Vous êtes sur la page 1sur 14

952

Finite element studies of reinforced concrete


slab – edge column connections with openings
Hong Guan and Maria Anna Polak

Abstract: An extended nonlinear layered finite element method (LFEM) is used to investigate the influence of openings
and shear stud reinforcement (SSR) on the behaviour of reinforced concrete slab – edge column connections. In all, ten
large-scale slab – edge column connections tested previously are analyzed. The laboratory test variables were the size
and location of the openings in the vicinity of an edge column and the existence of SSR. The numerical results of the
load–deflection response, the ultimate strength, and the crack patterns are compared with the experimental findings and
good agreement is achieved. A numerical study on two connections is also carried out to determine the influence of the
locations of opening in slab – edge column connections with SSR. Discussion on code provisions for slabs with openings
is provided. The comparative and numerical studies confirm the accuracy, reliability, and effectiveness of the LFEM in the
analysis of slab – edge column connections with both openings and SSR.
Key words: slab–column connection, punching shear, opening, shear stud reinforcement, finite element analysis.

Résumé : Une méthode élargie d’éléments finis multidimensionnelle (« LFEM ») non linéaire est utilisée pour étudier
l’influence des ouvertures et des renforcements des goujons (« SSR ») sur le comportement des connexions dalle –
colonne de bordure en béton armé. Un total de dix connexions dalle – colonne de bordure à grande échelle testées
précédemment ont été analysées. Les variables des essais en laboratoire étaient la dimension et l’emplacement des
ouvertures aux environs d’une colonne de bordure et la présence de « SSR ». Les résultats numériques de la réaction
charge-déformation, la résistance à la rupture, ainsi que les patrons de fissures sont comparés aux résultats expérimentaux;
une bonne corrélation a été obtenue. Une étude numérique de deux connexions a également été réalisée pour déterminer
l’influence des emplacements de l’ouverture dans les connexions dalle – colonne de bordure avec « SSR ». Nous discutons
aussi des dispositions des codes pour les dalles avec ouvertures. Les études comparatives et numériques confirment la
précision, la fiabilité et l’efficacité de la « LFEM » dans l’analyse des connexions dalle – colonne de bordure comportant
des ouvertures et des « SSR ».
Mots-clés : connexion dalle–colonne, cisaille–poinçonneuse, ouverture, renforcement des goujons en cisaillement, analyse
par éléments finis.

1. Introduction In the flat-plate floor system, it is often necessary to con-


struct openings in the vicinity of columns. Such openings are
A reinforced concrete flat-plate floor system is widely used in required for sanitary reasons, ventilation, heating, air condition-
various types of building structures including condominiums, ing, and electrical ducts. The area within the critical perimeter
parking garages, and office buildings. This floor system is ad- around the column experiences high transverse shear stresses
vantageous in terms of simpler formwork, shorter construction resulting from concentrated forces and unbalanced moments.
period, reduced storey height, flexible room arrangement, more The existence of an opening reduces the volume of concrete
headroom, better air circulation, and better light penetration. In that can resist such transverse shear. This further weakens the
spite of the advantages, the system has its problems, one of slab–column connections that are prone to localized and brittle
which is the possibility of punching shear failure in the vicinity punching shear failure. To strengthen such connections and to
of the slab–column connections. For corner and edge columns provide adequate ductility, properly placed and anchored shear
in particular, it is often the focus of studies due to the unbal- stud reinforcement (SSR) is often utilized.
anced loading condition as well as the unsymmetrical stress Research on flat-plate concrete slabs with openings is lim-
distribution in these locations. ited. Moe (1961) studied the shear strength of concrete slabs
with holes near the column faces. Hognestad et al. (1964) and
Received 16 July 2006. Accepted 6 December 2006. Published on Mowrer and Vanderbilt (1967) carried out further laboratory
the NRC Research Press Web site at cjce.nrc.ca/ on 28 August 2007.
tests on slab – interior column connections with openings with
H. Guan.1 Griffith School of Engineering, Griffith University Gold particular emphasis on lightweight aggregate concrete slabs,
Coast Campus, Queensland 4222, Australia. and Roll et al. (1971), on perforated concrete slabs. Not until
M.A. Polak. Department of Civil Engineering, University of Water- the mid 1990s did the study on the punching shear behaviour of
loo, 200 University Avenue West, Waterloo, ON N2L 3G1, Canada. slab–column connections with openings regain researchers’ at-
Written discussion of this article is welcomed and will be received tention. Various laboratory investigations have been conducted
by the Editor until 31 December 2007. including openings in the vicinity of square edge columns by El-
Salakawy et al. (1999, 2000) and openings adjacent to square
1
Corresponding author (e-mail: h.guan@griffith.edu.au). and rectangular interior columns by Gomes and Silva (2003)

Can. J. Civ. Eng. 34: 952–965 (2007) doi: 10.1139/L06-169 © 2007 NRC Canada
Guan and Polak 953

and Irawan et al. (2003). The effects of shear stud reinforce- Fig. 1. Typical dimensions, loading and boundary conditions
ment were also examined in some of these studies. Theoretical of test specimens (modified from El-Salakawy et al. 1999):
research in the subject area has also been reported by Gomes (a) elevation; (b) plan view. All dimensions in millimetres.
and Silva (2003).
Laboratory tests are labour-intensive, time-consuming, and
costly. Yet, various empirical and code methods, based heavily
on model test results, inevitably involve gross approximations
that are not always reliable and, by nature, their scope of ap-
plications is limited. Therefore, nonlinear finite element (FE)
analysis is a rational and needed method by which structures can
be analyzed to progressive failure. For slab–column structures,
the FE analysis can be undertaken using either detailed three-
dimensional elements (Staller 2001; Ožbolt and Vocke 2001) or
shell elements specifically formulated for reinforced concrete.
The first approach requires substantial time in numerical model
development and computational effort in calculating the struc- The test variables were the size and location of openings and
ture’s response under loads. Its strength and applicability lies the existence of SSR. Two additional specimens were analyzed
in the ability for analyzing connections and small portions of to study the effect of the opening location for slabs with SSR.
real structures, where it can produce detailed information about The accuracy and effectiveness of the LFEM are demonstrated
initialization, location, and progression of cracking, straining, by comparing the analytical solutions with the experimental
and damage. The second approach, based on layered shell ele- load–deflection responses, ultimate strengths, and crack pat-
ments, is by its nature less detailed because it is usually based terns. Discussions regarding the effect of openings and SSR on
on a smeared cracking approach and considers reinforcement as slab–edge connections strength and stiffness are presented.Also
either a layer within the concrete material or as additional stiff- presented in the paper are comparisons between experimental
ness added to the concrete itself. The strength of this approach and theoretical results obtained using formulas in Australian,
is that, in comparison with the three-dimensional analysis, it al- American, and Canadian concrete design codes (SAA 2001;
lows for a much faster creation of a computational model (finite ACI 2005; CSA 2004).
element mesh with load and boundary conditions) and allows
the use of a much smaller number of elements. Thus, it is evi- 2. Experimental investigation
dent that the nonlinear finite element shell analysis of concrete
slab structures is becoming increasingly important since, with A series of laboratory tests was undertaken by El-Salakawy
the proper pre-processing and post-processing interface, it can et al. (1999, 2000) on the punching shear behaviour of rein-
be utilized in the practical concrete building design situation. forced concrete slab – edge column connections with open-
Few layered finite element approaches have been proposed to ings and the effectiveness of shear stud reinforcement (SSR)
model punching shear in flat plates, slabs, and slab–column in resisting shear stresses around the connection. The large-
connections (Harmon and Ni 1989; Rericha 1991; Guan and scale specimens represented edge columns connected to a slab
Loo 1997; Polak 1998a, 1998b), where shell elements are used bounded by the theoretical contraflexure lines around the col-
encompassing concrete and smeared steel layers. Polak (1998a, umn. Typical dimensions of the specimens together with the
1998b) used the nonlinear elastic secant material model formu- loading and boundary conditions are presented in Fig. 1. The
lation based on the modified compression field theory (Vec- specimens were loaded through the column stub. They were
chio and Collins 1986), while Guan and Loo (1997) adopted simply supported along the three sides with the slab edge be-
the nonlinear elastic-plastic-fracturing tangential approach for ing free. Corners of the slabs were restrained from lifting. Each
modelling concrete. Regardless of the material modelling ap- specimen was subjected to a vertical shear force, V , through the
proach adopted within the shell element formulation, it has been column stub and two horizontal forces, H , applied simultane-
shown that shell finite elements can be successfully used in the ously to the column ends that produced the unbalanced moment
analyses of slabs to detect both flexural and punching-shear M. The moment to shear ratio, M/V , was kept constant at 0.3 m.
failure modes. The presented test specimens were divided into two series. Each
The purpose of the work presented is to report the perfor- series consisted of one specimen without an opening and the re-
mance of layered degenerate shell elements incorporated into a maining specimens with openings. The opening layout of all the
layered finite element method (LFEM) by Guan and Loo (1997), specimens is presented in Fig. 2. Series I, including specimens
in the analysis of slab–column edge connections with and with- XXX, SE0, CF0, SF0, SF1, and SF2 without shear reinforce-
out openings next to the column area. The paper presents analy- ment (Fig. 2a), were tested and analyzed by LFEM. The first
ses of slabs with and without shear reinforcements. Successful four specimens (Fig. 2b) in series II were also tested and ana-
finite element modelling of edge connections with openings is lyzed by LFEM. They are XXX-R, SE0-R, CF0-R, and SF0-R,
of practical importance for the global analysis of continuous being the counterparts of the first four specimens in series I,
flat plate – column structural systems. The finite element for- where R refers to the existence of SSR. XXX and XXX-R were
mulation is presented first. Twelve reinforced concrete slab – both control specimens without an opening.
edge column connections are analyzed to determine the influ- Also shown in Fig. 2b are two specimens, SF1-R and SF2-R,
ence of openings and shear stud reinforcement (SSR) on the which were not tested but analyzed using LFEM. These were
punching-shear behaviour of such connections. Ten of the ana- used for the numerical study of the influence of the locations of
lyzed specimens were tested by El-Salakawy et al. (1999, 2000). the opening in slab – edge column connections with SSR. The

© 2007 NRC Canada


954 Can. J. Civ. Eng. Vol. 34, 2007

Fig. 2. Opening layout of test specimens: (a) specimens without SSR; (b) specimens with SSR. (SF1-R and SF2-R were not tested, only
used for numerical analyses.) All dimensions in millimetres.

Table 1. Details of test data.


Specimen Compressive Tensile strength, Size of opening Location of D/d
strength, fc (MPa) ft (MPa) (mm × mm) opening
Series I
XXX 33.0 3.38 — — —
SE0 30.5 2.86 150 × 150 Edge 0
CF0 31.5 2.85 250 × 250 Front 0
SF0 32.5 3.49 150 × 150 Front 0
SF1 33.0 3.01 150 × 150 Front 1
SF2 30.0 3.04 150 × 150 Front 2
Series II
XXX-R 32.0 2.81 — — —
SE0-R 30.0 3.04 150 × 150 Edge 0
CF0-R 32.5 3.49 250 × 250 Front 0
SF0-R 32.0 3.47 150 × 150 Front 0

geometric configuration, boundary condition, material setting, Fig. 3. Arrangement of stud rails in SF0-R (modified from
and loading type of SF1-R and SF2-R are identical to those of El-Salakawy et al. (2000)). All dimensions in millimetres.
SF1 and SF2, respectively.
All the specimens were cast using normal density concrete of
approximately 30 MPa compressive strength (fc ). The concrete
tensile strength (ft ) was obtained from the split cylinder tests.
The material properties of each specimen are summarized in
Table 1. Also included in Table 1 are the size and location of
the openings. All the specimens were reinforced by tension
and compression steel meshes, having an average reinforcement
ratio of 0.0075 and 0.0045, respectively, for both directions of
steel in tension and compression zones. The columns were over-
reinforced. The shear studs were positioned along six strips
around the column, two at each side, as shown in Fig. 3 for
specimen SF0-R. Detailed reinforcement arrangement and the
test setup information can be found elsewhere (El-Salakawy et
al. 1999, 2000).

3. Layered finite element modelling


strategy
3.1. Layered element approach
Punching shear failure at slab–column connections of flat
plates has a distinct three-dimensional feature. The LFEM (Guan

© 2007 NRC Canada


Guan and Polak 955

Fig. 4. Layered finite element.

and Loo 1997; Loo and Guan 1997) was developed to encom- Fig. 5. Concrete constitutive model: (a) one-dimensional
pass three-dimensional (in-plane and out-of-plane) stress com- representation for concrete in compression and tension; (b)
ponents in its finite element formulation, thereby being capable tension stiffening effect.
of analyzing both flexural and transverse shear cracking, up to
failure.
The Mindlin plate hypothesis is used to derive the elements
from three-dimensional elasticity. Two primary assumptions are
adopted: (1) lines normal to midsurface before deformation re-
main straight but not necessarily normal after deformations,
and (2) normal stress in the transverse direction is equal to zero
(σz = 0). These assumptions allow us to formulate elements
where each nodal point located on the midreference plane has
five degrees of freedom viz. the in-plane displacements, u and
v; transverse displacement w; and two independent bending
rotations about the x and y axes, i.e., θy and θx , respectively.
The element has five nonzero stresses (σx , σy , τxy , τxz , and
τyz ) and five independent strains (εx , εy , γxy , γxz , and γyz ).
The transverse shear deformations are taken into account in the
deformation field because the rotations and displacements are
uncoupled. The transverse strain εz is found from equilibrium
and the condition of σz being zero.
The construction of a typical layered element of an eight-
node degenerate shell element type is shown in Fig. 4. In each
element, the concrete is subdivided into a number of layers of
varying but uniform thickness. The layer thickness gradually
s = ρ (f )/(ε ), where ρ is the average transverse re-
is D11
reduces from the midplane of the element towards the surfaces st st st st
to provide a more detailed account of the extensive cracking inforcement ratio for the element, fst is its strength, and εst is
near the bottom and top surfaces. The layers are fully bonded the strain in the direction of the reinforcement.
together. Each layer has constant material properties over the Each layer contains Gauss points on its midsurface. The
layer thickness. However, different layers may assume different stresses within each layer are computed at these points and
material properties. The bottom and top layers of flexural re- are assumed to be constant over the layer thickness. These lead
inforcing bars are represented by smeared layers of equivalent to a stepwise approximation of the stress distribution over the
thickness having directions in accordance with those of steel thickness of the element, as shown in Fig. 4. In the presented
bars. The transverse reinforcement is modelled as a property FE approach, the material state at any Gauss point located at
of a concrete layer associated with the normal strain (εz ) of the midsurface of a layer can be elastic, plastic, or fractured
concrete in the transverse direction. Through the perfect bond, according to the loading history. To account for the mechani-
it is assumed that transverse reinforcement undergoes strain cal change of the materials throughout the incremental loading
corresponding to its directions (for typical reinforcement nor- process, cracking and nonlinear material response are traced
mal to the plate surface, the strain is equal to (εz )). Transverse layer by layer.
reinforcement is smeared through the element. Then its contri-
bution to the appropriate diagonal term of the material stiffness 3.2. Material constitutive relations
matrix of a reinforced concrete element, in the coordinate sys- The concrete failure is identified as a result of either tension
tems corresponding to the transverse reinforcement direction, cracking or plastic yielding (crushing). A three-dimensional

© 2007 NRC Canada


956 Can. J. Civ. Eng. Vol. 34, 2007

stress state is considered in the model. According to the con- crete in both tension and compression. The constitutive equa-
ventional plane stress assumption based on the Mindlin plate tion for isotropic material, in the material coordinate system
bending theory, the constraint of the stress component normal (x  y  z ), can be expressed as
to the midreference plane is imposed (i.e., σz = 0) at the global,
element level, as is required for degeneration from the three- [1] d{σ } = Dc d{ε}
dimensional to the shell element. Figure 5a depicts the one-
dimensional representation of the constitutive model for con- where Dc is given by

 
K + 43 G K − 23 G K − 23 G 0 0 0
 K + 43 G K − 23 G 0 
 0 0 
 
[2] Dc = 
 K + 43 G 0 0 0 

 G 0 0 
 
Symm G 0
G

is the material matrix in which K (= [E/3(1 − 2ν)]) and G (= [E/2(1 + ν)]) are the bulk and shear moduli, respectively. E and
ν are, respectively, the modulus of elasticity and Poisson’s ratio.
In eq. [1], the stress and strain components are {σ } = {σx , σy , σz , τxy , τxz , τyz } and {ε} = {εx , εy , εz , γxy , γxz , γyz }, respectively.
Due to the assumption of σz = 0, each element has five nonzero stresses and five independent strains. The strain component εz
can be determined from equilibrium with imposed σz = 0 and is a function of the five independent strains. With all the strain
components, the principal strains and their corresponding directions can be calculated, based on which the principal stresses and
directions can be determined, using appropriate constitutive models for concrete as described below.
An elastic brittle fracture behaviour is assumed for concrete in tension. Cracks are assumed to form in the plane perpendicular to
the direction of maximum principal tensile stress as soon as this stress reaches the specified concrete tensile strength ft . Cracked
concrete is treated as an orthotropic material using a smeared crack approach and the tension cut-off representation is utilized. The
constitutive equation for cracked concrete is given as

[3] d{σ } = Dcr d{ε}

where the material matrix Dcr is expressed as


 
Ei 0 √ 0 0 0 0
 E ν E E /(1 − ν − 2ν 2) 0 0 0 
 2 2 3 
 E3 0 0 0 
[4] Dcr =  
 Gc12 0 0 
 Symm Gc13 0 
Gc23
or
 
Ei 0 0 0 0 0
 Ei 0 0 0 0 
 
 E3 0 0 0 
[5] Dcr = 


Gc12 0 0


  
 Symm Gc13 0 

Gc23

for concrete cracked in one and two directions, respectively. In aggregate interlock, the shear stiffness deterioration in terms of
  
eqs. [4] and [5], E2 and E3 are the elastic moduli of concrete the reduced shear moduli Gc12 , Gc13 , Gc23 and Gc12 , Gc13 , Gc23 are
in directions 2 and 3, respectively. The elastic modulus in di- also taken into consideration after concrete is cracked.
rection 1, E1 , is replaced by the fictitious modulus of elasticity For cracked concrete, the amount of shear carried by aggre-
Ei when the concrete cracks in direction 1 (eq. [4]); and both gate interlock decreases with the crack width, which can be
E1 and E2 are replaced by Ei when the concrete cracks in both represented by the principal tensile strain. According to Hinton
directions 1 and 2 (eq. [5]). This is the so-called tension stiff- and Owen (1984), the shear moduli can be assumed to degrade
ening effect as a result of the bond mechanisms. Because of the linearly with the increase in principal tensile strains. For con-

© 2007 NRC Canada


Guan and Polak 957

crete cracked in the direction 1 only, Fig. 6. Steel constitutive model.


  ε1
0.25G 1 − for ε1 < 0.004
[6] Gc12 = Gc13 = 0.004
0 otherwise
5
Gc23 = G
6
where G is the uncracked shear modulus and ε1 is the current
principal tensile strain in direction 1. For concrete cracked in
both directions 1 and 2,
   
c 0.5Gc13 for Gc23 > Gc13
[7] G12 = 
0.5Gc23 otherwise where Dep is given by
  ε1
 0.25G 1 − for ε1 < 0.004 Dc aa T DTc
Gc13 = 0.004 [11] Dep = Dc −
0 otherwise H  + a T DTc a
  ε2
 0.25G 1 − for ε2 < 0.004 and
Gc23 = 0.004
0 otherwise [12] Dc = TTε Dc Tε
where ε1 and ε2 are the principal tensile strains in directions 1 if the principal axes do not coincide with the local reference
and 2, respectively. axes x, y, and z; a is the flow vector; and H  is the hardening
The tension stiffening effect, used in Hinton and Owen’s parameter associated with the expansion of the yield surface.
study (1984), is adopted herein, and assumes a gradual release When the compression type of failure transpires in concrete
of the concrete stress component normal to the cracked plane. (when the ultimate strain εu is reached), some but not all strength
This is illustrated in Fig. 5b as a descending branch of the stress– and rigidity of the material is lost. This is represented by the
strain curve. Also depicted in Fig. 5b is the process of loading effect of bulk modulus in the concrete material matrix. Or,
and unloading of cracked concrete. This process is assumed to  
follow the linear behaviour with the fictitious elasticity modulus K K K 0 0 0
Ei given as:  K K 0 0 0 
 
 K 0 0 0 
αt ft (1 − εi /εm ) Dcrs = 
0 0 0 
[13]
[8] Ei = , εt ≤ εi ≤ εm  
εi  Symm 0 0 
where αt and εm are the tension stiffening parameters as shown 0
in Fig. 5b, and εi is the maximum value reached by the ten-
Numerical modelling of either cracking or crushing of concrete
sile strain at the point currently under consideration. Certain
involves the modification of material stiffness and partial or full
assumptions exist in the model; in particular, the tension stiff-
release of the appropriate stresses in the fractured elements.
ening parameters that are determined based on the experimental
The reinforcing steel is assumed to be uniaxial elastic-plastic
measurements of concrete and the arrangement of steel. The ef-
material. The reinforcing bars at a given level in an element are
fect of variations in steel arrangement can be captured by having
modelled as a smeared steel layer of equivalent thickness. For
refined meshes for dense areas of reinforcing steels. The influ-
a steel layer, the constitutive equation is given as
ence of different tension stiffening parameters in eq. [8] on the
predictions of punching shear behaviour has been examined by
Polak (2005). In that study, αt = 0.6 and εm = 0.002 are sug- [14] d{σ } = Ds d{ε}
gested as the values that produce the best predictions and thus where the material matrix in the material coordinate system
they are also used in the presented work.
 
The material matrices for cracked concrete Dcr as given in ρs Es 0 0 0 0 0
eqs. [4] and [5] must be transformed from the principal axes  0 0 0 0 0 
 
to the local xyz coordinate system through the transformation  0 0 0 0 
Ds = 
0 0 0 
[15]
matrix Tε for strain components. Or,  
 Symm 0 0 
[9] Dcr = TTε Dcr Tε 0
The compressive behaviour of concrete is modelled using the in which ρs (s = x  ,y  ) is the reinforcement ratio (in the in-
strain-hardening plasticity approach that determines the bound- plane x  - and y  -directions) in the steel layer and Es is the
aries of elastic and plastic regions (when the concrete compres- Young’s modulus of steel. A trilinear idealisation of reinforcing
sive stress reaches 0.3fc where fc is the compressive strength steel is shown in Fig. 6. When the stress in the direction of
of concrete) and the progress of damage growth in the plastic the reinforcement exceeds 80% of the yield strength fy , Es is
zone. The elasto-plastic constitutive equation is expressed as replaced by Es1 ; and if fy is reached, the steel is assumed to have
[10] d{σ } = Dep d{ε} yielded and Es2 is then assumed. The values of Es1 and Es2 are

© 2007 NRC Canada


958 Can. J. Civ. Eng. Vol. 34, 2007

taken based on the experimental stress–strain measurement of Fig. 7. Finite element idealisation: (a) SF0-R (symmetrical half);
the steel bars. If the directions of the steel bars do not coincide (b) SE0-R.
with the x-axis, Ds must also be transformed into the local
coordinate system as

[16] Ds = TTε Ds Tε

The total material matrix D containing the contributions of con-


crete and steel can be determined for each element and the
stiffness matrix for the corresponding element can be evaluated
using the Gaussian integration technique where the selective
integration rule is adopted. D is expressed as

nc

ns
[17] D= Dc + Ds
i=1 i=1

where Dc can be taken as Dcr (eq. [9]), Dep (eq. [11]), or Dcrs
(eq. [13]) depending on the actual stress condition; nc and ns ,
respectively, are the total number of concrete and steel layers
(see Fig. 4).
In eq. [17], D is of size 6×6. However, due to the conventional
plane stress assumption, σz = 0, in the constitutive equations, D
must be condensed into 5 × 5 in size which contains a diagonal
term in the form of
[18] D(3, 3) = Dc (3, 3) + ρz Es
This equation implies that the effect of the out-of-plane (trans-
verse) reinforcement (in the z -direction) can be included by
adding its contributions to the concrete material matrix that
corresponds to the normal strain in the transverse direction.
The shear studs are included in the model by treating them as
transverse reinforcement where ρz equals ρst and Es equals
fst /εst .
Finally, the global stiffness matrix is assembled using the
standard procedure. The Newton–Raphson method, an incre-
mental and iterative procedure, is used to obtain the nonlinear
solution due to both material and geometric nonlinearities in
slab–column connection problems.

4. Numerical investigation
Ten test specimens are analyzed using the LFEM and the
numerical predictions are compared with the experimental re-
sults of El-Salakawy et al. (1999, 2000). The vertical load V
is applied through the centre of the column and the two hori-
zontal forces, H , are applied to the column centre as a bending
moment (see Fig. 1). The column region is modelled using the
same 8-node shell elements with much higher stiffness to ac-
count for the longitudinal steels and closed ties. Each element is
subdivided into eight concrete layers of varying thickness. The To better simulate the concentrated effect of the shear studs,
normalized layer thicknesses are taken as 0.1, 0.2, 0.3, and 0.4 the element meshes at the stud locations are made relatively
from the surface to the midplane. Such thickness combination narrow (same as the width of stud rail, see Figs. 3 and 7) and
has shown to yield a satisfactory solution (Guan and Loo 2003). all the studs on the same rail are located at the centre line of
The top and bottom flexural reinforcing steels are smeared into the elements covering the stud rail (for example, elements 43
four layers of equivalent thickness. By virtue of symmetry, the to 49 in Fig. 7a). The corresponding equivalent (smearing) stud
control specimens and those having the opening in front of the ratios, ρst , for these elements are calculated. The presence of
column are analyzed based on half of the respective models. the anchor head and the rail, located near the top and bottom
For specimens SE0 and SE0-R, a full model is analyzed. The layers, are also considered in calculating corresponding ρst .
finite element idealisation for specimens SF0-R and SE0-R is Hence, different layers may have different ρst depending on the
presented in Fig. 7. relative position of the stud rail in the transverse direction.

© 2007 NRC Canada


Guan and Polak 959

Fig. 8. Deformed shape: (a) SF2 (symmetrical half); (b) SE0-R. shear strength by an average of 41.5% for series I and 31.5% for
series II, respectively. The predictions by the Canadian standard
include the difference in assumed concrete shear strength be-
tween the slabs with the stirrups and the mechanically anchored
shear studs.
In comparing the ultimate loads for specimens with and with-
out SSR, it is evident from the experimental observations that
the strengths of the specimens with SSR were higher than their
counterparts without SSR. This is also true for the LFEM and
code predictions. It is further noted that the existence of an
opening in the specimen reduces its ultimate strength. Such re-
duction is less in specimens SE0-R and SE0 (where the opening
is located next to the edge face of the column) than specimens
CF0-R, CF0, SF0-R, and SF0 (where the opening is immedi-
ately adjacent to the front column face and in the direction of
unbalanced moment). Also, there is a significant reduction in
ultimate strength in specimen CF0 where the opening is of the
size of the column.

4.3. Crack pattern


In Fig. 11, the predicted and observed crack patterns of spec-
imens SF0 and SE0-R are compared. Note that the continuous
lines in Figs. 11a1 and 11a2 indicate only the crack direction
at specific Gauss points and they do not offer information on
4.1. Deformed shape and load–deflection characteristics crack length and width. The LFEM appears to predict satisfac-
As a typical example, the deformed shape of specimens SF2 torily the crack pattern at the tensile surface of both specimens.
and SE0-R at ultimate load is depicted in Fig. 8. As can be seen, In the laboratory test, the formation of flexural cracks on the
the maximum vertical deflection occurs at the column stub. This tensile surface was initiated at the vertical load of 49–55 kN.
of course is a result of the experimental setup where the slab This corresponds, approximately, to the change from uncracked
was supported along the contraflexure lines while the loaded to cracked stiffness on the load–deflection curves, as evident in
column was free to displace vertically. The experimental and Figs. 9d1 and 9b2 for SF0 and SE0-R, respectively. The numer-
LFEM load–deflection curves are compared in Fig. 9 for all the ical study also confirms that where the opening is immediately
ten specimens. The LFEM deflection predictions are accurate. adjacent to the column, the first cracks start at the corners of
The effectiveness of SSR on the load–deflection behaviour the opening away from the column and propagate towards the
of the connections is demonstrated in Fig. 10 where the results supported corners of the slab at a 45˚ angle. These can be seen
for the specimens with SSR (i.e., XXX-R, CF0-R, SE0-R, and in Figs. 11a1 and 11a2.
SF0-R) are compared with their counterparts without SSR (i.e., A larger number of cracks spread over a greater area are dis-
XXX, CF0, SE0, and SF0). The results confirm that the prop- played in the LFEM results compared with the experimental
erly placed and anchored shear stud reinforcement not only observations. This is because in the LFEM, a crack is displayed
strengthen the slab but also increase the ductility (larger de- at any Gauss point at which the tensile strength of concrete (ft )
flection) of the slab–column connections. This is also in good is exceeded regardless of the length or width of the crack. In
agreement with the experimental findings in that the specimens the experimental work, however, many of the smaller cracks are
with SSR experienced a ductile flexural failure, whereas those either not visible to the human eye or merge together forming
without SSR exhibited a brittle punching failure mechanism, as a larger and more localized crack. Nevertheless, together with
demonstrated in Fig. 10, and reached lower ultimate load with the deformed shape and load–deflection response, the punch-
smaller deflection. ing shear strength and failure behaviour of the slab–column
connections can be simulated satisfactorily by the LFEM.
4.2. Ultimate strength
The predicted and the experimental ultimate loads V for both
series I and II are presented in Table 2. Also included are the 5. Effect of openings on deflections and
predictions according to the Australian standard AS3600-2001 strength
(SAA 2001), the American code ACI 318 (ACI 2005), and the
Canadian standard CSA-A23.3 (CSA 2004). The ratios of the The influence of the size of the openings on the load–deflection
experimental and the predicted values are also presented in Ta- behaviour of the slabs is demonstrated in Fig. 12, where the re-
ble 2. As can be seen in Table 2, the LFEM is able to predict the sults for CF0 (with a 250 mm × 250 mm opening) and SF0
ultimate load with good accuracy having mean ratios of 0.92 (with a 150 mm × 150 mm opening) are compared with XXX,
and 1.03, respectively, for series I and II. The ACI code, on the the control specimen without an opening. The results, both the-
other hand, consistently underestimates the shear strength of the oretical and experimental, confirm that the larger the opening,
specimens, by approximately 39% and 30%, respectively, for the larger the reduction in slab stiffness, or the lower the ulti-
the two series (not considering the results for CF0 and CF0-R). mate punching shear strength and the larger the deflection at
For all specimens, the Australian standard underestimates the the same load level.

© 2007 NRC Canada


960 Can. J. Civ. Eng. Vol. 34, 2007

Fig. 9. Comparison of load-deflection curves: (a1) XXX; (b1) SE0; (c1) CF0; (d1) SF0; (e1) SF1; (a2) XXX-R; (b2) SE0-R;
(c2) CF0-R; (d2) SF0-R; (e2) SF2. Experimental results from El-Salakawy et al. (1999, 2000).

© 2007 NRC Canada


Guan and Polak 961

Fig. 10. Effectiveness of SSR on load–deflection characteristics: (a) XXX and XXX-R; (b) CF0 and CF0-R; (c) SE0 and SE0-R;
(d) SF0 and SF0-R. Experimental results from El-Salakawy et al. (1999, 2000).

Table 2. Ultimate loads Vu in kN. LFEM, layered finite element method.

Experiment LFEM AS3600 (SAA 2001) ACI318 (ACI 2005) CSA-A23.3 (CSA 2004)
Vu,EXP Vu,EXP Vu,EXP Vu,EXP
Specimen Vu,EXP (kN) Vu,LFEM (kN) Vu,AS (kN) Vu,ACI (kN) Vu,CSA (kN)
Vu,LFEM Vu,AS Vu,ACI Vu,CSA
Series I
XXX 125 135 0.93 73 1.71 87 1.44 87 1.44
SE0 120 125 0.96 64 1.88 76 1.58 76 1.58
CF0 87 102 0.85 56 1.55 33 2.64 36 2.42
SF0 110 121 0.91 63 1.75 56 1.96 56 1.96
SF1 115 126 0.91 69 1.67 69 1.67 69 1.67
SF2 114 121 0.94 68 1.68 75 1.52 76 1.50
Mean 0.92 1.71 1.80 1.76
Standard deviation 0.04 0.11 0.45 0.37
Series II
XXX-R 154 144 1.07 108 1.43 129 1.19 186 0.83
SE0-R 150 144 1.04 95 1.58 112 1.34 163 0.92
CF0-R 105 112 0.94 82 1.28 50 2.10 71 1.48
SF0-R 146 134 1.09 93 1.57 84 1.74 120 1.22
Mean 1.03 1.46 1.59 1.11
Standard deviation 0.07 0.14 0.41 0.30

The effects of the opening location in terms of its distance The influence of the location of an opening in the slab – edge
from the front column face is relatively small as evident in column connections with SSR is examined numerically herein,
Fig. 13a, which compares the experimental and theoretical with two models SF1-R and SF2-R (Fig. 2b). Such influence
(LFEM) load–deflection curves for specimens SF0, SF1, and on the load–deflection response predicted by the LFEM is pre-
SF2. However, the specimen with edge face opening (SE0) was sented in Fig. 13c. It is evident that the further the opening from
stiffer than that with front face opening (SF0) although the ul- the column, the higher the ultimate strength of the connection
timate strengths were similar. This is illustrated in Fig. 13b. (with a slight increase). This is true for both the experimental

© 2007 NRC Canada


962 Can. J. Civ. Eng. Vol. 34, 2007

Fig. 11. Comparison of crack patterns of SF0 and SE0-R: (a1) tension layer — predicted (SF0); (b1) tensile surface — observed (SF0);
(a2) tensile layer — predicted (SE0-R); (b2) tensile surface — observed (SE0-R). (Observed patterns from El-Salakawy et al. (1999,
2000).)

Fig. 12. Influence of the size of opening. the SSR increases the strength of the slab – edge column con-
nections. All codes are conservative in predicting the strength
of slabs with openings. Both experimental results and LFEM
tend to suggest that the existence of a square opening, of size
up to 70% of the column cross section (the openings analyzed
herein were 70% of the column cross section), does not signifi-
cantly change the punching strength of the slab. The maximum
reduction of strength (for openings next to the column) as com-
pared to the specimen without an opening was 12% for the
experimental study and 10% for LFEM for slabs without SSR,
and 5% and 7%, respectively, for slabs with SSR. At the same
time, all codes predict reduction of strength of approximately
14% (SAA 2001), 35% (ACI 2005), and 36% (CSA 2004) for
all slabs, without and with SSR. The method for accounting of
openings in the presented codes seems to be overly conservative
and theoretical (LFEM) results. Also, from a deflection view- and further studies should be undertaken to clarify this issue.
point, the specimen with an opening immediately adjacent to
a column (i.e., SF0-R) has lower stiffness than the specimens
with the openings away from the column face. 6. Conclusions
The ultimate loads for specimens with openings in front of the
column predicted by the LFEM and the codes (SAA 2001; ACI The nonlinear layered finite element method (LFEM) has
2005; CSA 2004) are compared in Table 3. The codes adopt been used for the analytical study on the punching shear be-
a strategy of removing a portion of the critical perimeter de- haviour of slab – edge column connections with and without
pending on the size and location of the opening (Fig. 14). Both openings and shear stud reinforcement (SSR). The finite ele-
the LFEM and code predictions confirm that the existence of ment formulation is based on layered shell elements, which in-

© 2007 NRC Canada


Guan and Polak 963

Fig. 13. Influence of the location of opening (a) in the vicinity Table 3. Comparison of ultimate load Vu for specimens
of front face of column, (b) adjacent to the front versus edge face with varying locations of opening.
of column, and (c) with SSR in the vicinity of the front face of
the column. Experimental results from El-Salakawy et al. (1999, Specimen Expt. Vu,LFEM Vu,AS Vu,ACI Vu,CSA
2000). (kN) (kN) (kN) (kN)
XXX 125 135 73 87 87
SF0 110 121 63 56 55
SF1 115 126 69 69 69
SF2 114 121 68 75 76
XXX-R 154 144 108 129 186
SF0-R 146 134 93 84 120
SF1-R — 135 101 105 149
SF2-R — 138 98 113 160
Note: Expt., experiment; LFEM, layered finite element method;
AS, AS3600 (SAA 2001); ACI; ACI 318 (ACI 2005); CSA,
CSA-A23.2 (CSA 2004).

Fig. 14. Critical shear perimeter: (a) SF0; (b), SF1; and (c) SF2.

it shows that the layered nonlinear finite element analysis can


be used in design of reinforced concrete slab–column structural
systems.
Also, based on this work, a benchmark model is established
that enables a numerical investigation on the influence of the
locations of opening in slab – edge column connections with
SSR. The findings of this investigation are summarized below.
It should be noted that the study included only square openings
located centrally at any side of the column.
• The influence of the opening on the ultimate strength is
small when the size of opening is smaller than 70% of the
column size. This is true for slabs with and without SSR.
However, the reduction in strength is less when SSR is
present.
• The reduction of the slab strength is smaller when the
opening is located not in the direction of the unbalanced
moment (compare specimens SE0 and SF0).
• The connections with openings have lower stiffness than
the connections without openings. This reduction of stiff-
ness decreases with the distance of the opening from
clude shear deformations and consider three-dimensional states
the column face. The existence of the SSR increases the
of stress within each layer. Appropriate constitutive models for
strength and ductility of edge connections with openings.
concrete and reinforcement are adopted. The presented study
shows that the LFEM is able to predict, with good accuracy, the • When an opening of the size of a column is to be con-
deflections, the ultimate strengths, and the progressive crack structed, a different (than a flat plate on column) structural
patterns, up to failure. The study has also confirmed the exper- carrying system must be employed.
imental findings with respect to the influence of the opening
and the SSR on the ultimate strength of such connections. This Further, this study has demonstrated that the LFEM is an
has an important value for practical industrial applications as effective tool in global analysis suitable for the design-type

© 2007 NRC Canada


964 Can. J. Civ. Eng. Vol. 34, 2007

problems associated with reinforced concrete flat-plate floor Polak, M.A. 1998a. Shear analysis of reinforced concrete shells using
systems. Thus, design engineers can have such a tool by which degenerate elements. Computers and Structures, 68(1-3): 17–29.
they can determine punching and (or) flexural failures of this Polak, M.A. 1998b. Modelling punching shear of RC slabs using lay-
important class of concrete structures. ered finite elements. ACI Structural Journal, 95(1): 71–80.
Polak, M.A. 2005. Shell finite element analysis of RC plates supported
on columns for punching shear and flexure. Engineering Computa-
References tions, 22(4): 409–428.
Rericha, P. 1991. Layer model of bending-shear failure in RC plates and
ACI. 2005. Building code requirements for reinforced concrete (ACI
beams. Journal of Structural Engineering, ASCE, 117(10): 2865–
318-05) and commentary - ACI318R-05. American Concrete Insti-
2883.
tute (ACI), Farmington Hills, Mich.
Roll, F., Zaidi, S.T.H., Sabnis, G., and Chuang, K. 1971. Shear resis-
CSA. 2004. Design of concrete structures. Standard CSA-A23.3-04,
tance of perforated reinforced concrete slabs. In Cracking, deflec-
Canadian Standards Association (CSA), Toronto, Ont.
tion, and ultimate load of concrete slab systems. ACI Publication,
El-Salakawy, E.F., Polak, M.A., and Soliman, M.H. 1999. Reinforced
American Concrete Institute, Detroit, Mich. SP-30, pp. 77–101.
concrete slab–column edge connections with openings. ACI Struc-
SAA. 2001. Concrete structures. Standard AS3600-2001, Standards
tural Journal, 96(1): 79–87.
Association of Australia (SAA), Sydney, Australia.
El-Salakawy, E.F., Polak, M.A., and Soliman, M.H. 2000. Reinforced
Staller, M. 2001. Numerical analysis with the finite element program
concrete slab–column edge connections with shear studs. Canadian
MARC. In Punching of structural concrete slabs. International Fed-
Journal of Civil Engineering, 27(2): 338–348.
eration for Structural Concrete, Lausanne, Switzerland. fib Bulletin
Gomes, R.B., and Silva, J.A. 2003. Punching shear resistance of rein-
12, pp. 103–109.
forced concrete flat slab with openings and rectangular columns. In
Vecchio, F.J., and Collins, M.P. 1986. The modified compression field
Proceedings of the 9th East Asia-Pacific Conference on Structural
theory for reinforced concrete elements subjected to shear. Journal
Engineering and Construction, RCS session, Bali, Indonesia, 16–18
of Structural Engineering, ASCE, 83(6): 925–933.
December 2003. Edited by D. Hoedajanto, I. Imran, and M. Abduh.
Institut Teknologi Bandung, Bali, Indonesia. pp. 231–236.
Guan, H., and Loo,Y.C. 1997. Layered finite element method in crack- List of symbols
ing and failure analysis of RC beams and beam-column-slab con-
nections. Structural Engineering and Mechanics: An International
a flow vector
Journal, 5(5): 645–662.
d effective depth of slab
Guan, H., and Loo, Y.C. 2003. Failure analysis of column–slab con-
D total material constitutive matrix in local
nections with stud shear reinforcement. Canadian Journal of Civil
coordinate system
Engineering, 30(5): 934–944.
Harmon, T.G., and Ni, Z.Y. 1989. Shear-strength of reinforced- Dc constitutive matrix for concrete in material
concrete plates and shells determined by finite-element analysis coordinate system
using layered elements. Journal of Structural Engineering, ASCE, Dc constitutive matrix for concrete in local
115(5): 1141–1157. coordinate system
Hinton, E., and Owen, D.R.J. 1984. Finite element software for plates Dc constitutive matrix for concrete at different
and shells. Pineridge Press Limited, Swansea, UK. stage
Hognestad, E., Elstner, R.C., and Hanson, J.A. 1964. Shear strength Dcr constitutive matrix for cracked concrete in
of reinforced structural lightweight aggregate concrete slabs. ACI material coordinate system
Structural Journal, 61(6): 643–656. Dcr constitutive matrix for cracked concrete in local
Irawan, P., Liew, J.T., Kong, K.L., and Teng, S. 2003. Punching shear coordinate system
strength of slab-column connections with opening and strength- Dcrs constitutive matrix for crushed concrete
ened with shear stud reinforcement. In Proceedings of the 9th East Dep elasto-plastic constitutive matrix
Asia-Pacific Conference on Structural Engineering and Construc- Ds constitutive matrix for steel in material
tion, RCS session, Bali, Indonesia, 16–18 December 2003. Edited coordinate system
by D. Hoedajanto, I. Imran, and M. Abduh. Institut Teknologi Ban- Ds constitutive matrix for steel in local coordinate
dung, Bali, Indonesia. pp. 97–102. system
Loo, Y.C., and Guan, H. 1997. Cracking and punching shear failure D distance of the opening from the column face
analysis of RC flat plates. Journal of Structural Engineering, ASCE, E1 , E2 , E3 elastic moduli of concrete in principal
123(10): 1321–1330. directions
Moe, J. 1961. Shearing strength of reinforced concrete slabs and foot- Ei fictitious modulus of elasticity
ings under concentrated loads. Development Department Bulletin Es , Es1 , Es2 elastic moduli of steel
D47, Portland Cement Association, Skokie, Ill. pp. 1–130. fc compressive strength of concrete
Mowrer, R.D., and Vanderbilt, M.D. 1967. Shear strength of ft tensile stress of concrete
lightweight aggregate reinforced concrete flat plates. ACI Structural fy yield strength of steel
Journal, 64(11): 722–729. fst yield strength of transverse reinforcement
Ožbolt, J., and Vocke, H. 2001. Three dimensional numerical analysis G, G12 , G13 , G23 shear moduli of concrete
of punching failure. In Punching of structural concrete slabs. Inter- Gc12 , Gc13 , Gc23 reduced shear moduli for cracked concrete
  
national Federation for Structural Concrete, Lausanne, Switzerland. Gc12 , Gc13 , Gc23 reduced shear moduli for cracked concrete
fib Bulletin 12, pp. 71–78. H  hardening parameter

© 2007 NRC Canada


Guan and Polak 965

K bulk modulus of concrete εst strain in the direction of reinforcement


Tε strain transformation matrix εu ultimate compressive strain of concrete or ultimate strain
u critical shear perimeter of steel
u, v, w nodal displacements εz normal strain in transverse direction
Vu ultimate load of slab–column connection ε strain component
V, H vertical and horizontal forces applied to column stub θx , θy nodal rotations
x  , y  , z material coordinate system ν Poisson’s ratio
x, y, z local coordinate system ρs reinforcement ratio
α t , εm tension stiffening parameters ρst average transverse reinforcement ratio for the element
ε 1 , ε2 principal tensile strains in directions 1 and 2, σ stress component
respectively σz stress component normal to midreference plane

© 2007 NRC Canada

Vous aimerez peut-être aussi