You are on page 1of 12

Engineering Structures 31 (2009) 697–708

Contents lists available at ScienceDirect

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

Tying capacity of web cleat connections in fire, Part 2: Development of


component-based model
Hongxia Yu a,∗ , I.W. Burgess a , J.B. Davison a , R.J. Plank b
a
Department of Civil and Structural Engineering, University of Sheffield, Sheffield S1 3JD, UK
b
School of Architecture, University of Sheffield, Sheffield S10 2TN, UK

article info a b s t r a c t

Article history: The companion paper has reported the results from a test programme in which web cleat connections
Received 13 June 2008 were subjected to various combinations of shear, tying and moment actions at elevated temperatures.
Received in revised form These tests showed that web cleat connections have very good tying resistance and rotational capacity,
28 August 2008
mainly due to the large deformation of which the web cleats are capable. In this paper a mechanical
Accepted 3 November 2008
Available online 27 December 2008
model is developed to predict the behaviour of web cleats subjected to tying forces. This model considers
the formation of four plastic hinges on each angle and the effect of the angles opening in enhancing their
Keywords:
resistance. It is capable of representing the action of the angles in component-based models for web cleat
Web cleat connection connections, in which algorithms for other components, including bolts in tension, bolts in double shear
Fire and holes in bearing, are already available. Failure criteria determined from the tests have been introduced
Tying capacity into the models for components such as web cleats and bolts in double shear. This enables the component-
Component-based method based assembly to predict the occurrence and the sequence of connection failure. The behaviour of the
connection predicted by the component-based model shows good correlation with the test results, which
indicates that the developed model can be adopted in structural frame analysis to consider connection
failure.
© 2008 Elsevier Ltd. All rights reserved.

1. Introduction to higher moments, but collapse was generally determined by


the capacity of the connection to resist the tying forces. All
The behaviour of a real structure in fire involves an intimate the sub-frames failed at their connections when the beams had
interaction between the behaviour of its connections and that developed large deflections, and the failure temperatures were
of the connected members, in that the internal forces generated highly dependent on the connection type.
cannot be ascribed to either in isolation. Inclusion of a model In the past, mainly flush end plate and flexible end-plate con-
that describes the realistic connection behaviour in a structural nections have been experimentally studied at elevated temper-
analysis is essential in both understanding the structural behaviour atures [4,5]. These studies covered only moderate deformations
and determining its fire resistance. Connection failures have and were limited to moment–rotation relationships. Taking into
been observed in fire on a number of occasions [1,2]. These account the fact that connection failure may result in progressive
are a consequence of tying forces transferred through the collapse of the whole structure, their real behaviour should be con-
connection, which may be the result of cooling (and shortening) sidered and incorporated in structural fire safety design analysis.
of members, or as a result of large deflections developed in the The companion paper [6] reports a test programme aimed at in-
steel or composite beam because of heating. However, traditional vestigating the behaviour of web cleat connections subjected to
understanding of connection behaviour has tended to be limited to combinations of shear, tying and moment. Finite element analy-
ses were then performed to simulate these test results. These ana-
moment–rotation characteristics. Tests [3] performed on the same
lytical results have helped in understanding the test observations,
structural assembly, but with different connection types, have
and gave a better indication of the resistances of individual con-
shown that the forces transferred through the connections tested,
nection components. However, 3D solid modelling of connections
including compression/tying and moment, changed considerably
is inconvenient for normal structural designers involved in analyt-
with temperature. Stiffer connections were generally subjected
ical design. A more practical solution lies in the development of
component-based models to simulate the connection behaviour.
The concept of a component-based model to predict the
∗ Corresponding author. Tel.: +44 114 222 5726; fax: +44 114 222 5793. ambient-temperature moment–rotation (M–φ ) behaviour of steel
E-mail address: H.Yu@sheffield.ac.uk (H. Yu). and composite joints was first proposed in the 1980s [7] and
0141-0296/$ – see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2008.11.006
698 H. Yu et al. / Engineering Structures 31 (2009) 697–708

connections in steel and composite joints. Spyrou [12] developed


Notations a new model for T -stub behaviour at high temperatures based
on beam theory, considering the compatibility of displacement
d Bolt diameter between the bolts and the T -stub, in which bolt fracture was
fy Yield strength of normal steel considered as the ultimate failure mode for all the T -stubs.
fu Ultimate strength of normal steel For the compression zone of endplate connections, the tra-
fub Ultimate strength of the bolt ditional approach had been mainly concerned with its plas-
g Distance between the bolt lines tic resistance and initial stiffness. Spyrou [13] proposed a
RH Radius of bolt head force–displacement relationship for the compression zone based
t Thickness of the web cleat on the sequential yielding of the column web and flange. Block [14]
w Effective width of the web cleat model proposed the use of an elliptical curve to connect the elastic phase
A The tensile stress area of the bolt and the ultimate resistance. Combining his model for the compres-
B Width of the bearing surface of bolts sion zone with the T -stub model by Spyrou, Block developed a
C1 , C2 Constants determined by boundary conditions. component-based model for endplate connections, which works
E Young’s modulus well when compared with a number of M–φ tests from various
Et Tangent modulus sources. Sarraj [15] performed a thorough investigation of the be-
F Load applied to a web cleat haviour of fin plate connections and proposed models for com-
Fb Bearing resistance of a bolt ponents such as bolt holes in bearing, bolts in shear, and friction.
Fb,Rd Nominal bearing resistance of a bolt His model has been shown to compare well against fin plate con-
FV Double shear resistance of a bolt nection tests at elevated temperatures, but his model for bolt be-
FV ,Rd Nominal double shear resistance of a bolt haviour can now be improved [16]. Most recently, Ramli-Sulong
H Internal force generated in a web cleat that is et al. [17] implemented a component-based model for several con-
perpendicular to the applied load F nection types into the computer program ADAPTIC for structural
I Second moment of area of a cross-section analysis.
Kr Rotational stiffness of plastic hinges The previous work on the development of component-based
L The distance between the two ends of Beam 1 in the models can be roughly divided into two categories. The first aims
web cleat model to provide an approximate estimation of the initial stiffness and
L0 Span length of an idealised Beam1 in the web cleat the moment resistance by a method simple enough for hand
model calculation, so that it can conveniently be included in conventional
L1 , L2 Original length of idealized Beam 1 and Beam 2 in structural design. The second aims to provide a nonlinear curve
the web cleat model which describes the behaviour of the connection up to its ultimate
M0 Moment at the free ends of Beam 1 and Beam 2 in resistance. This is generally implemented into structural analysis
the web cleat model programs as a ‘‘connection element’’. Research into the behaviour
My Yield moment capacity of the plastic hinge structures in fire [16,6] has shown two characteristics. One is that
Mn Nominal plastic moment capacity of a cross-section fracture of the connection components determines the ultimate
Mu Ultimate moment capacity of the plastic hinge resistance of simple connections, and must be considered in
Mr1 , Mr2 Moments at the fixed end of Beam 1 and Beam 2 in design. The other is that the behaviour of the connection cannot
the web cleat model be solely described by a moment–rotation relationship due to
N The axial force in Beam 1 in a web cleat model the complexity of the loads to which it is subjected in fire
Nn Nominal tensile capacity of a cross-section cases. This means that the behaviour of each of the components
εp Proportional strain of steel of the connection, during the whole loading or heating phase,
εy Yield strain of steel is important as it will affect the load distribution between
εu Strain at the ultimate stress components and the failure mode.
δ General symbol used for displacements Most previous connection tests have been stopped at a rotation
δ1 , δ2 Deflections of Beam1 and Beam 2 in the web cleat of less than 60 mrad, which is only able to provide the plastic
model resistance. It is an objective of this research project to study
δ1,end , δ2,end Deflections at the free ends of Beam 1 and Beam the behaviour of the connections up to fracture. Following tests
2 in the web cleat model in which specimens were loaded until rotations up to 18◦ , this
δh Horizontal movement of the free end of Beam 1 in project aimed to develop component-based models capable of
the web cleat model predicting the behaviour up to the failure of commonly-used
θ Inclination of Beam 1 in the web cleat model steel connections. These models are mainly to be used in the fire
θ1,end , θ2,end Rotations at the free ends of Beam 1 and Beam 2 situation and, therefore, the behaviour at high deformation, rather
in the web cleat model than initial elastic behaviour, is the major concern. Hence, failure
criteria are incorporated wherever available information allows. To
describe the nonlinear behaviour at high deformation, the methods
has since then been adopted in design guidance [8]. This method
may appear too complicated for hand calculation, but can be easily
divides the joint into an assembly of independent components.
incorporated into numerical analysis.
As each component can be described as a (possibly nonlinear)
spring, the overall behaviour of the joint can be calculated from In this paper, a new mechanical model is developed to describe
the equilibrium and compatibility of the assembly of springs. Most the behaviour of the angle-cleat in a web cleat connection. The be-
previous work has been concerned with endplate connections. haviour of the components common to all bolted connections, such
Leston-Jones [4] summarised previous work and applied it to as bolts in shear/tension and bolt holes in bearing, at both ambient
(M–φ ) curves at elevated temperatures. His model included an and elevated temperatures, has been studied previously [15,16].
elastic phase and a strain-hardening phase after yield. Da Silva, This has enabled an almost complete component-based model to
et al. [9] proposed a model based on the equations of EC3: Part be constructed for the web cleat connection. The results from the
1.8 [8], but with the thermal effect integrated. Al-Jabri et al. [10,11] component-based model are seen to agree well with the test re-
developed a component-based (M–φ ) model for flexible endplate sults reported in the companion paper.
H. Yu et al. / Engineering Structures 31 (2009) 697–708 699

Fig. 1. Mechanical model for the web cleat.

2. Development of a mechanical model for the web cleat cantilever beams with concentrated forces at their ends, as shown
in Fig. 1. Each beam can form a plastic hinge at each end. The
Previously, one simple model has been developed by Owens and original lengths of these two beams are L1 and L2 respectively.
Moore [18] to predict the tying capacity of web cleat connections. When the angle is pulled by the force F applied to the beam web,
The load capacity was calculated on the basis of a plastic hinge it is assumed that Beam 1 is subjected to large deflection, and
mechanism without consideration of deformation history. The its effective span shortens to L0 . However, Beam 2 is assumed to
capacities predicted by this method were around 60% of those have only a small deflection, and its span remains at L2 . The total
found under tests when the connections were subjected to pure deformation of the angle due to the force F is therefore equal to
tying force, normal to the column face. Considering that in reality the deflection of Beam 1. A tri-linear material model with infinite
connections are rarely subjected to pure tension, it is apparent that ductility, shown schematically in Fig. 1, is adopted. The moments
the bolt rows cannot all reach their maximum resistances at the at the fixed ends of Beams 1 and 2 are denoted as Mr1 and Mr2 . The
same time. A model which can relate the displacement of each bolt moments at the ends of the two beams at the heel are the same,
row to the equivalent force generated by that bolt row, considering and are denoted as M0 . The moments My and Mu are the yield and
the deformation behaviour of the relevant parts of the web cleats, ultimate moment capacities of the cross-section.
is necessary to predict the connection behaviour. Ramli-Sulong In the subsequent formulations it is assumed that plastic hinges
et al. [17] state that they have incorporated a component-based can form at the ends of the legs. Before the moments at these
model for web cleat connections into the program ADAPTIC, positions reach My the behaviour at these positions is controlled
although no detail is given of the model’s principles or formulation. by elasticity and continuity; after they reach Mu , a plastic hinge
Their model appears to be based on the traditional two parameters is created. Between My and Mu , the behaviour at these positions
of initial stiffness and plastic resistance. This would not be enough is assumed as being described by a linear variation of moment
to describe the behaviour of the angles, which have the potential and rotation, as shown in Fig. 2. Before the force–deformation
for considerable enhancement of their resistance as the angles behaviour of the two beams can be resolved, the tangent rotational
open at large deformation. In this section, a mechanical model is stiffness Kr is derived on the basis of a bilinear stress–strain
developed on the basis of beam theory for a single bolt row working relationship, assuming the Young’s modulus to be infinitely large.
together with a finite-width strip of the web cleat angle section. A The ultimate moment Mu is practically unachievable. Taking the
complete connection can be treated as an assembly of a number of stress state shown in Fig. 2 as the final state, the tangent rotational
such bolt rows; each of which can have different but compatible stiffness can be defined as
deformations. Kr = Mu0 − My /θu0 .

(1)
2.1. Development of the mechanical model For a beam with equal but opposite moments at its ends, the length
of a plastic hinge can be assumed to be the length of the yielded
Fig. 1 shows one of the angles from a double web cleat part at each end. The moment and rotation of the hinge can then
connection. It has a finite width w and has one bolt through each be calculated as
leg. In the mechanical model, this is considered as two orthogonal 1

2

beams connected at the middle of the heel. The bolts attached to Mu0 = w t 2 f y + E t εu (2)
the column flange are assumed to be able to provide full fixity, 4 3
whereas those attached to the beam web allow movement in the L Mu − My
plane of the web. The two legs of the angle are then treated as θu0 = εu . (3)
t Mu
700 H. Yu et al. / Engineering Structures 31 (2009) 697–708

Fig. 2. Model to derive the behaviour of a plastic hinge.

Substituting Eqs. (2) and (3) into Eq. (1) and expressing My as Substituting Eq. (9) into Eq. (6) gives
1
4
wt 2 fy , the tangent rotational stiffness can be calculated from 1
L1 = L0 +
Et I Mu 2 (EI )2
Kr = 2 . (4)  
L Mu − My 1 F 1
× F 2 L50 − (FL0 − M0 ) L40 + (FL0 − M0 )2 L30 . (10)
20 4 3
In the following sections, the forces and deformations of the angles
in web cleat connections are solved by considering the changes of Calculating θ1,end and θ2,end from Eq. (9) and substituting them into
the state in the regions of these plastic hinges. Eq. (8) gives

Phase a. Elastic L20 L22


Ignoring the effect of axial force, the deflection profile of Beam F − M0 L0 + H − M0 L2 = 0. (11)
2 2
1 can be established as
This gives
L0 x 2 x3 x2
   
1
δ1 = L20
 
F − − M0 + C1 x (5) 2
EI 2 6 2 H = M0 (L0 + L2 ) − F . (12)
L22 2
where C1 is determined by the boundary condition. For Beam 2,
The deflection of Beam 2 can be written from Eq. (5)
the deflection profile is given by the same equation, but F and L0
should be replaced by H and L2 . L32 L22
 
1
At any applied load F , in order to calculate the deflections, three δ2,end = H − M0 . (13)
EI 3 2
equilibrium equations need to be established to solve for the three
variables H, M0 , and L0 . These are shown below: Substituting δ2,end from Eq. (13) into Eq. (7) gives

L32 L22
 
1. Deformation compatibility for Beam 1 dictates that the curved 1
L1 − L0 = H − M0 . (14)
length of Beam at any deflection should be equal to its original EI 3 2
length L1 .
Substituting Eq. (12) into Eq. (14) and solving for M0 gives
L0 L0
Z Z  
1
q
L1 = 1 + δ102 dx ≈ 1 + δ102 dx. (6) (L1 − L0 )EI + 13 FL20 L2
0 0 2 M0 = 1 2
. (15)
L
6 2
+ 23 L0 L2
2. Deformation compatibility at the joint of the two beams implies
that. Substituting M0 into Eq. (10) gives an equation in L0 , which can
be solved iteratively. When L0 is known, M0 and δ1,end can be
L1 − L0 = δ2,end (7) evaluated accordingly. The end of this phase is determined by
θ1,end + θ2,end = 0. (8) Mr1 = My .
Phase b. When My < Mr1 ≤ Mu and M0 ≤ My
These equations are solved for various yielding phases. During the
In the subsequent phases, the expression for the integration
loading, it is assumed that Mr1 > M0 > Mr2 . This is generally true.
in Eq. (6) becomes too complicated to be solved directly. An
In the elastic phase, for both Beams 1 and 2, δ 0 = 0 x=0 ;

incremental method is adopted instead. The applied force F is
therefore, C1 = 0. For Beam 1, increased by a small amount ∆F in every step. Each step uses
the value of L0 from the previous step, and this value is updated
x2
   
1 again at the end of this step for use in the next. For convenience in
δ10 = F L0 x − − M0 x . (9)
EI 2 calculation, M0 is used as the controlling incremental parameter.
H. Yu et al. / Engineering Structures 31 (2009) 697–708 701

Hence, for each increment in M0 , Eqs. (24), (14), (16) and


(17) together give

3EI (L1 − L0 ) + 3
2
M0 L22
H = (25)
L32
 
1
+ (Kr + HL0 ) 2∆M0 KEIr − 12 HL22 + M0 (L0 + L2 )

EI M0 + My − 2
HM0 L22
F = 1
EIL0 − 6
HL30 + 21 Kr L20
(26)
EI 1 1
C1 = 2∆M0 − FL20 − HL22 + M0 (L0 + L2 ) (27)
Kr 2 2
 
1 1 1
Fig. 3. Deformation of Beam 1 when both ends reach ultimate moment. δ1,end = FL30 − M0 L20 + C1 L0 . (28)
EI 3 2

Once these parameters are known, L0 is updated using Eq. (22).


In this phase, a plastic hinge is formed at x = 0 for Beam 1. The
This phase terminates when Mr1 = Mu .
boundary condition at x = 0 therefore becomes
Phase d. When Mr1 = Mu and M0 ∈ My , Mu


FL0 − M0 − H δ − My

In this phase, Eq. (17) should be replaced by
= Kr (16)
C1 /EI FL0 − M0 − H δ1,end = Mu . (29)
in which dM = FL0 − M0 − H δ − My is the increment of the Then, for each increment in M0 , Eqs. (24), (14), (16) and
hinge moment, referred to the end of the linear–elastic phase, and (29) together give
dθ = C1 /EI is the increment of the corresponding rotation.
From Eq. (5), 3EI (L1 − L0 ) + 32 M0 L22
H = (30)
L32
1 1
EI δ1,end = FL30 − M0 L20 + C1 L0 (17) δ1,end
3 2
2∆M0 · EI
· L0 − 16 Mu L20 + M0 1 2
+ L0 L2 − 12 HL0 L22

Kr
L
3 0
Eqs. (11) and (14) remain valid for this phase. Eqs. (11), (14), (16) = (31)
and (17) together give the following parameters for each increment EI + 16 HL20
of M0 : Mu + M0 + H δ
F = (32)
3EI (L1 − L0 ) + 32 M0 L22 L0
H = (18)
L32 EI δ1,end − 13 FL30 + 21 M0 L20
C1 = . (33)
1 H Kr H
− 12 HL22 + M0 (L0 + L2 ) L0
  
M0 + My − 2 EI
M0 L22 + EI
+ L
EI 0
F = 1 H 3 1 2 Kr
+ EIH L0

L0 − L + L When the values of these parameters are known, L0 is updated
3 EI 0 2 0 EI
using Eq. (22). This phase terminates when M0 = Mu .
(19)
From Phases b to d, all the above equations have been derived
1 1 based on the assumption that the fixed end of Beam 2 remains
C1 = M0 (L0 + L2 ) − FL20 − HL22 (20)
2 2 elastic. Whenever it is detected that Mr2 > My , then the
appropriate Eqs. (18), (25) and/or (30) giving H should be modified
 
1 1 1
δ1,end = FL30 − M0 L20 + C1 L0 . (21) accordingly:
EI 3 2

As F , M0 and C1 are known for this step, the integration in Eq. (6)
 Kr
∆Mr2 = HL2 − M0 − My = C2 (34)
can be expanded to give EI
1 1
1 EI δ2,end = HL32 − Mu L22 + C2 L2 . (35)
L1 − L0 = 3 2
2 (EI ) 2

  Replacing δ2,end with L1 − L0 , and eliminating C2 from Eqs. (34) and


2 5 1
F 2 L50 − FM0 L40 + M02 + 2FC1 L30 − M0 C1 L20 + C12 L0 . (22) (35) gives H as

×
15 12 3
EI (L1 − L0 ) + 12 M0 L22 + M0 + My L2 KEI

The new L0 can now be calculated iteratively from Eq. (22). This H = 1 3
r
. (36)
phase terminates when M0 = My . L
3 2
+ L22 KEIr
Phase c. When My < Mr1 ≤ Mu and My < M0 ≤ Mu Phase e. When Mr1 = Mu and M0 = Mu
As the ends of the two beams reach their yield moments, they When both Mr1 and M0 have reached Mu , further enhancement
start to rotate relative to one another. The continuity equation (8) of the resistance can only come from geometric nonlinearity. As
becomes both ends of Beam 1 have reached their ultimate moment capacity,
it actually rotates as a ‘‘link’’. An illustration of the effect of change
∆M0 of geometry for Beam 1 is shown in Fig. 3. With any further rotation
θ1,end + θ2,end = 2 . (23)
Kr dθ , the change in the vertical deflection is

Substituting the expressions for θ1,end and θ2,end into Eq. (23) gives dδ = L · dθ · cos θ . (37)
Therefore, the total deflection is
L20 L22
 
EI
F − M0 L0 + H − M0 L2 + C 1 = 2∆M0 . (24) δ1,end = δ1,end + dδ. (38)
2 2 Kr
702 H. Yu et al. / Engineering Structures 31 (2009) 697–708

The change in the horizontal deflection is


dδh = L · dθ · sin θ . (39)
The total shortening of Beam 1, or the lateral deflection of Beam 2,
is
δ2,end = δ2,end + dδh . (40)
From force equilibrium of Beam 1, the total applied force F should
be
(2Mu + H δ)
F = . (41)
L1 − δ2,end
In Eq. (41), the value of H depends on whether the fixed end of
Beam 2 is purely elastic, elasto–plastic or fully plastic. The bending
moment at its fixed support can be approximately calculated as
Mr2 = HL2 − Mu . (42)
At the end of each step the status of Mr2 should be checked. If it lies
between 0 and My , then for the next step

3EI δ2,end + 32 M0 L22


H = . (43)
L32
If Mr2 lies between My and Mu ,, then

Kr
∆Mr2 = Mr2 − My = C2 (44) Fig. 4. Determination of the equivalent support position.
EI
1 1 head/nut is positioned on the web cleat as shown in Fig. 4, the
EI δ2,end = HL32 − Mu L22 + C2 L2 . (45)
3 2 bolt head/nut is able to provide full restraint against lifting of the
Substituting Eq. (42) into Eq. (44) and eliminating C2 between web cleat along its inner edge. Beyond the circumference of the
Eqs. (44) and (45) gives bolt head/nut, the restraint is assumed to extend in the directions
of the hexagonal bolt head faces. The equivalent support position
EI δ2,end + 12 Mu L22 + Mu + My EI

Kr
L2 is calculated as the average distance of the supported edge to the
H = . (46)
1 3
L + EI
L22 centre-line of the bolt. If the horizontal distance between the two
3 2 Kr
lines of bolts connected to the column is defined as g, then the
If Mr2 = Mu , then length of Beam 1 can be calculated as
2Mu √ √ !
H = . (47) (g − tw ) t 3 3
L2 L1 = − − RH − w . (50)
2 2 2 12
During this phase the axial force is increased gradually, which
causes a progressive reduction in the moment capacity of the
plastic hinges. The interaction equation relating the capacity of 2.2.2. Determination of the material properties
a plastic hinge in a rectangular solid section subjected to a
This model assumes a tri-linear stress–strain relationship
combination of axial force and moment can be written as [19]
for the steel. When used in real applications, the stress–strain

N
2 
Mu
 relationship needs to be approximated to a tri-linear model.
+ = 1. (48) This section will discuss the determination of these properties,
Nn Mn especially Et on the basis of the stress–strain relationships
For every step, the axial force in Beam 1 can be calculated as proposed by EC3: Part 1.2 [20] for hot rolled steel. Two methods
are investigated to approximate the stress–strain curves into a
N = F sin θ + H cos θ . (49)
tri-linear model. The first is to simply use the secant modulus,
The moment capacity Mu of the plastic hinges should be and the other is based on a principle of minimum deviation. The
recalculated, considering the co-existence of axial force. The effects of these two approximations on the results produced by
ultimate capacity of the whole web cleat is reached when the angle the mechanical model are compared. At ambient temperature,
is in pure tension. the increase of stress beyond the yield point is due to strain-
hardening. Two typical stress–strain curves according to EC3: Part
2.2. Determination of parameters 1.2 are shown in Fig. 5. The higher curve represents stress–strain
relationships at 350 ◦ C and below, including a hardening phase,
2.2.1. Determination of L1 and the lower one represents the curves at higher temperatures.
When one bolt row of a web cleat connection is subjected It is easy to obtain both the linear–elastic phase and the plateau
to tension, the bending behaviour of the web cleat is three- phase from these two curves. When using the secant modulus
dimensional, as the restraint from the bolt diminishes towards the approach, the secant stiffness is determined by simply connecting
free edges at both sides. When this behaviour is approximately these two lines. When based on the minimum deviation principle,
considered using a two-dimensional model, an equivalent support the transition phase always ends at the start of the plateau.
position as the supported edge of Beam 1, as shown in Fig. 4. Its intersection with an extension of the linear phase gives its
Ignoring the effect of the washer and assuming that the bolt beginning. The slope of this phase is determined such that the area
H. Yu et al. / Engineering Structures 31 (2009) 697–708 703

Fig. 5. Transformation of the EC3 material model into a tri-linear model.

Fig. 8. Comparison of the mechanical model against finite element analysis at


various web cleat widths.

Fig. 6. Comparison of the mechanical model using different tri-linear model against
the finite element analysis results.

Fig. 9. Comparison of the mechanical model against finite element analysis at


various bolt line distance.

e, in which geometric nonlinearity dominates. It can be seen that


only the responses in Phases b, c and d are affected by selection of
the tri-linear model. The minimum-deviation model simulates the
shape of the finite element analysis results in the transition phase,
but with much smaller total deflection. This is due to the stiffness
of the linear elastic phase predicted by the mechanical model. By
using the secant modulus, the resistance at medium deflections
is obviously reduced. However, as the mechanical model always
gives too stiff a response in the linear–elastic phase, this material
model has in effect shown a better correlation with the finite
element analysis results, by intentionally using a weaker material
model. Therefore, for simplicity, the secant modulus is used in
subsequent calculations. At large deflections, the responses from
the mechanical model are not affected by the selection of the
material model, and they are both very close to the finite element
analysis results.
Fig. 7. Details of the finite element analysis model.

2.3. Verification of the mechanical model


covered by the original shape from εp to εy (εu for 350 ◦ C and under)
is the same as that enclosed by the new lines, as shown in Fig. 5.
Fig. 6 shows a comparison of the results from the simple tri- 2.3.1. Verification against finite element analysis
linear material models against the finite element analysis results Finite element analysis provides a suitable tool for testing
using the original EC3 material model at 300 ◦ C and 600 ◦ C the validity of the mechanical model, because it can simulate
respectively. The geometry of the analysed case is shown later in some of the idealised assumptions adopted, such as the material
Fig. 7, with w = 70 mm, g = 134 mm. The first four points on all properties. It can also trace the occurrence of each event (such
the mechanical model curves correspond to the ends of Phases a, b, as the onset of yield) and give a clear indication of its impact on
c and d respectively. The responses thereafter correspond to Phase the force–displacement relationship. In addition, finite element
704 H. Yu et al. / Engineering Structures 31 (2009) 697–708

analysis allows parametric studies to be performed at little cost, 3.1. Introduction to the component behaviour
and hence the validity of the model can be checked for a wide
range of applications. This section compares the mechanical model A typical web cleat connection involves 4 springs in series for
to finite element analysis results for a simple case, in which two each bolt row, as shown in Fig. 11. These represent bolts in tension,
web cleats were used to connect the beam web to a fixed rigid web cleats in bending, bolts in shear and the beam web in bearing.
plate, as shown in Fig. 7. A pulling force normal to the rigid In this section, this model is used to simulate the twelve web cleat
plate was applied to the beam web. A three-dimensional finite connection tests reported in the companion paper [6]. Considering
element model was created using solid elements, and all bolts were that, in the current set of tests, the behaviour of the web cleat in
explicitly modelled. The material properties of the web cleats are bearing is not critical, it is not considered, but can be added where
shown in the figure. To prevent bolt failure, the bolts were assumed necessary. The properties of the springs are introduced separately.
to remain elastic. Both the beam web and the fixed plate were
assumed to be rigid. a. Bolts in tension
Two parameters were studied; the width w of the web cleat The properties of bolts in tension are simply derived from
and the distance g between the bolt lines. A comparison of the the tested stress–strain relationships [21] at high temperatures
results from the mechanical model against those from the finite for the bolt material. Given the existence of threads, the net
element analysis is shown in Figs. 8 and 9. The displacements from cross-sectional area of 254 mm2 is used for the M20 bolts. The
the finite element model were measured at the axis of the bolt effective length of the bolt is calculated as the total connected
connecting the two web cleats to the beam web. The mechanical plate thickness plus half of the nut thickness. This component
model is very stiff in the elastic phase, because it assumes that the is generally both stiff and strong, and its properties have little
two bolts connecting the angles to the plate create a fixed boundary influence on the overall behaviour of the connection.
condition. At larger deflections the mechanical model generally b. Web cleat in bending
gives good correlation with the finite element analysis. It can be The force–displacement relationship for the web cleat has been
seen from Fig. 8 that the equivalent restraint position works very derived in the previous section.
well for different web cleat widths.
c. Bolts in double shear
2.3.2. Verification against test results Generally, bearing failure of bolts is not critical. However, it
It is always best to test the applicability of a simple model has been observed in the current tests that bolts can experience
to be used in practical design against test results. Owens and significant bearing deformation at elevated temperatures. It has
Moore [18] performed 11 tests on web cleat connections subjected been explained in the companion paper [6] that this is due
to normal tension. Among these, five tests were repeated, and to the combined effect of the bending and bearing forces. One
therefore a total of six tests were modelled using the mechanical inherent shortcoming of the component-based model is that it
model. The geometrical and material data of the tested connections cannot consider the interactions between different components.
are given in their paper, and this data will not be repeated here. As the bolt bending is caused by the opening of the angles, this
No information was provided about the steel’s elastic properties effect cannot be considered directly. However, a simple model is
and so, in the mechanical model, the Young’s modulus was set to introduced to consider the possible bearing deformation of the
210 kN/mm2 and the tangent modulus was calculated assuming bolts. Considering the bolt behaviour described in the previous
that the ultimate stress was achieved at a total strain of 0.04. paper, it is assumed here that the bolt starts to deform by bearing.
A comparison of the predictions from the mechanical model With further deformation, this bearing resistance increases, but the
against the test results is shown in Fig. 10. For connections with bolt’s shear resistance decreases. Beyond the point at which they
multiple rows of bolts, the total resistance is simply calculated become equal, the deformation of the bolt is controlled by shear
as the sum of the resistances of all the bolt rows. It should alone. According to EC3: Part 1.8 [8], the nominal shear and bearing
be mentioned that the reported test displacements were total resistances of an individual bolt are given by
displacements measured at the beam web. For small deflections,
it is to be expected that the test results will show much larger FV ,Rd = αV fub A (51)
deformations than the mechanical model. The failure mode for Fb,Rd = k1 αb fub dt . (52)
each test is shown in the figure. At large deflections, when the
failure is by fracture of the web cleat, the contribution from To simulate the current tests, use 0.6 for αV , 1.0 for αb and 2.5 for k1 .
other components to the total deformation is probably negligible. The behaviour of Grade 8.8 M20 bolts in double shear has been
Therefore, Tests 4 to 6 seem to indicate that the mechanical investigated experimentally at the University of Sheffield. One test
model gives high resistance, or smaller displacements, at large at ambient temperature was reported by Yu et al. [22]. The bolt
deflections. When the failure is not controlled by the web cleat, the reached a maximum resistance of 311 kN and fractured suddenly;
displacements from the tests can include significant contributions its pure shear deformation at fracture was about 2.5 mm. At
from other components, which partially explains why the test elevated temperatures, bolts in shear behaved differently, in that
result for Test 3 gives significantly higher deflection. the parts on either side of the shear planes were sheared gradually
with a progressively decreasing resistance. Several double-shear
3. Component-based model tests were performed; one of the test curves, at 450 ◦ C, is
shown in Fig. 12. Based on these test results, it is proposed that
The test results are of limited direct use in practical design, bolts are assumed to reach their maximum shear resistance at a
since in real structural detailing, the type of a connection and its displacement of 2 mm, and maintain this resistance up to 3 mm.
loading condition can differ significantly from those tested here. From 20 ◦ C to 400 ◦ C, bolts fracture in a brittle manner at about
One of the major objectives of this research project was to develop
3 mm; from 400◦ C upwards, the shear resistance reduces linearly
a set of component-based models for commonly-used types of
to zero. A mathematical description is
connection. The component-based method divides the joint into
a set of uncoupled standard components, the behaviour of which −0.2δ 2 + 0.9δ δ ∈ (0, 2)


can be described mathematically as nonlinear springs. The overall 
1 δ ∈ (2, 3)
FV

behaviour of the joint can then be calculated by assembling these = 0 δ ∈ (2, 3) &T < 400◦ C (53)
springs. The test results can be used to assess the validity of the FV ,Rd
 1 (20 − δ)

δ ∈ (3, 20) &T > 400 C . ◦

component-based method.

17
H. Yu et al. / Engineering Structures 31 (2009) 697–708 705

Fig. 10. Comparison of the predictions from the mechanical model against test results.

Fig. 12 shows a comparison of Eq. (53) against test results. resistance, the net resistance is assumed to drop linearly to zero
Little research has been done on the behaviour of bolts in at 20 mm displacement.
bearing deformation. The force–displacement relationship for a d. Beam web in bearing
bolt in bearing is simply derived by considering the change of the Sarraj [15] proposed a model for plates in bearing at the edges
bearing surface with the bearing deformation as shown in Fig. 13. of bolt holes. This model has been used successfully to construct
At any displacement δ , the bearing width perpendicular to this a component-based model for fin plate connections, and this has
displacement is B. The bearing resistance is then been directly adopted here. However, it should be borne in mind
that this model was developed for fin plate connections, and
therefore considers out-of-plane bending of the connected plates.
2.5 × fub × B × t δ < d/2

Fb = (54) In web cleat connections, the beam web is subjected to pure in-
2.5 × fub × d × t δ > d/2 plane loading and the bearing resistance may be higher.
p The bottom spring in Fig. 11 simulates the effect of the bottom
where B = 12 d2 + (d − 2δ)2 . flange of the beam coming into contact with the column flange at
With the increase of the bearing deformation δ , the shear high rotation. The stiffness of this spring is initially zero, but when
resistance is then calculated based on the reduction of the its displacement is equal to the initial gap between the beam end
shear area. After the bearing resistance has exceeded the shear and the column flange, its stiffness becomes infinitely large.
706 H. Yu et al. / Engineering Structures 31 (2009) 697–708

Table 1
Tri-linear material properties for the web cleat.
Temperature (◦ C) fy fu E Et

20 354.5 455 135 000 5784.5


400 220 395 67 700 2405.5
500 177 236 58 800 1311.4
600 94.7 119 37 400 884.7
700 44.1 55.6 12 100 436.3

the companion paper. The effective tri-linear material properties


adopted for prediction of the web cleat behaviour are shown in
Table 1. The force–displacement relationships for each of the four
springs in series at each bolt level, shown in Fig. 11, based on the
equations developed in the previous section, are shown in Fig. 14
for the four test temperatures. Both forces and displacements are
shown as negative for tension. The compressive stiffness of each
component is assumed to be very large. Comparison of the four
Fig. 11. A component-based model for web cleat connection.
curves at each temperature shows that the weakest components,
which control the failure, are the beam web in bearing at 20 ◦ C and
the bolts in shear at higher temperatures. If fracture of web cleats
is not considered, the failure mode indicated by the component-
based model is actually in agreement with the test observations.
If allowed to develop large deflections, web cleats can have very
high resistances. However, a comparison with the test observations
shows that the predicted deformation of the web cleat is lower
than the test results. It can be seen from this figure that, at the
failure point of each bolt row, the maximum deformation of the
web cleat is around 20 mm. It has been observed from the tests that
the deformation of the web cleats before fracture is around 30 mm
for tests at 450 ◦ C and 550 ◦ C and is higher at 20 ◦ C and 650 ◦ C. From
450 ◦ C to 650 ◦ C it can be seen that, with increasing temperature,
the bolts in shear become progressively weaker in comparison to
the beam web in bearing. This also agrees with test observations.
Fig. 12. Behaviour of bolts in double shear at elevated temperature. Initially, these force–displacement relationships were intro-
duced into the general FEM program ABAQUS [23] simply as prop-
erties of spring elements. However, the analysis was unable to
reach the peak resistance of the connection, due to transient in-
stabilities caused by the failure of some components. A simple
method was then implemented in a self-written program, in which
the overall behaviour of each bolt row is defined as an equivalent
spring. Rotation-controlled analyses were run by increasing the ro-
tation of the connection step by step and solving for the force in
each bolt row. The centre of rotation was initially set at the mid-
dle bolt row for small rotations, and then shifted to the bottom
flange when the axial compressive movement of the bottom spring
reached the value of the initial gap between the column flange and
the beam end. The total force acting on the connection was then
calculated as the resultant of the forces in all the bolt rows.
The force-rotation relationships from the component-based
model are compared to the test results in Figs. 15–17. Several
observations can be made from these comparisons:
Fig. 13. Model for bolts in bearing deformation.
The component-based model in general gives a stiffer response
than the test. One of the reasons for this has been explained
3.2. Simulation of the tests using the component-based model above as an under-estimation of the deflection of the web cleats.
This could be partially due to the inherent characteristics of the
The material properties of the bolts, web cleat angle and the mechanical model. It can be seen from the tests that the bolts
beam were tested at ambient temperature. It is difficult to obtain connecting the cleats to the column flange achieved significant
a reliable and complete stress–strain curve for the steel; the bending deformation, which indicated little rotational restraint to
parameters which are easily identifiable are the yield stress and the the web cleat. The mechanical model, however, always assumes
ultimate stress. For the angles used, these were fy = 340 N/mm2 full restraint. Another reason could be a degree of uncertainty
and fu = 455 N/mm2 respectively; for the beam they were fy = about the material properties. At 650 ◦ C, the finite element
356 N/mm2 and fu = 502 N/mm2 . The bolts were Grade 8.8 simulation results presented in the companion paper and this
M20, zinc-plated and fully-threaded. Their average ultimate tensile simple model both give significantly lower deformation of the web
capacity, obtained from three specimens at ambient temperature, cleats.
was 224 kN. Their properties at elevated temperatures have been The predicted resistances are generally unconservative, and
extrapolated on the basis of other heated tests, as discussed in differences between the simple model and the tests become
H. Yu et al. / Engineering Structures 31 (2009) 697–708 707

Fig. 14. Properties of the components at various temperatures.

progressively higher with the increase of the load angle. This is


because shear resistance is not considered in the component-based
model. The effect of the applied shear force is simply considered to
generate a bending moment. The effect of shear force in reducing
the resistance of each individual component in the direction of the
tying force has not been considered.
The differences between the peak resistances given by the
simple model and the test results are more significant at 450 ◦ C and
550 ◦ C. This is because, for these two tests, the bolts in shear have
not developed their full resistances, and the failure was controlled
by fracture of the web cleat.
In general, this component-based model has given an accept-
able prediction of the test behaviour. Work which might be done
to improve its performance in the future includes:
Consideration of the bending deformation of the bolts connect-
ing to the column flange;
Inclusion of a shear spring in the component-based model and
consideration of its interaction with the tensile spring;
Experimental investigation of the double-shear behaviour of
bolts with the co-existence of other actions at high temperatures. Fig. 15. Comparison of the component-based model against the test results at 35◦ .
708 H. Yu et al. / Engineering Structures 31 (2009) 697–708

shear force applied to a connection can reduce its tying capacity.


The current component-based model is unable to consider this and
overestimates overall connection resistances.

Acknowledgments

The authors gratefully acknowledge the support of the En-


gineering and Physical Sciences Research Council of the United
Kingdom under Grant EP/C510984/1. Provision of the steel sections
by Corus Ltd is also acknowledged.

References

[1] NIST. Final report on the collapse of the World Trade Center towers. USA:
National Institute of Standards and Technology; 2005.
[2] Newman GM, Robinson JT, Bailey CG. Fire safety design: A new approach
to multi-storey steel-framed buildings. UK: The Steel Construction Institute;
2004.
Fig. 16. Comparison of the component-based model against the test results at 45◦ . [3] Ding Jun. Behaviour of restraint concrete filled tubular columns and their joints
in fire. Ph.D. thesis. UK: The University of Manchester; 2007.
[4] Leston-Jones LC. The influence of semi-rigid connections on the performance
of steel framed structures in fire. Ph.D. thesis. UK: The University of Sheffield;
1997.
[5] Al-Jabri KS. The behaviour of steel and composite beam-to-column connec-
tions in fire. Ph.D. thesis. UK: The University of Sheffield; 2000.
[6] Yu Hongxia, Burgess IW, Davison JB, Plank RJ. Typing capacity of web cleat
connections in fire part 1: Test and finite element simulation. Eng Struct 2009;
31(3):651–63.
[7] Tschemmernegg F, Tautschnig A, Klein H, Braun Ch, Humer Ch. Zur
Nachgiebigkeit von Rahmenknoten – Teil 1. Stahlbau 56. Heft 10. 1987.
p. 299–306.
[8] European Committee for Standardization (CEN). BS EN 1993-1-8. Eurocode
3: Design of steel structures. Part 1.8: Design of Joints. UK: British Standards
Institution; 2005.
[9] Da Silva LS, Santiago A, Real VP. A component model for the behaviour
of steel joints at elevated temperatures. J Construct Steel Res 2001;57:
1169–1195.
[10] Al-Jabri KS, Burgess IW, Plank RJ. Spring-stiffness model for flexible end-plate
bare-steel joints in fire. J Construct Steel Res 2005;61:1672–91.
[11] Al-Jabri KS. Component-based model of the behaviour of flexible end-
plate connections at elevated-temperature. Compos Struct 2004;66:
215–221.
Fig. 17. Comparison of the component-based model against the test results at 55◦ . [12] Spyrou S, Davison JB, Burgess IW, Plank RJ. Experimental and analytical
investigation of the tension zone component within a steel joint at elevated
temperatures. J Construct Steel Res 2004;60:867–96.
4. Conclusions [13] Spyrou S, Davison JB, Burgess IW, Plank RJ. Experimental and analytical
investigation of the compression zone component within a steel joint at
Component-based models, which can describe the behaviour elevated temperatures. J Construct Steel Res 2004;60:841–65.
[14] Block Florian. Development of a component-based finite element for steel
of a connection using an assembly of springs, provide a practical beam-to-column connections at elevated temperatures. Ph.D. thesis. UK: The
solution to include realistic behaviour and possible failure of University of Sheffield; 2006.
connections in structural analysis for fire safety. A component- [15] Sarraj Marwan. The behaviour of steel fin plate connections in fire. Ph.D. thesis.
UK: The University of Sheffield; 2007.
based model for web cleat connections is proposed in this paper. [16] Yu Hongxia, Burgess IW, Davison JB, Plank RJ. Experimental investigation
It includes four springs for each bolt row, representing bolts in of the behaviour of fin plate connections in fire. J Construct Steel Res
tension, web cleats, bolts in shear, and the beam web in bearing. A [in press].
[17] Ramli-Sulong NH, Elghazouli AY, Izzuddin BA. Behaviour and design of
mechanical model has been developed to describe the behaviour of beam-to-column connections under fire conditions. Fire Safety J 2007;42:
the angle in a web cleat connection. This model is based on classical 437–451.
beam theory, and considers the formation of four plastic hinges for [18] Owens GW, Moore DB. The robustness of simple connections. Struct Eng 1992;
each angle. The effect of large deformation and opening of the angle 70:37–46.
[19] European Committee for Standardization (CEN). BS EN 1993-1-2. Eurocode 3:
is included. This model has been validated against finite element Design of steel structures. Part 1.2: General rules and rules for buildings. UK:
simulations and previous test results, and is found to give a very British Standards Institution; 2005. p. 54.
good prediction of the responses of web cleats in pure tension up [20] European Committee for Standardization (CEN). BS EN 1993-1-2. Eurocode 3:
Design of steel structures. Part 1.2: General rules- structural fire design. UK:
to very high deflections. British Standards Institution; 2005.
Failure criteria have been introduced for most components and, [21] Theodorou Y. Mechanical properties of Grade 8.8 Bolts at elevated tempera-
when used to simulate the behaviour of the tests on connections tures. Master’s Dissertation. UK: The University of Sheffield; 2003.
[22] Yu HX, Burgess IW, Davison JB, Plank RJ. Experimental investigation of the
reported in the companion paper, have enabled predictions of behaviour of fin plate connections in fire. In: Proceedings of ICSCS 2007; 2007.
both their peak resistance and rotational capacity. The simulation p. 541–548.
results are generally satisfactory. It should be highlighted that the [23] ABAQUS Analysis User’s Manual Version 6.5. ABAQUS Inc.; 2004.

You might also like