Engineering Structures 76 (2014) 177186

Contents lists available at ScienceDirect

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

Numerical evaluations of codied design methods for progressive

collapse resistance of precast concrete cross wall structures
M. Tohidi a,b, J. Yang c,d,e,, C. Baniotopoulos f
a

School of Civil Engineering, University of Birmingham, Birmingham B15 2TT, UK

Azad University of Sananadaj, Iran1
c
School of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR China
d
State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR China
e
School of Civil Engineering, University of Birmingham, Birmingham, UK2
f
School of Civil Engineering, University of Birmingham, UK
b

a r t i c l e

i n f o

Article history:
Received 23 July 2013
Revised 18 June 2014
Accepted 19 June 2014

Keywords:
Cross wall construction
Progressive collapse
Catenary action
Ties force method
Ductility
Translator element
Bond
Post-bond-failure analysis

a b s t r a c t
Progressive collapse of building structures typically occurs when an abnormal loading condition causes a
sudden loss in the structural capacity of one or more critical members, which leads to a chain reaction of
failure and ultimately catastrophic collapse. The tensile tie force (TF) method is one of the main design
approaches for considering progressive collapse. As this method does not take into account factors such
as dynamic effect, the load redistribution mechanism, the effect of steelconcrete interfacial properties,
or the size and embedment length of tie bars on bond behaviour, it can be considered as a simplied
method, and hence a thorough examination of the adequacy of this method is needed. This paper reports
such a study including numerical evaluation of the codied methods of progressive collapses for precast
concrete cross wall buildings. To this end, detailed three-dimensional nite element models of the pullout behaviour of strands in the keyway of precast concrete blocks and of the ductility behaviour of oor
joints subjected to uniform and line loads exerted from upper walls were developed. Through a calibration process for a series of laboratory pullout tests carried out by the Portland Cement Association (PCA),
the interfacial bond properties were established using numerical modelling. The same modelling method
was then used in the subsequent three dimensional non-linear numerical analyses to simulate the ductility behaviour of precast concrete oor joints in the absence of underlying wall supports. In both modelling processes, the simulation of the bondslip behaviour at the steelconcrete interface was realised by
using the translator element embedded in ABAQUS. The numerical analyses showed a close agreement
between FE analyses and test results. The tie force developed during the collapse process was particularly
examined. Discrepancies in the tie force between the numerical and the codied specications have
suggested an underestimate of tie force in the TF method that may lead to an unsafe design. Finally,
an improved model based on the numerical results has also been proposed to address this problem.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
As is dened by the Portland Cement Association (PCA) [1], the
term large-panel concrete structure is used to describe a building
system consisting of vertical wall panels together with precast concrete oors and/or roofs. Large panel buildings are featured as
examples of wall panels being used as the load-bearing structure.
Corresponding author at: School of Naval Architecture, Ocean and Civil
Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR China. Tel.: +86
13917654726.
E-mail addresses: j.yang.3@bham.ac.uk, yangjian2000@hotmail.com (J. Yang).
1
On study leave.
2
On leave.
http://dx.doi.org/10.1016/j.engstruct.2014.06.034
0141-0296/ 2014 Elsevier Ltd. All rights reserved.

In the usual arrangement, a wall that is perpendicular to the longitudinal axis of a structure is referred to as the cross wall and that is
parallel to the longitudinal axis is termed the spine wall. In the
cross wall system, oor/roof slabs are typically one way hollow
core precast concrete slabs, and only cross walls carry the oor
loads (see Fig. 1).
To avoid the progressive collapse of a precast concrete cross wall
structure, in the event of a sudden loss of a support wall, it is required
that damage will be limited to the affected zone only, e.g. in the
vicinity of the damaged wall (see Fig. 2). This means that the remaining structures will stay in place without any chain-reaction type of
collapse. One of the typical responses of oor-to-oor joints in the
above scenario is that, due to the dynamic nature of the event, the

178

M. Tohidi et al. / Engineering Structures 76 (2014) 177186

Fig. 1. Examples of precast concrete wall construction (courtesy of Bison).

Damaged zone

Catenary action of
floor system

Ineffective
wall support

Fig. 2. A typical damage scenario for a cross wall structure.

impact force from the upper oors will exert on the oor-to-oor
joint immediately above the damaged wall. This impact force will
rapidly crush the in-situ lled grout and produce a large deection
in the oor joint. The damage to the grout and the development of
a large deection will dissipate the impact energy. The latter also
renders a reduction in the force from the upper walls and diverts
loads to the adjacent walls. In so doing, a catenary action is developed in the oor slabs adjacent to the damaged wall. It is believed
that the key for the catenary action to work successfully is that longitudinal ties at the joint have sufcient strength and deformation
capability, which closely depends on the bond performance of ties
in the grout (see Fig. 3).
Following the partial failure of a precast concrete building
based in London, Ronan Point apartment [2], in 1968, the British
Standards for concrete structures [3] started to incorporate provisions to deal with the problem of progressive collapse. The Portland Cement Association [1] conducted a series of comprehensive
investigations to form an underpinning knowledge basis supporting the stipulated minimum detailing requirements to ensure the

development of an alternative load path (ALP) in the event of any

local damage [46]. These attempts led to a tie-force (TF) design
method which is a rst of its kind in the world. This method, which
is mainly of a prescriptive nature, requires the inclusion of internal,
peripheral and vertical ties (see Fig. 4) to provide different alternative load paths, e.g. catenary, cantilever, vertical suspension
and diaphragm actions, in the event of the loss of underlying wall
support. These prescriptive tie requirements may have proven adequate in engineering practice but are not fully scientically
justied, so substantial efforts are still needed to improve the
understanding, at a fundamental level, of how the mechanism of
post-collapse resistance is developed through these tie provisions.
This need has also been supported by a number of researchers in
the last decade.
Dusenberry [7] indicated the necessity of a better understanding of the mechanism by which progressive collapse can be
resisted. The UK Building Research Establishment (BRE) has conducted a series of quarter-scale tests to verify the adequacy and
reliability of the tie force method [8]. To show the adequacy of
the codied methods for progressive collapse, an evaluation on
three well known collapsed building cases was performed by Nair
[9] based on ve current codes of practice or standards. Results
revealed that all three studied structures are susceptible to
progressive collapse. Abruzzo et al. [10] has also indicated the
inadequacy of the TF method to prevent progressive collapse of
structures. The necessity of developing an improved TF method
has also been recommended by the US Department of Defense
(DoD) [11]. According to experimental study on single beams,
Merola [12] showed that the tie rules are effective against progressive collapse when steel of ductility class C is used. To investigate
the efciency of the TF design method, Li et al. [13] also conducted
comprehensive numerical studies on two reinforced concrete (RC)

Precast concrete wall

Grout in joint
and keyway

Dry pack
Longitudinal tie in
keyway

13 mm
50 x 3 mm
bearing pad

Precast concrete
hollow core slab
50 mm

(a) Section view of a floor-to-floor joint

(b) Examples of horizontal ties

Fig. 3. Typical longitudinal ties arrangement at the oor-to-oor joint.

M. Tohidi et al. / Engineering Structures 76 (2014) 177186

L=Longitudinal
T=Transverse
V=Vertical
P=Peripheral

Fig. 4. Overall tie requirements [1].

structures of 3 and 8 stories, respectively; results were veried by

the experimental work of Yi et al. [14]. The numerical results
revealed that the current tie force method cannot provide a
safeguard to progressive collapse for all RC structures that have a
different number of stories and experience damage in different
locations; accordingly, an improved TF method was proposed.
Based on numerical assessment results of the disproportionate collapse, Gerasimidis et al. [15] suggested that a structure could
respond better if damage is distributed in two adjacent elements
rather than in only one element. Based on a recent study of the
analytical evaluation of the TF method, Tohidi et al. [16] found that
the current TF method remarkably underestimates the tie force
requirement at the oor joints in the precast concrete cross wall
structures. This conclusion has been further conrmed by the
latest edition of DoD design criteria published in 2013 [17], in
which the required tie force has been increased signicantly compared to the previous recommendations quoted in DoD [11] and
British standard [3].
Another criticism of the TF method has been recently raised by
several researchers who believe that the TF method does not take
into account the overall behaviour of the structure as a whole, particularly for the structures susceptible to buckling failure, e.g., steel
frame structures [1820]. In addressing this issue, they proposed
that a global analysis of the progressive collapse should be performed by considering the loss of stiffness in local regions. However, in order to do so, the key step is to characterise the real
behaviours of the local regions, such as joints, during the progressive collapse. This paper reports such a study, and the outcomes
can be utilised in the future exercise of the global stability analysis
of precast concrete cross wall structures.
Through the review of the TF method, it has also been found
that this method has not taken into account the effect of bond
behaviour between tie bars/strands and the surrounding grout.
Such behaviour is inuenced by many factors, such as bar/
strand-grout interface characteristics, stress-slip relationship, the
material properties, the diameter, interfacial modulus, and embedment length of tie bars/strands. This type of lack may also attribute
to the abovementioned inadequacy and will be addressed in this
study.
With regards to the subject of the bond behaviour of steel bars
in concrete, although a large body of research work has been available in the published literature, most of this was related to the
crack damage analysis and the pre-bond-failure performance of
concrete reinforced members [2130].
Similarly, research with the focus on the progressive collapse, in
particular relative to the post-bond behaviour, is rather limited for
precast concrete (PC) structures. For reinforced concrete (RC)
structures, however, several numerical analyses can be found,
which are based on the global 3-D models of skeleton structures,
where the detailed joint effect was ignored [3134].
The rst step of the present study is to address the abovementioned gap as to the deciency of post-bond failure analysis and it

179

is then followed by the second step, where the mechanism of forming a catenary action in relation to the bond behaviour in precast
concrete cross wall structures is examined. To that end, numerical
studies on the pull-out behaviour of strands in concrete and the
catenary behaviour of oor joint system due to the removal of wall
support are performed, from which key inuencing factors are
identied. By using the obtained results, the adequacy of the current TF method as recommended by codes of practice is evaluated.
2. Tie force (TF) method
The TF method requires that in each direction ties should be
designed to carry a tensile force of P (kN/m) equal to the greater
of the following two values [3,11,35]:

P1

g k qk Lr
Ft
7:5
5

P2 F t

1
2

where (gk + qk) is the sum of the characteristic permanent and variable oor loads (in kN/m2); Ft is the lesser of (20 + 4no) or 60 kN/m,
where no is the number of storeys; Lr is the length of the oor span.
In implementing the TF method that is adopted in most codes or
standards, an indeterminate structure is usually simplied to a
determinate one by introducing hinges at connections. This will
enable the minimum tie forces to be calculated. Based on the calculated results of tie force, sufcient tie arrangements are made
to provide sufcient strength to establish overall structural integrity, continuity and redundancy. This method is suitable for hand
calculation and results are inevitably of an approximate nature.
Recently, with the advancement of computer tools, an alternative
load path (ALP) method has become more popular. In this method,
following the removal of a critical element, the structure should be
capable of redistributing loads to the remaining undamaged
structural elements.
3. Catenary action
According to the current code specications, in order to prevent
progressive collapse of building structures, four types of alternative load path should be provided, i.e.





catenary action of oor-to-oor system,

cantilever and beam action of wall panels,
vertical suspension of wall panels, and
diaphragm action of the oor slabs.

In this study, only the catenary action of oor-to-oor systems

(see Fig. 5) is considered, so it is assumed that all other load paths
have been effectively provided. If an underlying wall support is
suddenly removed due to an abnormal load, in order to bridge
out the load exerted by the upper walls and hence retain the structural integrity, a continuity requirement at the oor-to-oor joints
must be provided so that an alternative load path can be found, i.e.
via catenary action in this case. Unlike the normal service condition, a much larger deection in the affected zone is allowed in
an accidental event. Therefore, sufcient ductility of these connections is essential to satisfy the deection demand. In precast crosswall constructions, these requirements can be facilitated by the tie
strands/bars embedded in the cast in-situ grout in keyways and the
side edge gap of oor slabs (see Fig. 5). After an underlying wall
support is removed, the grout will be crushed immediately under
the increased loads and these ties will experience tensile forces
and develop large deections in the oor slabs. This process forms
a catenary action (see Fig. 5).

180

M. Tohidi et al. / Engineering Structures 76 (2014) 177186

Wall panels

Floor slab

T Fv

UDL q

lm

Fv

Fl

ls

Fl
2lb

Longitudinal ties

(a) Damage affected zone

(b) Idealised catenary system

Fig. 5. Catenary action facilitated by longitudinal ties.

An equilibrium equation of the catenary system can be derived

by taking moments about a side support in the free body diagram
of the half system as shown in Fig. 5b.

Fl

wlb aqbp lb
2ds

3a
2

F l 1 a

Let q wlb ;

wbp lb
2ds

3b

where w is the uniformly distributed load (including permanent

and variable loads); bp the spacing of ties; lb the oor span length;
Fl the force in the longitudinal tie joining adjacent slabs; ds the vertical displacement at the middle wall support; q the line load
exerted by the upper wall and a the percentage increase of the line
load considering the number of storeys (see Table 1).
Based on the compatibility condition of deformation in
Fig. 5(b):

q
2
2
lb d2s  lb
s
!
 
ds
1
1
dl lb
db

dl

 2
dl 1 ds

lb 2 lb

if

4a
4b

 
dl
1
lb

where dl represents the increase in the length of each oor slab,

which consists of the extension of ties at both ends of the oor slab.
If we use dls and dlm to represent the extension experienced at the
side and middle supports of one of the affected oor slabs, we have

dl dls dlm

During the development and evolution process of the catenary

system, the tie force will reduce with an increase in vertical deection as indicated by Eq. (3). The increase in deection is facilitated
by the extension of oor-to-oor joints including the elongation of
tie strands and the slip of the strand out of the surrounding grout.
Failure of the catenary system occurs when the extension reaches a
certain level. The corresponding deection at the joint has often
been set as the failure criterion. At failure, the tensile force in the
tie has usually reduced below the yield stress and therefore most
of the extension is provided by the slip due to the pullout action.
To analyse a catenary action mechanism, the following assumptions are considered [1]:
 Local damage, and therefore the initial failure of slab, will not
affect the ability of the system to develop a catenary action
mechanism.

 After establishing the catenary action, a static behaviour of the

system can be assumed.
 In the event of the removal of the underlying support walls, sufcient transverse, vertical, and peripheral ties have been provided, so the whole structure remains stable.
 All extension demand is provided by the elongation of the longitudinal ties and the slip between the longitudinal ties and the
surrounding grout.
 Adequate longitudinal continuity has been provided to establish
catenary action for the oor-to-oor system.
It is noted that to satisfy the large deection requirement for
the catenary action mechanism, according to Eqs. (3b) and (4b), a
large extension should be provided at each joint but at the same
time the tie force can be quite small. As a result, it is anticipated
that the catenary action will occur in the post-bond-failure stage.
Fig. 5 also shows that after removing the middle supporting
wall, a high level of tie force at both side walls will be induced.
The horizontal components of the tie forces will be primarily
transferred to the adjacent slab and eventually to the bracing
system of the building. The vertical components will be sustained
directly by the side supporting walls.
4. FEA modelling
The PCA conducted pullout tests for strands embedded in the
grout lled in the keyways of 41 precast concrete blocks to study
their pre- and post-bond behaviours [1]. The blocks were cut from
the precast hollow core slabs and have the following dimensions:
1000 mm or 600 mm in width, 200 mm in height, and a variable
length, with an aim to study the effect of embedment length (see
Fig. 6). Two strand sizes are considered in the test, i.e. 9.5 mm
(3/8 in.) and 12.7 mm (1/2 in.). In order to seek an appropriate
numerical approach to model the grout-steel interface, these tests
were reproduced by using a commercial FEA package, ABAQUS, in a
preliminary phase of the study. In this stage, 3D models of steel
strand in the grouted keyways of a concrete slab block were generated. The dimensions of this model were chosen based on the pullout test congurations.
Once the modelling technique for the grout-steel interface has
been veried by test results, it is then adopted in the modelling
process to reproduce the full-scale oor-to-oor joint tests, also
carried out by PCA [1]. In these tests, the system consisting of
two hollow core precast concrete slabs of full width, which were
connected through two or three steel ties placed into two keyways
(see Fig. 7), was subjected to uniform and central line load to
Precast slab

Keyway

Strand
Table 1
The percentage increase of the line load with the number of storeys a (BS 811011:1997).
Storey no.

10

a%

17

33

50

67

83

100

117

133

150

Grouted Keyway

(a) Plan view

(b) Longitudinal sectional view

Fig. 6. Illustrative diagram of the pullout test by PCA (19751979).

M. Tohidi et al. / Engineering Structures 76 (2014) 177186

imitate the load exerted by the upper walls. All longitudinal ties
were seven wire strands, which were placed symmetrically into
keyways in the middle and side joints.

181

Fig. 8a. This connector dictates kinematic constraints by combining

connection types with the options of SLOT and ALIGNS [49].
4.2. Mesh description and boundary conditions

4.1. Modelling techniques for the bondslip behaviour

Since bond is the key factor in the analysis and design of RC
structures, and it governs most RC performances, not just progressive collapse, seeking a technically reliable and economically viable
bond modelling technique remains a challenging issue. To date,
numerous research papers have been published which study the
bondslip behaviour between the tie and the surrounding grout,
a large proportion of which were carried out by numerical modelling. In these papers, a wide range of modelling techniques has
been adopted to simulate such behaviour. Bresler and Bertero
[36] rst introduced a layerwise model. Since bond only occurs
in the concrete zone near the reinforcement surface, to differentiate the inelastic deformation and fracture damage in this zone
from the bulk concrete, the concrete is divided into two zones,
i.e. an inner boundary layer and an outer layer. It was assumed that
both zones have a linearly-elastic isotropic behaviour, but with different material properties. Reinhardt et al. [37] later introduced a
slip layer, which was divided into two layers, one with a thickness equal to the bar diameter and the other equal to the outer
zone of concrete. The steel bar was assumed to be elastic. The nonlinearity of the concrete layer was described by an elastic-softening constitutive law in the tension zone, and an elasticplastic
law in the compression zone. The chosen element for the steel
bar can represent exactly the shape of a ribbed bar.
An alternative treatment to model the interfacial zone is to
assume a negligible thickness of the interface layer, and thus transfer the bond problem into a category of contact issue [23]. A useful review was presented by Keuser and Mehlhorn [38] in respect
of this type of work, in which the normal stress between the steel
bar and the concrete, and the bondslip behaviour, was modelled
by using a double spring with one movement in the longitudinal
axis and the second in the perpendicular direction. The spring does
not have dimension and the relevant stiffness is calculated based
on the bondslip characteristics. The bond strength is a structural
behaviour rather than just a material property, and hence Darwin
et al. [39] proposed a full scale reinforced concrete model and simulated the interface layer by using a 3D interface link, which acts as
a contact-slip element.
From the 1980s onward, varieties of new types of FE element
emerged and were applied successfully to simulate the bondslip
relationship [4048]. In this study, a 3D reinforcement concrete
model with a translator element and a contact surface appears to
be the best approach after a number of different trials. The translator element is a special type of FE element that has been built in
the ABAQUS programme [49]. It has two nodes, which can be
attached to two substrates (see Fig. 8a). Like other types of contact
elements, it can be assigned a force together with corresponding
relative displacement between these two nodes. It also can receive
slot constraints and align them in the local direction u1 as shown in

330 340 330 mm

50mm

4.3. Material properties

As stated in Section 3, the catenary action occurs at the postbond-failure stage, at which both the concrete and steel reinforcement have been unloaded. As a result, the stress in both materials
will be below the yield level, so elastic material properties were
employed.
One of the key challenges encountered in modelling is to dene
an appropriate and efcient bondstress relationship. The damage
initiation criteria and the damage evolution laws are also
important to simulate the degradation behaviour of the bondslip
relationship. To determine the non-linear property of the translator elements, the pullout test of the concrete block tests were used
to derive the force-slip relationship. The measured results from the
test were pullout force and overall displacement. It was assumed
that the stiffness for the translator along the embedment length
is uniform. According to the pullout test results, and using four
translators at the interface and with an interval of 100 mm along
the embedment length, the translator properties were dened as
shown in Fig. 11.
4.4. Analysis solution strategy
The translator element is only available in ABAQUS/Explicit, and
the contact condition and other discontinuous problems can be
readily formulated in the Explicit module. Hence it is used in this
study to perform a non-linear quasi-dynamic analysis.
4.5. Verication of models

50 mm joint

lb = 4, 6, 8, 10 m

Both concrete and steel were modelled by the 8-node solid element with reduced-integration. The model was discretised in such
a way that the mesh density varies at different locations where
stress distributions are different. Three locations have been chosen
to apply different mesh densities in the hollow core concrete slabs,
and these are the steelgrout interface zones within the embedment length and the middle of the block (see Fig. 9). A mesh size
convergence analysis was carried out to determine the optimal
meshing pattern. Table 2 presents three mesh trials with various
mesh sizes at the circumference of the steelgrout interface, along
the embedment length and the middle of the block. The results of
slip and tie-force were examined for the convergence check. Table 2
indicates that meshing trials B and C yield very close results and
hence trial B has been chosen for the following modelling work.
The boundary conditions applied in the concrete block and the
oor-to-oor joint models are displayed in Fig. 10(a and b). In the
former case, only one degree of freedom of two end nodes
remained free, i.e. the longitudinal movement for the left end node.
In the latter case, the right end node remains restrained but the
middle point has been allocated a symmetry boundary condition
as only the right-hand side half is included in the model.

lb = 4, 6, 8, 10 m

50mm

Longitudinal ties
Keyways

Side
supports

Fig. 7. Illustrative diagram of the full scale oor-to-oor joint tests.

The FE models were validated by comparing pullout and full

scale oor-to-oor joint tests undertaken by PCA [1]. The PCA
experimental study was performed on a wide variety of strands
embedded in the keyways of precast concrete slabs of different
geometry and material properties. In the present study, in order
to validate the FE modelling, two pullouts and two full scale
oor-to-oor joint test results were used (Tables 3 and 4). The pullout load versus pullout displacement for pull-out test specimens
CP1 and CP2 (Table 3) are obtained from the FE modelling. They

182

M. Tohidi et al. / Engineering Structures 76 (2014) 177186

u2

u2
b

u1

u3

u1
u3

(a) Translator element (ABAQUS 2006)

(b) Implementation of translator element

Fig. 8. FE modelling of steelconcrete interaction.

Middle of block
Embedment length

Embedment length

Steel bars in keyways

Circumference

(a) Mesh configuration of a typical

pullout model

(b) Close-up view of the mesh configurations

of slab

(c) Model of the floor -to-floor joint system

Fig. 9. Finite element mesh pattern.

2500

Table 2
Mesh properties for pullout and full scale models.

A
B
C

Number of element or mesh size

Circumference
at interface

Embedment
length (mm)

Middle of
block (mm)

Slip
ratio

Tie
force
ratio

8 elements
16 elements
32 elements

50
25
12.5

150
150
150

1
1.09
1.10

1
1.08
1.08

Force (kN)

Mesh
trial

12.7 mm, Ld = 1500 mm

9.5 mm, Ld = 1140 mm

2000

9.5 mm, Ld = 760 mm

1500
1000
500
0

50

100

150

Pullout Displacement (mm)

Fig. 11. Translator properties used in the modelling process.

(a) Model for the pullout tests of concrete blocks

(b) Model for full scale floor- to-floor joint system

Fig. 10. Boundary conditions.

are compared with the corresponding experimental studies as presented in Fig. 12. As can be seen from Fig. 12, both sets of results
agree extremely well in the entire loading range, which indicates

the accuracy of FE modelling to simulate the pullout behaviour

of strands in the grouted keyways.
The results of the tie force in the strands at the mid-span vs. the
central vertical deection from the full scale oor-to-oor joints
tests modelling for FT1 and FT2 (Table 4) are presented in
Fig. 13. The comparison in Fig. 13 reveals that the FE modelling
provides a good estimate in terms of both peak load and ascending
or descending phases. The slight discrepancy can be attributed to
the measuring errors from the full scale test procedure and inherent errors associated with the assumptions introduced in the modelling process.
In the experiment, the tie force has been calculated based on
the measured strain results. Strain gauges were attached to the
steel strands at two discrete points at the loaded end. The grout
in the middle joint gap can contribute to the stiffness of the system
prior to the crushing. This happened before ds  250 mm [1]. In the

183

M. Tohidi et al. / Engineering Structures 76 (2014) 177186

Table 3
Properties of pullout test specimens undertaken by PCA [1].
ID

Strand size / (mm)

Ld//

Embedment length (Ld)

Loading angle

CP 1
CP 2

9.5
12.7

120
120

1140
1500

8
0

Table 4
Slab details from oor-to-oor joint tests.
ID

Dimension (mm)

Strand diameter(mm)

Ld/db

FT1
FT2

150  1000  6300

150  1000  6300

9.5
9.5

152
110

40
20
0

50

100

80
40
0

150

PCA
FE

120

Pullout Displacement (mm)

(a)

50

100

150

9.5mm, Ld=1140 mm

(b)

12.7 mm, Ld=1500 mm

Fig. 12. Pullout loaddisplacement from PCA experimental and FE results.

PCA
FE

75
50
25
0

0.2

0.4

0.6

0.8

Vertical deflection (m)

(a) 9.5mm, Ld = 1440 mm

Tie Force (kN)

Tie force (N)

FE modelling, the contribution from this part of the grout was not
included in the model. This explains why the stiffness from the
tests was slightly higher than in the numerical modelling before
the central deection reached 250 mm (see Fig. 13). However, after
the grout crushing, FE and experimental results show a close
agreement.The developed model has been conrmed as being able
to capture a complete tie force vs. vertical deection history with
good accuracy for different bar sizes, embedment lengths, and slab
lengths. From both full-scale tests and FE modelling results, during
the descending section in the tie force vs. deection curves, ties
undergo stable pullout damage until the pullout displacement
becomes excessive. According to the experimental study, the safe
region to establish catenary action described by the deection/
span ration is 5% 6 ds =lb 6 15% [1]. Hence, based on Eq. (4), the
upper limit for the pullout displacement can be dened as
dl/lb = 0.56%. This limit of the pullout displacement can be further
used to derive the corresponding limit of the tie force by using
tie force versus vertical deection graphs, e.g. Fig. 13. These limits
are included in dening the translators properties as the failure
criteria option.
The FE modelling also provides opportunities to investigate the
bond behaviour along the entire embedment length of the tie.
Fig. 14 shows the pullout displacement along the strand for the
CP2 specimen. Fig. 14(a) presents the result curves at two pullout
force levels before reaching the ultimate value, i.e. the prebond-failure stage. The rst curve corresponds to the case when
the pullout force is only half of the ultimate bond load, i.e.
P = 0.5Pmax, and no debonding along the entire strand occurs. The
100

5. Parametric study

Pullout Displacement (mm)

100

PCA
FE

75
50
25
0

0.2

0.4

0.6

0.8

Vertical Deflection (m)

(b) 9.5 mm, Ld = 1050 mm

Fig. 13. Experimental and FE results of tie force-vertical deection of FT1 and FT2.

Based on full scale experimental studies, the PCA [1] suggested

that the embedment length, bar size, concrete strength, slab
length, the number of keyways and the surface load had major
effects on the ductility behaviour of the oor-to-oor joint system,
and hence these should be considered as main design variables. To
keep the parametric study manageable, the following study is
limited to the oor-to-oor joint system subjected to a uniformly
distributed surface load only, as this type of loading always occurs
upon removal of the underlying wall supports caused by an accidental load such as explosion. The following geometrical variables
will be considered in the parametric study: the slab length lb, tie
spacing bp, embedment length ld, and bar size /. In addition, the
translator properties that are related to the strength of concrete
and interfacial properties between steel and concrete are also identied as a key inuencing variable.
A two-span slab system is modelled for four slab lengths of
4.075 m, 6.375 m, 8.075 m, and 10.075 m, respectively. The diameter of the strands is 9.5 or 12.5 mm. The embedment length of the
strand in the grouted keyways is considered to be 0.76 m, 1.14 m
or 1.5 m. The above gures are chosen with a view to resembling
PCAs experimental data.
6. FEA results
To show the inuence of the bar size and the embedment length
in establishing the full catenary action, the oor-to-oor assemblies with the above mentioned oor span lengths, bar sizes and
embedment lengths are analysed. Fig. 16 presents the tie force/
resistance ratio versus the vertical deection/span length ratio
for 9.5 mm and 12.7 mm strands and four different slab spans,
20

P = 0.9 Pmax
P = 0.5Pmax

15
Partial debonded

10
5
0

Fully bonded

250 500 750 1000 1250 1500

Pullout Displacement (mm)

60

Pullout Displacement (mm)

Pullout Load (kN)

PCA
FE

Pullout Load (kN)

160

80

second curve is plotted when the pullout force is close to the ultimate load, i.e. P = 0.9Pmax. At this level, the strand section near the
loaded end is debonded while the rest remains unbonded.
Fig. 14(b) represents the post-bond-failure stage, i.e. the descending section in the pullout force against displacement curves in
Fig. 12. In this stage, the entire strand has debonded and the pullout force has dropped to 60% of the ultimate value. At this level, all
data points in the strand experience excessive pullout displacement with a modest variation.
Fig. 15 shows the shear stress along the strand under the same
pullout force levels as indicated in Fig. 14. If no debonding is found
along the entire strand, the shear stress distribution follows an
exponentially decreasing pattern as is predicated by any conventional linear elastic analysis [20]. Once a part of the strand is debonded, the shear stress in the debonded zone tends to be almost
uniform, while in the bonded zone it still follows a similar pattern
to the un-bonded case. In the post-bond failure stage, when the
pullout forces descend to 0.6Pmax, the shear stress shares a similar
trend to the pullout displacement but with a smaller degree of
variation.

90

P = 0.6Pmax

85
80

Completely debonded

75
70
65
60

250

500

750 1000 1250 1500

Distance from loaded end (mm)

Distance from loaded end (mm)

(a) Pre-bond-failure stage

(b) Post-bond-failure stage

Fig. 14. Pullout displacement over embedment length for CP2.

184

M. Tohidi et al. / Engineering Structures 76 (2014) 177186

1

Partially debonded

1.5
1.2

Pre-failure; P=0.5Pmax

Completly debonded

0.9

Partial-failure; P = 0.9Pmax
Post-failure; P = 0.6Pmax

0.6

Fully bonded

F/Fy

Shear Stress (MPa)

1.8

0.3
0

250

500

750

1000

1250

1500

Distance from loaded end (mm)

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

L = 4.075 m
L = 6.375 m
L = 8.075 m
L = 10.075 m

10

15

Fig. 15. Shear stress over embedment length for CP2.

7. Proposed design method

Based on the present FE analysis results, for each chosen value
of ds =lb in the safe region, e.g. 5% 6 ds =lb 6 15%, the maximum

F/Fy

(a)

25

30

35

1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

9.5 mm, Ld = 760 mm

L = 4.075 m
L = 6.375 m
L = 8.075 m
L = 10.075 m

10

20

30

40

s /lb

F/Fy

(b)
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

9.5 mm, Ld = 1140 mm

L = 6.375 m
L= 8.075 m
L = 10.075 m

10

15

20

25

30

s /lb

(c)

12.7 mm, Ld = 1500 mm

Fig. 16. Tie force/resistance ratio versus vertical deection/span length ratio for
different oor span lengths and tie specications.

140

Tie Force (kN)

respectively. The predened three embedded lengths render Ld//

80, 120 and 120. Fig. 16(a and b) show that although the embedment length for the 9.5 mm strand is increased by 50%, the maximum tie force shows only a 15% increase. Reducing the span will
reduce the ductility expressed in terms of the central deection/
span ratio. However, all of them have exceeded 15% in the ds/lb
ratio. Fig. 16(c) shows a similar trend for the 12.7 mm strand with
Ld// = 120. The shortest slab span (4.075 m) has been removed as it
is considered inappropriate for this strand size. It can be seen, from
the comparison of Fig. 16(b and c) on the maximum tie force to
yield force ratio of the strand, that for the same Ld// value, this
ratio remains almost constant. This indicates that the embedment
length/bar ratio inuences the maximum tie force, in spite of the
actual bar size and the slab span. However, it will decline with a
reduction in embedment as indicated by Fig. 16(a). The above
observation conrms that it is the pullout behaviour that governs
the tie forcedeection behaviour. However, the ultimate deection varies signicantly with the slab span. From the design perspective, the necessary condition of forming the catenary action
is the adequate ductility, i.e. ultimate deection, often dened in
terms of the deection to span length ratio. Therefore the ability
of strands to provide efcient pullout displacement must be
considered as a signicant factor. As a result, the load-vertical
deection relationship in the post-bonding-failure stage needs to
be considered in the tie design. In the TF method, however, it is
only the tie strength requirements that have to be met.
According to the condition for a safe catenary mechanism established by PCA, ds =lb 15%; the pullout displacement limit dl at each
end, can be calculated as dl =lb 0:56% by using Eq. (5). It shows
that for a greater span length, the catenary action requires more
pullout displacement. However, as pullout behaviour governs the
ductility behaviour of the system, it can be seen that in the descending stage, for a specic ds/lb value, the relative tie force demand is
reduced when the span length is increased. The same conclusion
can be obtained through the FE analysis results (e.g. in Fig. 16).
In addition, by using the limit ds/lb = 15%, the required tie force
for different bar sizes and embedment lengths is analysed and presented in Fig. 17 for different span lengths. Also presented in the
same gure is the required tie force according to the TF method.
It can be seen that the TF method requires the tie force to increase
with increasing span length, whereas FE results suggest the tie
force requirements reduce with increasing span length.
By following the TF method together with Eq. (5), it is suggested
that a high vertical deection is required, while the PCA experimental study clearly indicates that the oor-to-oor joint system
will stop enduring loads for ds =lb P 15% [1]. Fig. 17 shows clear
discrepancies between the TF method and FE analysis in terms of
the tie force requirement to prevent progressive collapse, and suggests that the current TF method is signicantly underestimated.

20

s /lb

12.7 mm, Ld=1.50m

120

9.5 mm, Ld=1.14 m

100

9.5 mm, Ld=0.76 m

80

TF Method

60
40
20
0

10

12

14

16

lb (m)
Fig. 17. Tie force requirements vs. span length for different strand size and
embedment length.

uniform load on the oor wbp that will be endured by the system
can be derived by using Eq. (3). Fig. 18(ac) show the applied load
versus ds =lb for various strands diameters, embedment lengths and
oor spans. To develop a general design method according to the
PCA [1] experimental study, the maximum ds =lb ratio for an effective catenary action is chosen as 15%. Fig. 19 presents a design
chart for the maximum applied load wbp, under which a oor joint
design conguration can establish a safe catenary action
mechanism and prevent progressive collapse.

M. Tohidi et al. / Engineering Structures 76 (2014) 177186

8

w bp (kN/m)

7
6

L=4.075 m

L=6.375 m

L=8.075 m

L=10.075 m

5
4
3
2
1
0

10

15

20

25

s/lb

(a)

9.5 mm, Ld=760 mm

8
L=4.075 m
L=8.075 m

w bp (kN/m)

L=6.375 m
L=10.075 m

6
5
4
3
2
1
0

10

15

20

25

s/lb

(b)

185

oor joints were veried by comparing the results with test data
carried out by the PCA. Parametric analysis reveals that the bond
behaviour of ties governs the oor joint behaviour in developing
the catenary mechanism, and hence the maximum tie force for different span lengths is identical if the tie congurations are the
same. Results also indicate that the embedment length to bar
diameters ratio poses a more signicant impact on the ultimate
bond load of a system. However, the central deection to span ratio
will vary with the span, and the limit of this ratio is often chosen
for the safe design of the progressive collapse of the oor slab. Furthermore, the results show that the maximum applied loads that
oors can carry decreases signicantly with increasing slab length
from the progressive collapse design perspective.
The novelty of this study is that, from a design perspective, it is
the ductility rather than the tie strength which should be considered in the progressive collapse design, and the oor joint deection in the absence of supporting walls to span ratio is often
used quantify the design. The discrepancies in the tie force
between the numerical and the codied specications suggest that
an underestimate based on the TF method may lead to an unsafe
design. Hence, an improved model based on the numerical results
has also been proposed to address this concern.

9.5 mm, L d=1140 mm

Acknowledgements

L=6.375 m
L=10.075 m

wbp (kN/m)

L=8.075 m

The authors would like to express their gratitude to Bison Manufacturing Ltd. for their generous support by sponsoring test samples in this project. They are grateful for the funding from Shanghai
Pujiang Program, P.R. China (13PJ1405200). They also would like to
thank the anonymous reviewers for their constructive comments
to improve the quality of the paper.

5
4
3
2
1
0

10

15

20

s/lb

(c)

12.7 mm, Ld=1500 mm

Fig. 18. Maximum load wbp vs. vertical deection/oor span ratio.

w bp (kN/m)

12.5 mm, Ld=1.5 m

9.5 mm, Ld=1.14 m

9.5 mm, Ld=0.76 m

4
3
2
1
0

lb (m)

10

12

14

Fig. 19. Design chart for the maximum applied load against progressive collapse.

8. Conclusion
The tie force (TF) method is one of the most widely used methods to design concrete structures against progressive collapse. Due
to a high degree of simplication, this method is easy to use when
compared to the FE method. However, the study reported in this
paper has suggested that the design based on the TF method will
turn out to be unsafe for a certain range of oor spans.
Numerical models were developed to reproduce laboratory
tests on the pullout and full scale oor joint system. The interfacial
behaviour between the steel and the grout was modelled by using
the translator elements built into the ABAQUS software. The bond
stress-slip relationship was established by using the pullout tests
as calibration examples. The modelling results of the full-scale

References
[1] Portland Cement Association (PCA). Design and construction of large-panel
concrete structures. reports 16; Supplement Reports A, B, C. US Department
of Housing and Development; 19751979.
[2] Yagust VI, Yankelevsky DZ. On potential progressive failure of large-panel
buildings. Struct Eng 2007;131(11):1591603.
[3] British Standard BS 8110-11. The structural use of concrete in building Part
1: Code of practice for design and construction. London, UK; 1997.
[4] Pekau O, Cui Y. Progressive collapse simulations of precast panel shear walls
during earthquakes. Comput Struct 2006;84:40012.
[5] Scanlon A, Kianoush MR. Behaviour of large panel precast coupled wall
systems subject to earthquake loading. PCI J 1988:12437.
[6] Fintel M, Firnkas S, Speyer IJ. Comments on design against progressive collapse
by Alexander Popoff. PCI J 1977:1169.
[7] Dusenberry D. Review of existing guidelines and provisions related to
progressive collapse. Workshop on prevention of progressive collapse,
National Institute of Building Sciences, Washington (DC), USA; 2002.
[8] Moore DB. The UK and European regulations for accidental actions. In: Proc.
workshop on prevention of progressive collapse. National Institute of Building
Sciences. Washington (DC), USA; 2002.
[9] Nair RS. Progressive collapse basics. Modern Steel Constr 2004;44(3):3744.
[10] Abruzzo J, Matta A, Panariello G. Study of mitigation strategies for progressive
collapse of a reinforced concrete commercial building. Perform Constructed
Facil 2006;20(4):38490.
[11] DoD. Design building to resist progressive collapse. Unied Facilities Criteria
(UFC-04-023-03), USA: Department of Defence; 2005.
[12] Merola R. Ductility and robustness of concrete structures under accident and
malicious load cases. PhD. Thesis. School of Civil Engineering, University of
Birmingham, UK; 2009.
[13] Li Y, Lu X, Guan H, Ye L. An improved tie force method for progressive collapse
resistance design of reinforced concrete frame structures. Eng Struct
2011;33:293142.
[14] Yi WJ, He Q, Xiao Y, Kunnath SK. Experimental study on progressive collapse
resistant behaviour of reinforced concrete frame structures. ACI Struct
2008;105(4):4339.
[15] Gerasimidis S, Bisbos CD, Baniotopoulos CC. A computational model for full or
partial damage of single or multiple adjacent columns in disproportionate
collapse analysis via linear programming. Struct Infrastruct Eng
2013;9(1):114.
[16] Tohidi M, Yang J, Baniotopoulos C. An improved tie force method for
progressive collapse resistance of precast concrete cross wall structures.
World Acad Sci Eng Technol Int J Civil Archit Sci Eng 2014;8(1).

186

M. Tohidi et al. / Engineering Structures 76 (2014) 177186

[17] DoD. Design building to resist progressive collapse. Unied Facilities Criteria
(UFC-04-023-03), USA: Department of Defence; 2013.
[18] Yan Y, Gerasimidis S, Deodatis G, Ettouney M. A study on the global loss of
stability progressive collapse mechanisms of steel moment frames. In:
Ellingwood, Frangopol, editors. Safety, reliability, risk and life-cycle
performance of structure & infrastructures Deodatis. London: Taylor &
Francis Group. ISBN 978-1-138-00086-5; 2013.
[19] Spyridaki A, Gerasimidis S, Deodatis G, Ettouney M. A new analytical method
on the comparison of progressive collapse mechanism of steel frames under
corner column removal. In: Ellingwood, Frangopol (Eds). Safety, reliability, risk
and life-cycle performance of structure & infrastructures Deodatis. London:
Taylor & Francis Group. ISBN 978-1-138-00086-5; 2013.
[20] Ettouney M, Smilowitz R, Tang M, Hapij A. Global system considerations for
rogressive collapse with extensions to other natural and man-made hazards.
ASCE J Perform Constr Facil 2006:40317.
[21] Ren FF, Yang ZJ, Chen JF, Chen AWW. An analytical analysis of the full-range
behaviour of grouted rockbolts based on a tri-linear bondslip model. Constr
Build Mater 2010;24:36170.
[22] Shen SL, Hou DW, Zhao JL, Horpibulsuk S, Yin ZY. Assessment of internal forces
for intermediate anchorage zone of post-tensioned concrete structure. Constr
Build Mater 2014;64:3708.
[23] CEB-FIP. State-of-the-art report on bond of reinforcement in concrete. Stateof-art report prepared by task group bond models (former CEB Task Group 2.5)
FIB -Fd. Int. du Bton, Lausanne, Switzerland; 2000. p. 197.
[24] Den UJ. Bond modelling of prestressing strand. ACI Special Publication. vol.
180; 1998. p. 145170.
[25] Abrishami HH, Mitchell D. Analysis of bond stress distributions in pullout
specimens. Struct Eng 1996;122:25561.
[26] Balazs GL. Transfer control of prestressing strand. PCI J 1992:6071.
[27] Daoud A, Maurel O, Laborderie C. 2D mesoscopic modelling of barconcrete
bond. Eng Struct 2013;49:696706.
[28] Hao Q, Wang Y, Ou J. Design recommendations for bond between GFRP/steel
wire composite rebars and concrete. Eng Struct 2008;30:323946.
[29] Mazzarolo E, Scotta R, Berto L, Saetta A. Long anchorage bondslip formulation
for modeling of RC elements and joints. Eng Struct 2012;34:33041.
[30] Ogura N, Bolander JE, Ichinose T. Analysis of bond splitting failure of deformed
bars within structural concrete. Eng Struct 2008;3:42835.
[31] Salema HM, El-Fouly AK, Tagel-Din HS. Toward an economic design of
reinforced concrete structures against progressive collapse. Eng Struct
2011;33:334150.
[32] Kwasniewski L. Nonlinear dynamic simulations of progressive collapse for a
multistory building. Eng Struct 2010;32:122335.

[33] Shi Y, Li ZX, Hao H. A new method for progressive collapse analysis of RC
frames under blast loading. Eng Struct 2010;32:1691703.
[34] Valipour HR, Foster SJR. Finite element modelling of reinforced concrete
framed structures including catenary action. Comput Struct 2010;88:52938.
[35] BSI. BS EN1991-1-7: Action on Structures Part 17: General actionsaccidental action. London, UK; 2006.
[36] Bresler B, Bertero VV. Behaviour of reinforced concrete under repeated load.
Struct Eng 1968;94(6):157690.
[37] Reinhardt HW, Blaauwendraad J, Vos E. Prediction of bond between steel and
concrete by numerical analysis. RLLEM Mater Struct 1984;17(100):31120.
[38] Keuser M, Mehlhorn G. Finite element model for bond problem. J Struct Eng
1987;113(10):216073.
[39] Darwin D, McCabe SL, Brown CJ. Fracture analysis of steelconcrete bond. In:
Bittnar, Z. Jirasek M. and Mazars J, editors. Proc. Europe-US Workshop on
fracture and damage in quasi brittle structure. Pregue, Czech Repulic; 1994. p.
557566.
[40] Al-Zuhairi AHA, Al-Fatlawi WDS. Numerical prediction of bondslip behaviour
in simple pull-out concrete specimens. J Eng 2013;19:112.
[41] Sung HL, Sung CC, Bohwan O. Numerical analysis of pull-out behaviour of
headed bar in concrete. Architectural Technology Research Team, DAEWOO
Institute of Construction, Technology, Suwon 440-210 Korea; 2012.
[42] Valente M. Bond strength at the steel-concrete interface: experimental tests
and numerical analyse. In: 11th International conference on concrete
engineering and technology 2012 (CONCET2012), 1213th June 2012,
Putrajaya, Malaysia; 2012.
[43] Khalfallah S, Ouchenane M. Prediction of bond between steel and concrete by
numerical analysis. Open Civil Eng J 2008;2:18.
[44] Harajli MH. Numerical bond analysis using experimentally derived local bond
laws: a powerful method for evaluating the bond strength of steel bars. J Struct
Eng 2007;133:695705.
[45] Moreno C, Bastos AS. Experimental and numerical evaluation of bond
properties between reinforcement and concrete, Civil Engineering
Department, University of Porto, Porto, Portugal; 2006. p. 2426.
[46] Nardin SD, Filho FMA, Filho JO, Haach VG, El Debs ALHC. Non-linear analysis of
the bond strength behaviour on the steel-concrete interface by numerical
models and pull-out tests. New York: ASCE, Structure Congress; 2005.
[47] Girard C, Bastien J. Finite-element bondslip model for concrete columns
under cyclic loads. J Struct Eng 2002;128(12):150210.
[48] Yogarajah I, Yeo KC. Finite element modelling of pull-out tests with load and
strain measurements. Department of Civil Engineering, University of
Strathclyde, Glasgow, UK; 1994.
[49] Hibbit K. Sorensen, ABAQUS: Users Manual, Version 6.7; 2007.