compression of the component model is indicated by h. It P Td = Pti hi Pt0 h 0 (8)
should be noted that the position of the centre of rotation of the con-
nection varies within the bounds of the rigid tension link and the rigid Substituting the right side of Eq. (8) into the equilibrium Eq. (6), PT
compression link. The axial displacement (v) of the component model is is isolated as
given by the horizontal distance between rigid Bar 1 and the immovable M + Nz
PT =
support Bar 0, which is at z1 from the centre of compression, where z1 d (9)
represents the position of the geometric centreline of the connection. T
Combining Eqs. (5) and (7), and substituting P into Eq. (9), Pc is
When the position of the load application line coincides with the geo-
isolated as
metric centreline of the connection, z1 is equal to z. And if h is greater
than z1, the axial displacement (v) of the component model will be PC =
M + N(z d)
negative. d (10)

3.2. Force-balance relationship of the TSDWA component model 3.3. Deformation mode of the TSDWA component model

According to Fig. 5(c), the following force equations can be Many experimental studies [13,16,17] have shown that the ratio of

P P
Actual nonlinear curve

Simplified bi-linear curve


Intersection point -vcf -vcm -v cp -v ce 0
Ptm
k t2 v
k c1
e

Pte
u rv

k c1
rve

dc
cu

loa
d

un

k t1 -Pce
loa

k c2
k t1
un

vte v tp v tm vtf v Intersection point -Pcm


0
Actual nonlinear curve

Simplified bi-linear curve

(a)Equivalent tension spring (b)Equivalent compression spring


Fig. 6. Force-deformation response of the equivalent spring for theoretical analysis.

18
B. Meng et al. Engineering Structures 182 (2019) 13–28

the moment at the beam end to the axial force in the beam is not 3.4. Relationship of internal force to rotational angle in the TSDWA
constant under an interior column failure scenario, and that the ratio component model
changes with a conversion in the resistance mechanism. Therefore,
these changes in M/N require the TSDWA component model to present Based on the different deformation states of the component model
five different deformation modes successively, as shown in Fig. 7. discussed in Section 3.3 and the force–deformation responses of each
Deformation Mode I of the TSDWA component model represents the equivalent spring, explicit expressions for the M-N-θ and M-N-v re-
situation in which the internal axial tension in the connection is zero. In lationships of the component model can be developed as discussed in
this mode, the tension zone does no work, only the compressive zone is the following sections.
engaged. At this time, rigid Bar 1 is combined with rigid Bar 2, and the
M/N relationship meets the critical state of Eq. (11). 3.4.1. Angle of rotation in the tension zone
When the component model is in tension, the relationship between
M the angle of rotation of rigid Bar 1 (θ1) and the deformation of the
= 1
Nz (11) equivalent spring of the ith row (vi) can be obtained according to Mode
V as follows:
Deformation Mode II of the TSDWA component model indicates that
vti - e
the deformation state of the component model is in transition from 1 =
Mode I to Mode III. In this mode, the compression and tension zones of hi (14)
the connection are simultaneously engaged, and the component model where i is the identifying number of the equivalent spring.
undergoes a transition from compression to tension. The force–deformation curve of the equivalent spring of the ith row
Deformation Mode III of the TSDWA component model indicates and the bottom row can be derived using Eq. (2), as shown in Eqs. (15)
that all components in the tension zone (except the bottom component, and (16), respectively.
kb0) are engaged, the translation (e) of rigid Bar 1 is zero, and the
Pti Pte,i Pte,i
compressive force of the components in the compression zone is zero vti = + + vtp,i
(Pc = 0). Based on Eq. (10), the value of M/N can be obtained in this
k ti kt1,i (15)
mode using Eq. (12). Pt0 Pte,0 Pte,0
vt0 = + + vtp,0
M d z
kt0 k t1,0 (16)
=
Nz z (12) According to the geometry of the component model in Fig. 5, the
deformation of the equivalent spring of the bottom row is described by
According to Eq. (11), Eq. (12), and the deformation characteristics
vt0 = e - 1 h0 0. (17)
of the connection, the axial force in the component model under
compression and tension should meet the requirements of Eq. (13). Substituting vt0 from Eq. (17) into the equilibrium in Eq. (16) and
isolating Pt0 leads to
M M
<N< Pte,0
z d z (13) Pt0 = k t0 e - k t0 h 0 1 - (vtp,0 + ) kt0 + Pte,0 0
kt1,0 (18)
Deformation Modes IV and V of the TSDWA component model re-
Substituting vti from Eq. (15) into the equilibrium in Eq. (14) and
present situations in which the component mode is only affected by the multiplying kbi h i2 on both sides of the equilibrium in Eq. (14) produces
external axial tensile force, and the compressive force on the compo- the following expression:
nents is zero. When the bolt prying force is not considered, only the
tension zone of the connection is engaged. At this time, rigid Bar 2 is 2 k ti h i
1 k ti h i = (Pti - Pte,i) h i + Pte,i + vtp,i k ti h i ek ti h i
combined with support Bar 0, and the tensile deformation of the k t1,i (19)
component model can be expressed by the rotation angle (θ1) and According to Eq. (19), the sum of the rows of the equivalent tension
translation (e) of rigid Bar 1, corresponding to the catenary stage under springs can be expressed as
conditions of progressive collapse. Prior to Mode IV, the equivalent
spring kt0 has not been activated. Mode IV is notably a deformation- k ti h i
1 k ti h i2 = Pti h i Pte,i h i + Pte,i + vtp,i kti h i
critical state in which the displacement (vt0) of the equivalent spring kt0 k t1,i
is zero, after which the deformation mode of the TSDWA component ek ti h i (20)
model transitions to Mode V, when the equivalent spring kt0 is engaged,
potentially affecting the translation (e) of rigid Bar 1. Substituting Pt0 from Eq. (18) into the equilibrium in Eq. (6) and
isolating Pti h i leads to

r0 2 Bar
0 1 r0 r0 r0
Ba r r2 ar B a r1 Ba a r 1 Ba r1
Ba a
B v t1 B v t1 Ba v t1 B v t1 Ba

θ2
θ2 v t2 v t2 v t2 v t2
h1
h2 M N v t3 v t3 v t3 v t3
h3 d θ1 θ1 θ1
θ1
z

vc vc e e
h0 o v t0 =0 v t0

(Mode (Mode (Mode (Mode (Mode

Fig. 7. Deformation modes of the TSDWA component model.

19
B. Meng et al. Engineering Structures 182 (2019) 13–28

Pte,0 Pc Pce Pce


Pti h i = M + Nz + kt0 h 0 e kt0 h02 1 (vtp,0 + ) k t0 h 0 + Pte,0 h 0 vc = + + vcp
k t1,0 kc k c1 (28)
(21) Adding Eq. (5) to Eq. (7) and isolating Pc, then substituting PT from
Substituting Pti h i from Eq. (21) into the equilibrium in Eq. (20) Eq. (8) into the equilibrium in Eq. (7) results in the following expression
and isolating θ1 leads to for Pc:
M + Nz A + A3 A Pti hi Pt0 h 0
1 = + 2 + 4e 0 Pc = N
A1 A1 A1 (22) d (29)
where,
where, Pt0 is zero because the equivalent compression spring kto is not
vtp,i k ti h i ;
k hi
A1 = k ti h i2 + kt0 h02 ; A2 = Pte,i h i + Pte,i kti +
t1,i engaged.
P
A3 = Pte,0 h 0 - ( kte,0 + vtp,0 )k t0 h 0 ; A 4 = k t0 h 0 - kti h i It should be noted Then, substituting Pti h i Pt0 h 0 from Eq. (6) into the equilibrium
t1,0
that the equivalent spring kt0 is engaged when its displacement vt0 is in Eq. (29), and after substituting Pc from Eq. (6) into the equilibrium in
greater than zero, otherwise kt0 is zero. Eq. (27), the M-N-θ2 relationship can be obtained:
When the connection is in tension, the internal compressive force of
2 = D1 M + D1 (z - d)N +D2 0 (30)
the component model is zero and the tensile force on the component
model is balanced by all equivalent tension springs as follows: where, D1 = D2 =
1
; + d.
Pce 1
(
1
)
vcp
d2k c d k c1 kc

N = Pti + Pt0 As previously stated, the total rotation angle of the component
(23)
model consists of the relative rotation angles of rigid Bar 1 and rigid Bar
Dividing both sides of Eq. (19) by hi, the expression for the sum of 2, shown in the following equation.
the rows of equivalent tension springs can be developed by substituting
= + (31)
Pti + Pt0 from Eq. (23) and Pt0 from Eq. (18), then isolating e, pro- 1 2

ducing the following expression: Then, substituting θ1 from Eq. (26) and θ2 from Eq. (30) into the
1 equilibrium in Eq. (31) generates the following M-N-θ relationship for
e= (A 4 + N + A5 + A7 )
A6
1
(24) the component model under large deformation:

where, A5 = Pte,i +
k
Pte,i k ti + vtp,i k ti; A6 = k ti + k t0; A7 . = E1 M + E2 N + E3 (32)
t1,i

= - Pte,0 + ( k + vtp,0)k t0
Pte,0
where, E1 = C1 + D1, E2 = C2 + D1 (z d ), E3 = C3 + D2 .
The axial displacement of the connection (v) is also an important
t1,0
Substituting θ1 from Eq. (22) into the equilibrium Eq. (24) an iso-
lating e then yields: parameter that reflects the deformation value of the connection under
the interaction of bending moment and axial load. According to the
1
e= (A 4 1 + N + A5 + A7 ) deformation mode shown in Fig. 7, the axial displacement of the con-
A6 (25) nection can be approximately calculated by
A4 A1 + zA 4 A2 A4 + A3 A 4 + A1 A5 + A1 A7
where, B1 = ; B2 = ; B3 = .
A1 A6 - A 42 A1 A6 - A 42 A1 A6 - A 42 v= 1 z1 2 (d z1) + e (33)
It also should be noted that the translation (e) of rigid Bar 1 may be
negative when the axial load is less than the maximum value of Eq. Substituting θ1 from Eq. (26), θ2 from Eq. (30), and e from Eq. (25)
(13), indicating that the equivalent compression spring has been en- into the equilibrium in Eq. (33) leads to the following M-N-v relation-
gaged. In that case, B1, B2, and B3 are taken as zero, returning trans- ship:
lation (e) to zero.
v = F1 M + F2 N + F3 (34)
Substituting e from Eq. (25) into the equilibrium Eq. (22) and iso-
lating θ1, the M-N-θ1 relationship can be obtained as where, F1 = B1 + z1 C1 (d z1) D1; F2 = B2 + z1 C2 + (d z1)(d .
1 = C1 M + C2 N + C3 0 (26) z ) D1; F3 = B3 + z1 C3 - (d z1) D2
where, C1 =
1 + A4 B1
; C2 =
z + A 4 B2
; C3 =
A2 + A3 + A4 B3
. 3.5. Force-displacement relationship of an unequal span frame
A1 A1 A1

3.4.2. Angle of rotation in the compression zone To obtain the load capacity of an unequal span substructure under
The relationship between the angle of rotation of rigid Bar 2 (θ2) progressive collapse, the substructure is further simplified as a calcu-
and the deformation of the equivalent compression spring (vc) can be lation model with axial and rotational constraint springs, as shown in
obtained according to Mode I as follows: Fig. 8. In most practical cases, the loading types under a progressive
collapse scenario consist of uniform or concentrated loads, or a com-
vc
= bination of the two. These loads may be unequal on the beams adjacent
2
d (27)
to the failed column. Therefore, the above detailed situations of loading
The force–deformation curve of the equivalent compression spring on the substructure are considered and discussed in the following sec-
can be derived using Eq. (2) as follows: tions.

qL PL PR qR
kas kas
k rs k rs
bc l bc l2 bc
Fig. 8. Calculation model of an unequal span substructure.

20
B. Meng et al. Engineering Structures 182 (2019) 13–28

3.5.1. Axial deformation of the unequal span substructure Then, substituting ML from Eq. (41) into the equilibrium in Eq. (39)
When the deformation of the substructure is large, the catenary provides the axial force (NL) in the left beam:
action in the beams is more notable, with the approximate geometric
1 G2 G
deformation relationship shown in Fig. 9 (taking the left beam as an NL = 2 + + 3
G1 G1 G1 (42)
example, with the same applying to the right beam). Then, the axial
displacement of the left beam can be approximately calculated using its where,
geometrical deformation. By consulting Fig. 9, the following expression
2l1 2 24F1 E2 iL l1 24F1 iL 24F1 E3 iL l1
G1 = (4F2 + E b Ab
+ )l
kas 1 (1 + 6iL E1)
; G2 = (1 + 6iL E1)
; G3 = 1 + 6iL E1
.
can be derived:
4F3 l1
2 2
DL = , DR =
2l1 2l2 (35)
3.5.2. Collapse resistance of the substructure
where, DL and DR are the horizontal displacements of the left and right The axial force in the beam gradually develops with the increase in
beams, respectively; δ is the vertical displacement of the failed column the deformation of the substructure. Under these conditions, the equi-
of the substructure; and l1 and l2 are the net-span lengths of the left and librium of the substructure is shown in Fig. 10 (taking the left beam as
right beams, respectively. an example). In the explicit calculation model, NL is assumed to develop
The axial deformation of the unequal span substructure is mainly horizontally [5,11,30,31]. The equilibrium equation of the left beam of
composed of the connection, the beam itself, and the axial constraint of the substructure can be obtained as
the beam end. Therefore, the horizontal displacement of the sub- 1 2
structure can be expressed as follows: PL l1 + q l1 = 2ML + NL
2 L (43)
DL = 2vL + v aL + vLb, DR = 2vR + v aR + v Rb (36) Substituting ML from Eq. (41) and NL from Eq. (42) into the equi-
librium in Eq. (43) develops the relationship between the load capacity
where vL and vR are the axial displacements of the left and right beam-
and the displacement of the failed column as follows:
column connections, respectively, and can be calculated using Eq. (34);
v La and vRa are the axial deformations of the left and right beams 1 3 2
PL + q l1 = + + +
themselves, respectively, and can be obtained using Eq. (37); and v Lb 2 L
L L L L
(44)
and vRb are the axial deformations of the axial constraint of the left and 1 G2 12iL E2 G3 12iL
where, L = ; = ; = +
right beam ends, respectively, which can be calculated using Eq. (38): G1l1 L G1l1 (1 + 6iL E1)G1l1 L G1l1 (1 + 6iL E1)l12
12iL E2G2 12E3iL 12iL E2 G3
NL l1 NR l2 L =(1 + 6iL E1)G1l1
; (1 + 6iL E1)l1 (1 + 6iL E1)G1l1
v La = , v aR = Similarly, the load capacity of the right beam of the substructure
Eb A bL Eb AbR (37)
can be derived:
where NL and NR are the axial forces of the left and right beams, re-
1
spectively; Eb is the elastic modulus of the beams; and AbL and AbR are PR + q l2 = 3 + 2 + +
2 R
R R R R
(45)
the cross-sectional areas of the left and right beams, respectively.
NL N where the calculation of αR, βR, γR, and λR is similar to that of αL, βL, γL
v Lb = , vRb = R and λL, respectively.
k as k as (38)
The total load capacity of the entire substructure is the sum of the
In Eq. (38), kas is the equivalent axial stiffness of the beam ends; kac capacities of the left and right beams, as follows:
represents the axial constraint stiffness provided by the side columns; Ec
1 1
is elastic modulus of the columns; Ic is the cross-sectional moment of P + q l1 + qR l2 = 3 + 2 + +
2 L 2 (46)
inertia of the columns; lc is length of the side columns; and kaz re-
presents the axial constraint stiffness provided by the surrounding where,
members of the substructure. P = PL + PR ; = L + R; = L + R; = L + R; = L+ R
Assuming that the rotation angles and bending moments are equal The above formulas derived in Sections 3.1–3.5 form an explicit
at each end of the same beam, and that the axial force is constant along calculation method to describe the nonlinear static responses of a two-
the length of the beam, taking the left beam as the object of the study beam and three-column substructure with unequal spans. First, the
and substituting v from Eq. (34), DL from Eq. (35), v La from Eq. (37), and bilinear constitutive model of each equivalent spring is obtained by
v Lb from Eq. (38) into the equilibrium Eq. (36) results in the following fitting its actual nonlinear curve, which can be used to calculate the
ML-NL-δ2 relationship: parameters in Eqs. (22), (24)–(26), (30), (32), (34), (42), and (45). By
gradually increasing the displacement of the failure column (δ), the
2l1 2
2 = 4F1 l1 ML + (4F2 + + )NL l1 + 4F3l1 load–displacement curve of the substructure can be quantitatively
Eb Ab k as (39)
evaluated using Eq. (46) using the Excel spreadsheet software in a step-
If the rotation angle of the connection (θ) and the vertical dis- by-step manner. Similarly, a step-by-step procedure based on Eqs. (42)
placement of the failed column (δ) are obtained, according to the de- and (41) is used to track the axial force in the beam and the bending
flection method from structural mechanics, the bending moment (ML) moment of the beam end of the substructure, respectively. Besides, the
at the left beam end can be expressed as rotation and deformation of the connection can be assessed using Eqs.
iL iL
ML = - 4iL L - 2iL L +6 = - 6iL L +6 l
l1 l1 (40)
where iL is the linear stiffness of the left beam, iL = EbIb/l1; similarly, iR
is the linear stiffness of the right beam, iR = EbIb/l2; and Ib is the cross- D =l (1-cos ) δ
sectional moment of inertia of the beams.
cos ~ 1- 2/2
Substituting θ from Eq. (32) into the equilibrium in Eq. (40) pro-
vides the bending moment at the left beam end: ~ δ /l
D
6 i
ML = [L - iL E2 NL E3 iL]
1 + 6iL E1 l1 (41) Fig. 9. Approximated deformation of the left beam of the unequal span sub-
structure.

21
B. Meng et al. Engineering Structures 182 (2019) 13–28

D qL progressive collapse of a building caused by the failure of a local


N M member is likely to be a three dimensional problems. Because of the
inherent simplifications, the performance of the beam-column sub-
PL structure may not completely represent the whole structure following a
M δ column removal such as floor slab action, spatial action and the failure
N propagation of the whole structure. Although these factors should be
appropriately taken into account, for the purposes of deeply under-
standing and revealing resistance mechanisms including flexural me-
l chanism, compressive arching mechanism and catenary mechanism,
and thus making quantitative predict of the resistance of the sub-
Fig. 10. Equilibrium of the left beam of the substructure. structure with TSDWA connection, the proper use of the simpler ana-
lysis tool similar to the proposed method herein provides good com-
(31) and (34) respectively, which is used to make a comparison with putational efficiency and lays a foundation for the resistance prediction
corresponding deformation limits in each step and to determine whe- of the whole structure against progressive collapse.
ther the ultimate bearing capacity of the substructure is reached.
Therefore, the entire process of calculating the nonlinear static load of 4. Numerical analysis method
the substructure is established.
The proposed method is used for in–plane analyses of the unequal Generally, the more refined a numerical model is, the more accurate
span substructure after the failure of an internal column. In fact, the calculated progressive collapse resistance of the structure will be.
However, an overly refined model requires huge computing resources

Z Y

X O Faiure column
P
Ba Ba Ba Ba Ba Ba Ba Ba
r2 r1 r1 r2 r2 r1 r1 r2

A B C D E I J K L M
Left beam qL qR Right beam

FG H
Side connection Mid-span connection Side connection

bc /2 l1 bc/2 bc /2 l2 bc /2

(a)Schematic diagram of unequal span substructure

U1=U2=U3=0 U1=U2=U3=0

U3 U2
U1 O U1=U3=0

Left beam Right beam


Side column
Failure column
Side column

U1=U2=U3=0 U1=U2=U3=0

(b)Numerical model of unequal span substructure


Fig. 11. Numerical model of the unequal span substructure based on TSDWA component model.

22
B. Meng et al. Engineering Structures 182 (2019) 13–28

Upper

West East

Lower

Tension/
compression
load cell Upper side
column

bc
Horizontal hinged

B6
B6
connection
Horizontal hinged
connection
Upper beam

l2
Hydraulic actuator

B4
B4
Reaction
wall

bc
Lower beam Failure column

B1B3
B1B3
Horizontal hinged
connection Horizontal hinged

l1
connection

bc
Tension/
compression
load cell Lower side
column

Fig. 12. Test set-up and loading (from [19]).

Table 2
Dimensions of test specimens (from [19]).
Specimens Length of lower beam (l1 + bc) (mm) Length of upper beam (l2 + bc) (mm) Span ratios (l1 + bc)/(l2 + bc) Length of side column lc (mm)

TSDWA-0.6 1500 900 1: 0.6 1100


TSDWA-1.0 1500 1500 1: 1.0 1100
TSDWA-1.4 1500 2100 1: 1.4 1100

2524 52 2425

Ba Ba r0 1
r1 r2 r2 r1 Ba Bar
2 r
Ba Ba Ba
40 50 20 30 50 30 20 50 40

k t0 k t0 C k t1
t=10 kR
C
kR h0

k t3 kc k c k t3 h3
d h2 kTR k t2 h1

k t2 d k t3
k t2 h1 h2
kc h3
kR
C
t=10
k t1 k T
k T
k t1
50

R R
Faiure column
k t0 h0

30 45 45 30 40 60 40 10 3030 Internal connection Side connection


150

Fig. 13. Detailed layout of the TSDWA component model.

23
B. Meng et al. Engineering Structures 182 (2019) 13–28

Table 3 Fig. 11(b). The vertical concentrated load at the top of the failure
Geometric parameter values in component model of TSDWA. column and the uniform load on the beams were applied as shown in
Parameter d h1 h2 h3 h0 Fig. 11(a). The ABAQUS/Standard nonlinear static solver was used to
analyse the component model of the frame. To make the model more
Distance (mm) 150 205 105 55 45 convergent, the appropriate “Automatic Stabilization” parameter was
set in the analysis step, which ensured that the ratio of the viscous
damping energy to the total strain energy of the structural system was
and a great deal of time to analyse, particularly for a complex model sufficiently small [32].
with semi-rigid connections. Therefore, in this section, an efficient
method for numerical simulation based on the component model of the
5. Application and verification of explicit calculation and
TSDWA connection discussed in Section 2 is described in detail. This
numerical analysis methods
method can rapidly predict and assess the progressive collapse re-
sistance of the substructure under a column failure scenario. To ac-
The explicit calculation method derived in Section 3 and the nu-
complish this, the nonlinear analysis software ABAQUS was used to
merical analysis method based on the component model described in
establish the unequal span substructure model for some analyses.
Section 4 were applied and validated using relevant experimental data
To account for the unequal spans of the substructure, the entire
from the test results obtained by Zhong et al [19] as described in this
substructure was simulated, as shown in Fig. 11. The component model
section. The methods were evaluated in terms of their ability to du-
of the TSDWA connection was composed of a support Bar 0 (included in
plicate experimental test results for external and internal force-dis-
the column), rigid Bar 1, rigid Bar 2, rigid tension/compression links,
placement relationships.
and multiple unidirectional springs. The column and beam were si-
mulated using beam elements (B31), as were all rigid bars in the
5.1. Model verification tests
component model, but the higher stiffness of the rigid bars was ac-
complished by increasing the elastic modulus of their material (in this
In [19], three, 1/3-scale unequal span beam-column substructures
study, the elastic modulus of the rigid bars was assumed to be 10,000
with TSDWA connections were tested in the setup shown in Fig. 12.
times that of the beams). The size of beam element was approximately
Specimens H-150 × 150 × 8 × 10 (mm) and H-150 × 100 × 6 × 9
20 mm. In the substructure model, the beam ends were rigidly con-
(mm) [H-overall depth (d) × flange width (bf) × web thickness
nected by rigid Bar 1. The rigid tension/compression links were simu-
(tw) × flange thickness (tf)] were used for column section and beam
lated using the axial connector element (CONN3D2), as were the
section in the substructure, respectively. The top/seat angle and web
equivalent springs, indicated by the dashed line in Fig. 11(b). The po-
angle sections were L-140 × 90 × 10 (mm) and L-70 × 70 × 8 (mm),
sitions of the axial connector elements correspond to their geometric
respectively. Grade 10.9 M16 bolts were used to make all connections,
positions. The mechanical properties of each equivalent spring were
and their nominal yield strength and ultimate tensile strength were
accounted for by defining each axial connector element with its cor-
900 MPa and 1000 MPa, respectively. The dimensions of the test spe-
responding force–deformation curve.
cimens were as listed in Table 2, and the related material properties
In this three-bar component model, successful transfer of shear ac-
were as given in [19]. The locations of the sections (B1-B1, B3-B3, B4-
tion of the connection is the key to model analysis and convergence. It
B4, B6-B6) at the beam ends, shown in Fig. 12, were the same as in
is worth noting that shear coupling points of the side joint and internal
[19], and were mainly used to describe the development of the bending
joint are different (as shown in Fig. 11(a)), which is mainly related to
moments at the beam ends during the loading process.
the vertical displacement constraint of the column adjacent to the joint.
The beam-column connections of the substructure contained an
In this numerical model of the substructure, the vertical displacements
internal connection and a side connection in the detailed layout shown
of the side columns are constrained, while the vertical displacements of
in Fig. 13. The relative geometric parameters of the TSDWA component
the internal column are not constrained. Through analysis and trials
model are provided in Table 3. In the TSDWA component model, the
performed by the authors, only when the vertical displacements of
constitutive relationship of equivalent springs kt1 and kt0 are the same,
point pairs A and B, B and C, D and E, F and G, G and H, I and J, K and L,
as well as the constitutive relationship of kt2 and kt3, as shown in
and L and M are coupled, can the shear force of the connections be well
Fig. 14(a) and (b), respectively. Fig. 14(c) provides the for-
transferred. The boundary constraint conditions of the substructure
ce–displacement relationship of the equivalent compression spring kc. It
model were defined by the restraint in the U3 direction (Z direction) of
should be noted that the fracture of the tension angles (top/seat angle)
rigid Bar 1 and Bar 2, the U3 direction of the beams and the columns,
was not instantaneous after yielding was observed in the tests: the de-
the U1, U2, and U3 directions of the side column ends, and the U1 and
formation of the tension angles continued to develop, representing a
U3 directions of the upper end of the failure column, as shown in
slight departure from the Yang’s model of the bolt-angle connection

120 o o 100 -5 -4 -3 -2 -1 0
100 80 0
Failure point
80 60 Deformation/mm -20
Ultimate bearing -40
60
capacity 40
40 -60
20
F orce/kN
F o rce/k N

20 -80
0
0 0 5 10 15 20 25 30 35 40 -100
0 10 20 30 40 50 60 70 80
-20 Deformation/mm
-20 Deformation/mm -120
vu vf -40
-40 -140
-60 -60
F o rce/k N

-160
-80 -80 -180
-100 -100 -200

(a) Equivalent springs, kb1and kb0 (b) Equivalent springs, kb2 and kb3 (c) Equivalent spring, kc
Fig. 14. Force–deformation curves of equivalent springs.

24
B. Meng et al. Engineering Structures 182 (2019) 13–28

180 180 160


EC result EC result EC result
160
NA result
160 NA result 140 NA result
140 T result 140 T result 120 T result
120 120
100
Load/kN

Load/kN

Load/kN
100 100
80
80 80
60
60 60
40 40 40

20 20 20
0 0 0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 400 450 0 50 100 150 200 250 300 350 400 450 500
Displacement/mm Displacement/mm Displacement/mm
(a) TSDWA-0.6 (b)TSDWA-1.0 (c)TSDWA-1.4
Fig. 15. Comparison of the force-displacement curves of the subject unequal span steel frame.

Table 4 consider the failure of a critical individual component that first fails in
Comparison of maximum peak points. the component model and results in the continued damage of a frame
Specimens Maximum load/kN Displacement corresponding to the structure. In general, the proposed analysis methods can accurately
maximum load/mm capture the key performance of an unequal span steel frame with
TSDWA connections before the system loses its main load capacity.
EC NA T EC/T NA/T EC NA T EC/T NA/T

TSDWA-0.6 161 152 151 1.07 1.01 237 244 225 1.05 1.08 5.3. Internal force-displacement curve of the substructure
TSDWA-1.0 171 175 172 0.99 1.02 361 366 358 1.01 1.02
TSDWA-1.4 149 146 133 1.12 1.10 363 370 369 0.98 1.00
The axial forces and bending moment generated in the beams (at
sections B1, B3, B4, and B6) were calculated using Eqs. (42) and (41),
under tensile force [24]. Therefore, in the component model of this respectively, and a step-by-step analysis was carefully conducted and
connection, it was assumed that the top/seat angle in bending experi- examined using an Excel spreadsheet in which the displacement of the
ences a stable plastic phase after reaching the peak point predicted by failed column was gradually increased by 1-mm increments, starting
the Yang model, as shown in Fig. 14(a). Under this assumption, the from zero. Fig. 16 depicts the internal force–displacement responses
deformation continues to increase but the bearing capacity no longer determined by the explicit calculations and the numerical analyses
increases in the last stage. The deformation value of the plastic flow compared with those from the corresponding physical test results.
phase was obtained in this study by testing. Clearly, the results of the two methods are essentially identical to the
physical experiment results, indicating that the two methods of analysis
are well-equipped to reflect the development of internal forces (axial
5.2. Force–displacement curve of the substructure
force and bending moment) in an unequal span steel frame under
column failure conditions.
Based on the proposed TSDWA component model, the load-dis-
Fig. 17 depicts the deformation cloud diagram corresponding to
placement responses of the three unequal span substructures under a
maximum load capacity of the three evaluated unequal span steel
column failure scenario were obtained by both explicit calculation and
frames. This figure also demonstrates that the TSDWA component
numerical analysis. Only a concentrated load was considered and ap-
model is able to accurately reflect the performance of this connection
plied at the top of the failed column to match the lack of uniform loads
under large deformations, and that the numerical analysis methods can
(q) on the beams in physical experiments. When the explicit analysis
successfully simulate the large deformation behaviour of the entire
was conducted, the analytical load–displacement (P-δ) curves were
unequal span substructure.
determined using Eq. (46) and their coefficients were calculated using a
step-by-step calculation conducted in the Excel spreadsheet software.
This analysis process consisted of increasing the vertical displacement 6. Conclusions
of the failed column in 1-mm increments, starting at the no-load case,
and determining the corresponding vertical concentrated load. The Based on the basic framework of Izzuddin et al. [10] for assessing
numerical analysis was conducted by providing the constitutive re- the progressive collapse of steel frames with substructure levels, a more
lationship of each equivalent spring of the component model according explicit approach was derived and the anti-collapse behaviour of an
to Fig. 14. unequal span substructure with TSDWA connections was evaluated in
The load–displacement curves from the explicit calculations and the this study.
numerical analysis were then compared with the physical test results in Based on the component method, a three-bar spring component
Fig. 15. Table 4 provides a comparison of the maximum points on the model consisting of a series of individual components was constructed
curves, in which EC indicates the results of the explicit calculations, NA by dividing and simplifying a TSDWA connection according to its
indicates the results of the numerical analysis, and T indicates the re- geometry. The constitutive relationships of the equivalent springs were
sults of the physical tests. It can be seen that the results of the two determined using previous studies. The resulting equivalent component
analysis methods are in good agreement with the physical test results. model was not only able to provide good computational efficiency but
Clearly the two methods are able to reflect the main progressive col- was also able to predict the nonlinear mechanical behaviour of this type
lapse responses of an unequal span steel frame with TSDWA connec- of connection under an interior column failure scenario, including
tions, including the arching action and flexural action in the initial arching action and flexural action in the initial phase, and catenary
phase, and the catenary action in the larger deformation phase. It action in the larger deformation phase.
should be mentioned that the proposed analytical methods do not To describe the nonlinear static responses of the unequal span
completely reflect the influence of several TSDWA connection failures substructure with TSDWA connections, an explicit calculation method
following the maximum points. However, the proposed model is able to was proposed for an interior column failure scenario. This explicit

25
B. Meng et al. Engineering Structures 182 (2019) 13–28

200 50 EC-B1 EC-B3 EC-B4 EC-B6


EC-upper beam
FE-B1 FE-B3 FE-B4 FE-B6
EC-lower beam 40 T-B1 T-B3 T-B4 T-B6
150 NA-upper beam

Bengding moment/kN⋅m
NA-lower beam 30
100 T-upper beam EC-B6
20 EC-B4
T-lower beam
Axial force/kN

FE-B6
50 10 FE-B4
T-B4 T-B6
0
0 T-B1 T-B3
-10 EC-B1 EC-B4
FE-B1
FE-B3
-50 -20

-30
-100
0 50 100 150 200 250 300 0 50 100 150 200 250 300 350
Displacement/mm Displacement/mm

(a)TSDWA-0.6
350 EC-upper beam 40 TC-B1 TC-B3 TC-B4 TC-B6
EC-lower beam FE-B1 FE-B3 FE-B4 FE-B6
300
NA-upper beam 30 T-B1 T-B3 T-B4 T-B6
250 NA-lower beam

Bengding moment/kN⋅m
T-upper beam 20
200 T-lower beam FE-B4
FE-B6
Axial force/kN

TC-B4
150 10 TC-B6
T-B4
T-B6
100 0
T-B1T-B3
50 FE-B1
-10
FE-B3
0 TC-B1 TC-B3
-20
-50

-100 0 -30
50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450
Displacement/mm Displacement/mm

(b)TSDWA-1.0

EC-upper beam 40
250 TC-B1 TC-B3 TC-B4 TC-B6
EC-lower beam FE-B1 FE-B3 FE-B4 FE-B6
NA-upper beam 30 T-B1 T-B3 T-B4 T-B6
200
NA-lower beam
Bengding moment/kN⋅m

T-upper beam 20
150 FE-B4 TC-B4 TC-B6
T-lower beam FE-B6
Axial force/kN

10 T-B4
100
T-B6
0
50
T-B1
-10 T-B3
0 FE-B1
FE-B3 TC-B1 TC-B3
-50 -20

-100 -30
0 50 100 150 200 250 300 350 400 450 0 50 100 150 200 250 300 350 400 450
Displacement/mm Displacement/mm

(c)TSDWA-1.4
Fig. 16. Comparison of the internal forces of the subject unequal span steel frame.

calculation method is based on the TSDWA component model including structure showed good agreement between all methods. Of course,
explicit expressions to calculate internal forces (bending moments and when experimental data is difficult to obtain, numerical simulations can
axial forces) and the resistance of the unequal span substructure. By be confidently used as a basis for verification of the explicit method.
using the proposed TSDWA component model, a simplified numerical The proposed methods have therefore been demonstrated to provide a
model was developed, which can also obtain the nonlinear static re- simple and effective way to obtain the resistance of an unequal span
sponses of the unequal span substructure. substructure in the nonlinear static state following column removal. It is
The accuracy of the two proposed methods was then verified against believed that the present study lays a foundation for the resistance
the results of previous experimental tests, indicating that the proposed prediction and performance analysis of the whole structure against
explicit equations and numerical analysis process can be confidently progressive collapse.
used to rapidly and quantitatively evaluate the resistance of a steel
frame structure against progressive collapse in the case of interior Acknowledgements
column failure. Behaviours such as the load–displacement relationship,
the development of internal forces, and the large deformation of the The research presented in this paper is supported by the National

26
B. Meng et al. Engineering Structures 182 (2019) 13–28

(a)TSDWA-0.6

(b)TSDWA-1.0

(c)TSDWA-1.4
Fig. 17. Deformation cloud diagram of the subject unequal span substructure.

Natural Science Foundation of China through Grant No. 51678476, the [4] Guo L, Gao S, Fu F. Structural performance of semi-rigid composite frame under
Key Research and Development Project of Shaanxi Province through column loss. Eng Struct 2015;95:112–26.
[5] Stylianidis PM, Nethercot DA, Izzuddin BA, et al. Study of the mechanics of pro-
Grant No. 2018ZDXM-SF-097, China Postdoctoral Science Foundation gressive collapse with simplified beam models. Eng Struct 2016;117:287–304.
Funded Project through Grant No. 2018M643593, and Talent Science [6] EN 1991-1-7. Eurocode 1: Actions on structures-Part 1-7: General actions accidental
and Technology of Xi’an University of Architecture and Technology actions. Brussels: European Committee for Standardization; 2006.
[7] ASCE 2013. Minimum design loads for buildings and other structures. ASCE/SEI 7-
through Grant No. RC1827. All opinions, findings, conclusions, and 10. USA: American Society of Civil Engineers; 2013.
recommendations expressed in this paper are those of the writers and [8] GSA. Alternate path analysis & design guidelines for progressive collapse resistance.
do not necessarily reflect the views of the sponsors. Washington D C (USA): United States General Services Administration; 2013. p.
2013.
[9] UFC 4-023-03. 2013. Design of structures to resist progressive collapse. Washington
Declaration of interest D C (USA): Department of Defense; 2013.
[10] Izzuddin BA, Vlassis AG, Elghazouli AY, et al. Progressive collapse of multistory
buildings due to sudden column loss-part I: simplified assessment framework. Eng
None.
Struct 2008;30:1308–18.
[11] Stylianidis PM, Nethercot DA, Izzuddin BA, et al. Modelling of beam response for
References progressive collapse analysis. Structures 2015;3:137–52.
[12] Yang B, Tan KH. Experimental tests of different types of bolted steel beam-column
joints under a central-column-removal scenario. Eng Struct 2013;54(9):112–30.
[1] Fu F. Progressive collapse analysis of high-rise buildings with 3-D finite element [13] Zhong WH, Meng B, Hao JP. Performance of different stiffness connections against
modelling method. J Constr Steel Res 2009;65:1269–78. progressive collapse. J Constr Steel Res 2017;135(8):162–75.
[2] Alashker Y, Li H, El-Tawil S. Approximations in progressive collapse modelling. J [14] Coelho AMG. Characterization of the ductility of bolted end plate beam-to-column
Struct Eng 2011;137(9):914–24. steel connections. Portugal: University of Coimbra; 2004.
[3] Li H, El-Tawil S. Three-dimensional effects and collapse resistance mechanisms in [15] Guo LH, Gao S, Fu F, et al. Experimental study and numerical analysis of
steel frame buildings. J Struct Eng 2014;140(8).

27
B. Meng et al. Engineering Structures 182 (2019) 13–28

progressive collapse resistance of composite frames. J Constr Steel Res [23] Lemonis ME, Gantes CJ. Mechanical modeling of the nonlinear response of beam-to-
2013;89(5):236–51. column joints. J Constr Steel Res 2009;65(4):879–90.
[16] Li L, Wang W, Chen YY, et al. Experimental investigation of beam-to-tubular [24] Yang B, Tan KH. Robustness of bolted-angle connections against progressive col-
column moment connections under column removal scenario. J Constr Steel Res lapse: mechanical modelling of bolted-angle connections under tension. Eng Struct
2013;88(9):224–55. 2013;55(10):153–68.
[17] Lew HS, Main JA, Robert SD, et al. Performance of steel moment connections under [25] Faella C, Piluso V, Rizzano G. Structural steel semirigid connections: theory, design
a column removal scenario. I: experiments. J Struct Eng 2013;139(1):98–107. and software. CRC Press; 1999.
[18] EN 1993-1-8. Eurocode 3: Design of steel structures Part 1.8: design of joints. [26] Rex CO, Easterling WS. Behavior and modeling of a bolt bearing on a single plate. J
Brussels: European Committee for standardization; 2005. Struct Eng 2003;129(6):792–800.
[19] Meng B, Zhong WH, Hao JP. Anti-progressive collapse behavior of beam-to-column [27] Yu HX, Burgess IW, Davison JB, et al. Experimental investigation of the behaviour
assemblies with bolted-angle connections under different span ratios. Adv Struct of fin plate connections in fire. J Constr Steel Res 2009;65(3):523–36.
Eng 2018;21(6):891–905. [28] Yu HX, Burgess IW, Davison JB, et al. Tying capacity of web cleat connections in
[20] Savio AAD, Nethercot DA, Vellasco PCGS, et al. Generalised component-based fire, part 2: development of component-based model. Eng Struct
model for beam-to-column connections including axial versus moment interaction. 2009;31(3):697–708.
J Constr Steel Res 2009;65(2):1856–95. [29] Stylianidis PM, Nethercot DA. Modeling of connection behaviour for progressive
[21] Yuan Z, Tan KH, Ting SK. Testing of composite steel top-and-seat-and-web angle collapse analysis. J Constr Steel Res 2015;113(6):169–84.
joints at ambient and elevated temperatures, Part 1: ambient tests. Eng Struct [30] Li G, Wang K, Liu Y, et al. Catenary action of restrained steel beam against pro-
2011;33(10):2727–43. gressive collapse of steel frameworks. J Central South Univ 2012;19(2):537–46.
[22] Yuan Z, Tan KH, Ting SK. Testing of composite steel top-and-seat-and-web angle [31] Gao S. Progressive collapse behavior of planar steel frame with composite beam.
joints at ambient and elevated temperatures, Part 2: elevated-temperature tests. Eng Harbin: Harbin Institute of Technology; 2014.
Struct 2011;33(4):2093–109. [32] ABAQUS analysis user's manual version 6.12. ABAQUS Inc: 2012.

28

You might also like