Anda di halaman 1dari 10

Engineering Structures 56 (2013) 1718–1727

Contents lists available at ScienceDirect

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

Behaviour of partially encased composite columns with high strength


concrete
Mahbuba Begum a, Robert G. Driver b,⇑, Alaa E. Elwi b
a
Dept. of Civil Engineering, BUET, Dhaka, Bangladesh
b
Dept. of Civil and Environmental Engineering, University of Alberta, Edmonton, AB, Canada

a r t i c l e i n f o a b s t r a c t

Article history: The behaviour of partially encased composite columns under eccentric and concentric axial loading tends
Received 15 June 2011 to be complex because of the interaction of the concrete with the thin-walled steel section. When con-
Revised 8 July 2013 structed with high strength concrete, developing numerical simulations of the response of these columns
Accepted 29 July 2013
under load is particularly challenging. In this paper, a finite element model capable of simulating this
Available online 4 September 2013
response is presented. A constitutive model that has the capability of predicting the overall stress–strain
behaviour of high strength concrete, including the post-peak softening branch and residual strength, is
Keywords:
selected. The finite element model has been validated by conducting simulations of the behaviour of par-
Composite column
Explicit
tially encased composite test columns with high strength concrete. The load-deformation response, ulti-
Finite element mate capacity, and failure mode are predicted accurately by the model for both concentric and eccentric
High strength concrete axial loading conditions. Finally, a parametric study is presented that uses the validated finite element
Local buckling model to identify the key differences in strength and failure behaviour of partially encased composite col-
Nonlinear umns when using high strength instead of normal strength concrete.
Partially encased Ó 2013 Elsevier Ltd. All rights reserved.
Numerical simulation

1. Introduction columns indicated that the behaviour of this composite column


is significantly affected by the local instability of the thin steel
In recent years, high strength concrete has been used widely in flanges. The failure of the composite columns occurred by a combi-
buildings, bridges and other structures. Use of high strength con- nation local buckling of the steel flanges between the transverse
crete in columns can significantly reduce the size of the column links, yielding of the steel and crushing of the concrete. Experi-
and consequently reduce the dead load on the foundation system. ments on short PEC columns with high performance concrete un-
Moreover, the available floor space for a building can be increased der pure axial compression, as well as combined axial and
due to the reduction in column size. Partially encased composite flexural compression have been carried out by Prickett and Driver
(PEC) columns are one of the recent developments in steel- [6] under monotonic conditions. Behaviour of PEC columns under
concrete composite construction. This new composite system is axial compression and cyclic horizontal loads has been studied
composed of very thin steel plates that are welded together into by Bouchereau and Toupin [5] and Chen et al. [7]. Tests of beam-
an H-shape, as shown in Fig. 1. Transverse links are installed to-column connections were also performed for beams framing
between the flange tips to inhibit the occurrence of local buckling, into the weak axis of the PEC column by Muise [8].
and concrete is then poured between the web and the flange plates. Numerical simulations on partially encased composite columns
Extensive experimental research has been conducted on thin- have been performed by Maranda [9], Chicoine et al. [10], Begum
walled PEC columns with built-up sections by several research et al. [11] and Chen et al. [7]. Maranda [9] and Chicoine et al.
groups [1–8] to investigate their behaviour under various loading [10] modelled one-quarter of the column cross-section with a
conditions. A series of tests on PEC columns with normal strength length of one link spacing. The model developed by these research-
concrete under monotonic concentric and eccentric axial loads ers provided a very good representation of the capacity and load
have been performed by Filion [1], Tremblay et al. [2], Chicoine versus displacement response of short PEC test specimens with
et al. [3,4], Bouchereau and Toupin [5] and Prickett and Driver normal strength concrete [2,3] up to the ultimate load. However,
[6]. The results of these experimental investigations on PEC the researchers identified significant challenges in simulating the
local instability of the thin flanges and the triaxial behaviour of
the partially confined concrete in the column. Begum et al. [11]
⇑ Corresponding author. were able to overcome these challenges in the finite element mod-
E-mail address: rdriver@ualberta.ca (R.G. Driver). el through the implementation of a dynamic explicit formulation

0141-0296/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2013.07.040
M. Begum et al. / Engineering Structures 56 (2013) 1718–1727 1719

along with a damage plasticity model for concrete and a contact


pair algorithm at the steel-concrete interface. These researchers
implemented the model for reproducing the test results of a large Welded H
number of normal strength concrete PEC test columns [3,5,6] steel shape
encompassing a wide variety of geometric properties and loading t
conditions. The model has been shown to predict the behaviour
of tested PEC columns accurately when they are constructed with Link
normal strength concrete [11]. However, the suitability of this
d
model for predicting the response of PEC columns with high t
strength concrete had not been explored previously. Moreover, Concrete
the reference test database for PEC columns with high strength
concrete is not as rich as the database for those with normal
b
strength concrete. The primary difference in modelling PEC col-
umns with high strength concrete as opposed to normal strength bf
concrete lies in the unique nature of the constitutive relationship
of high strength concrete. Though the effects of high strength con-
crete on the behaviour of fully encased composite columns are rel- (a) (b)
atively well-understood through tests as well as numerical
simulations [12], the results cannot be used directly for PEC col- Fig. 1. PEC columns: (a) cross-section; (b) 3D view of the steel configuration.
umns where the effect of confinement on the concrete is negligi-
ble [3–6]. Therefore, the effects of high strength concrete on
the strength and ductility of PEC columns with various geomet- b ¼ 0:058f cu þ 1:0 ðfcu in MPaÞ ð3Þ
ric properties needs to be explored through parametric
investigations. Wee et al. [22] reported that Eqs. (1)–(3) give good predictions of
the ascending portion of the stress–strain curves for both normal
and high strength concretes. However, for high strength concrete
2. Constitutive models for high strength concrete the descending branch and the post-peak residual strength at high
strains are not adequately represented by the Carreira and Chu [16]
2.1. Uniaxial compression model. Therefore, two modification factors, k1 and k2, were intro-
duced into Eq. (1) by Wee et al. [22] for modelling the descending
Complete stress–strain curves for concrete under uniaxial com- branch of the stress–strain response. The expression for the
pression and tension are necessary to predict the structural re- descending branch becomes:
sponse of PEC columns by nonlinear finite element analysis. 2   3
There have been many attempts [13–23] to develop analytical for- k e
1 b ecu
6 7
mulations to represent stress–strain relationships for normal and fc ¼ fcu 4  k2 b 5 ð4Þ
high strength concrete in uniaxial compression. These models tend k1 b  1 þ ecue

to be good at predicting the overall mechanical behaviour of nor-


mal strength concrete; however, the strain softening behaviour The factors k1 and k2 were determined empirically by Wee et al.
and post-peak residual strength for high strength concrete (greater [22] based on their experimental investigations of high strength
than about 50 MPa) are not adequately represented by these con- concrete behaviour (with fcu ranging from 50 to 120 MPa), as
stitutive models, with the exception of the one proposed by Wee follows:
et al. [22], which has been adopted for the finite element model.  3:0
50
Wee et al. adopted an equation proposed by Carreira and Chu k1 ¼ fcu
[16], and applied two correction factors to better represent the  1:3 ð5Þ
50
post-peak descending branch for high strength concrete. The k2 ¼ fcu
ascending branch of the concrete stress, fc, versus strain, e, response
Thus, for a given concrete strength, fcu, and corresponding strain, ecu,
is represented by the following equation:
2 3 the complete stress–strain curve for high strength concrete under
 
b eecu compression can be generated by Eqs. (1)–(5).
6 7
fc ¼ fcu 4  b 5 ð1Þ
b  1 þ ecue 2.2. Uniaxial tension

where b is a material parameter that depends on the shape of the For nonlinear finite element analysis of PEC columns under
stress–strain curve, and is given by: generalised loading, the stress-displacement relationship for con-
crete members in the cracked state under uniaxial tension is re-
1 quired. In an uncracked state, the stress–strain relationship
b¼   ð2Þ
fcu
1 ecu Eit under uniaxial tension can be expressed by a linear function
[24,25]. After cracking, however, the stress versus crack width re-
Here, Eit is the initial tangent modulus of concrete, fcu is the uni- sponse is more appropriate for representing the material behav-
axial compressive strength of concrete, and ecu is the strain at the iour under uniaxial tension, since no unique stress–strain
ultimate stress. Several researchers [16,22] have proposed rela- relationship exists in the post-peak softening region [26]. Most
tionships between Eit and fcu based on experimental investiga- researchers [26,25,27–29] use exponential functions to represent
tions of the uniaxial compressive behaviour of concrete. Since the post-cracking behaviour of concrete. The current study uses
the initial tangent modulus, Eit, is not normally readily available the model proposed by Li and Ansari [28] for high strength con-
from standard material tests, a simplified expression for b as a crete, wherein the post-peak stress response, ft, is represented as
function of fcu only, proposed by Popovics [14], is used in this an exponential function of the ratio of crack width, w, to final
research: crack width, wf, as follows:
1720 M. Begum et al. / Engineering Structures 56 (2013) 1718–1727

Table 1
Geometric properties of reference test specimens.

Specimen Plate size Length Plate Link Eccentricity


bf  d  t (mm) Slenderness ratio Spacing Diameter
L (mm) b/t s (mm) Ratio of d / (mm) ex (mm) ey (mm)
H3 400  400  7.99 2000 25 120 0.3d 12.8 – –
H4 400  400  8.01 2000 25 200 0.5d 12.8 – –
H5 400  400  8.02 2000 25 400 1.0d 15.9 – –
H8 400  400  7.95 2000 25 240 0.6d 12.8 23 –
H9 400  400  7.98 2000 25 240 0.6d 12.8 100 –
H10 400  400  8.01 2000 25 240 0.6d 12.7 – 25
H11 400  400  7.95 2000 25 240 0.6d 12.7 – 74

8 " #0:9 9
< 0:03 = Link
ft ¼ ftu 1  exp w ð6Þ
: w
; t b
f Welded H
steel shape bf
where ftu is the ultimate tensile strength of concrete. t
While modelling the PEC columns in the current research, the
values of final crack width are chosen in such a way that consistent d = bf
fracture energy density value for concrete, as reported in the liter-
ature [26,28–30], are obtained. The valuesof the final crack widths
(a) End plate 3d

used in current study are 1.5 mm and 0.3 mm respectively, for nor-
mal strength concrete and high strength concrete.
d

3. Properties of PEC column test specimens with high strength


concrete

Prickett and Driver [6] conducted a comprehensive experimen-


tal research project to study the behaviour of thin-walled PEC col- 3d Test
umns made with high performance concrete. Nine short PEC region
columns, designated H3 to H11 and described in Table 1, con-
structed with high strength concrete were tested under concentric
and eccentric axial loading conditions. The test columns had a s
gross cross-section of 400 mm  400 mm, with the steel plates
having a b/t ratio of 25. Three link spacing values – 0.3d, 0.5d End region (with closer link
and 1.0d – were provided in columns H3 to H7 (see Table 1), which d spacing and high strength
were subjected to axial compression only. On the other hand, a link concrete)
spacing of 0.6d was used in the eccentrically loaded columns (H8
through H11) to just meet the maximum spacing requirement in (b)
standard S16 of the Canadian Standards Association [31]. These
four specimens were designed to have identical geometric proper- Fig. 2. Geometry of a typical PEC test column, (a) cross-section, and (b) elevation.

ties, as shown in Table 1. The parameters that varied for specimens


H8 through H11 were the load eccentricity and the column orien-
tation. Columns H8 and H9 were loaded at 23 mm and 100 mm strength concrete of 60 MPa (nominal strength) was used in the
eccentricities, respectively, inducing bending about the strong axis, test region of these specimens. Two of the nine specimens (H6
whereas columns H10 and H11 were loaded to have weak axis and H7) had 1% steel fibres by volume in the high strength con-
bending at 25 mm and 74 mm eccentricities, respectively. Table 2 crete, and these specimens are excluded from the numerical simu-
provides the mechanical properties for the steel section and the lations presented in this paper. The end regions of all the test
test region concrete in these specimens. The steel section was fab- specimens had very high-strength concrete (nominally 80 MPa)
ricated with CSA-G40.21 grade 350W [32] steel plate. High to restrict failure to the test regions. A typical test column is shown

Table 2
Material properties of reference test specimens.

Specimen Properties of concrete in test region Properties of steel plate


fcu Ec ecu m Fy Fsh Fu ey esh eu
(MPa) (MPa) (le) (MPa) (MPa) (MPa) (%) (%) (%)
H3 60.0 28,000 2880 0.16 394 394 528 0.19 1.67 15.80
H4 58.9 28,800 2995 0.16 394 394 528 0.19 1.67 15.80
H5 61.7 28,500 3165 0.16 394 394 528 0.19 1.67 15.80
H8 62.4 29,800 2900 0.17 394 394 528 0.19 1.67 15.80
H9 64.5 28,600 3040 0.15 394 394 528 0.19 1.67 15.80
H10 65.7 29,000 2745 0.15 394 394 528 0.19 1.67 15.80
H11 65.1 28,700 3225 0.16 394 394 528 0.19 1.67 15.80
M. Begum et al. / Engineering Structures 56 (2013) 1718–1727 1721

in Fig. 2. Prickett and Driver [6] reported that the high strength the flange plates. To account for the shape change, the nodal coor-
concrete PEC columns ultimately failed by concrete crushing com- dinates are updated at the beginning of each increment to reflect
bined with local flange buckling. Similar failure modes were ob- current positions in space, and all the shape functions and deriva-
served in PEC columns with normal strength concrete [3,6]. tives are re-evaluated using the updated nodal coordinates.
However, the failure of a high strength concrete column was ob- The transverse links were modelled using B31 beam elements.
served to be sudden, as compared to an equivalent PEC column Fig. 3(a) shows the complete model for a short PEC test column
with normal strength concrete. with the spacing s = 0.5d. This figure displays the boundaries be-
tween the different parts of the column. The mesh itself is shown
4. Finite element model in Fig. 3(b) for a typical part in the test region of the column. A de-
tailed description of the mesh configuration can be found in Begum
4.1. Geometric properties et al. [11]. The end plates used in the test specimens are repre-
sented by rigid body surfaces and are fixed to the adjacent nodes
4.1.1. Mesh description of the end surfaces of the column. The local flange imperfections
The finite element model for this new composite system is were excluded from the numerical model since Begum et al. [11]
implemented using the ABAQUS [33] finite element code. The full found negligible effect of local imperfection on the overall behav-
length and cross-section of the test specimens are included in iour of PEC columns.
the numerical analysis to ensure that all features of the response
are simulated accurately. The steel section of the PEC column is
constructed with thin plates, which are susceptible to local buck- 4.1.2. Modelling of steel-concrete interactions
ling causing large rotations at the flange plates and, hence, adding One of the most challenging aspects of this study is to model
geometric nonlinearity to the behaviour. In the tests [2–6], the successfully the steel-concrete interaction at their interfaces with
short PEC columns reached their ultimate capacity at the simulta- a contact algorithm. In PEC columns, initially there is contact be-
neous occurrence of local buckling of the thin flanges and crushing tween the steel plates and the adjacent surfaces of the concrete in-
of the concrete. In order to capture this behaviour, finite strain S4R fill. As the loading progresses, the flanges of the steel section
shell elements were used to model the steel plates and C3D8R ele- between two consecutive links may experience local buckling
ments were selected to model the concrete blocks between the resulting in a separation between the flange and the concrete.
flange plates and the web of the composite section. The C3D8R ele- However, neither local buckling of the web nor separation between
ment is an eight-node reduced-integration brick element with the web plate and concrete block was observed [2,3,5,6]. Therefore,
three translational degrees of freedom at each node. A mesh den- a contact algorithm was used only at the interior flange surfaces.
sity sensitivity analysis was performed with the S4R and C3D8R To implement the contact pair algorithm in ABAQUS, the two con-
elements to optimise the mesh in order to produce proper repre- tact surfaces are first defined geometrically. The steel plate surface
sentations of local buckling of the steel flange and simultaneous is defined as the master surface, whereas the concrete surface in
concrete crushing, while maintaining reasonable computing econ- contact is defined as the slave surface. While the two surfaces
omies. For all the elements except the elements defined at the are in contact, the shear and normal forces are transmitted across
web-flange junction, the aspect ratio was close to 1.0. At the corner the interface. To allow separation, the tensile bond strength be-
where the plates meet, narrow elements with a width equal to tween the contact surfaces is assumed to be zero. A mechanical
one-half the thickness of the plates are defined to match the mesh interaction model, including friction, is defined to model this
of the steel plates with that of the concrete. All continuum and interaction between the steel and concrete surfaces. A balanced
plate elements in ABAQUS/Explicit [33] are based on an updated master-slave contact formulation was used, which prevents the
Lagrangian formulation. This formulation is useful for the current penetration of the contacting bodies and provides an appropriate
problem because the elements experience considerable shape simulation of the interaction between the steel flange and concrete
changes resulting from large rotations due to local buckling of of an imperfect PEC column [11].

Flange 2 2
Web P Rigid
P P
beam

End u1=u3=0 u1= u3=0


u1= u3=0
region θ1= θ2= θ3=0 θ2= θ3=0 θ1= θ2=0
Links ex ey
s = 0.5d Column
centreline
Test Web Flange
region Concrete

2 s
Rigid
beam
End
region
Flange Concrete
3 1 3 1
u1=u2=u3=0 u1=u2=u3=0 u1=u2=u3=0
θ1= θ2= θ3=0 θ2= θ3=0 θ1= θ2=0
(b) Strong axis Weak axis
bending
bending

(a)
Concentrically Eccentrically
loaded columns loaded columns

(c)
Fig. 3. Finite element model for PEC column, (a) typical PEC column displaying the parts between consecutive links; (b) mesh configuration of a typical part in the test region
of the column and (c) end boundary conditions applied in the FEM model.
1722 M. Begum et al. / Engineering Structures 56 (2013) 1718–1727

4.1.3. Modelling of end boundary conditions quasi-static problems with some special considerations in the ap-
The boundary conditions applied in the model to simulate the plied loading history. An accurate and efficient quasi-static analy-
end conditions for both the concentrically and eccentrically loaded sis requires the loading to be as smooth as possible. ABAQUS/
specimens are shown in Fig. 3(c). These constraints are applied Explicit has a simple, built-in type of amplitude function, called
through a reference node situated in each of the top and bottom ri- ‘‘smooth step’’ that limits spurious oscillations. In this study, the
gid surfaces of the finite element model. For the concentrically load was applied using a smooth amplitude displacement history
loaded columns, translations and rotations at both ends were fixed, at a rate of 8 mm/s that rendered the inertial effects negligible. This
except for the vertical displacement at the top. In the finite ele- rate was selected through an iterative process to not only minimize
ment model for eccentrically loaded test specimens, pinned-pin- the inertial effects, but also to achieve economy in the solution
ned end conditions are applied at the eccentric load points time.
located on the rigid end planes.

4.2. Material properties 5. Performance of the finite element model

The steel material properties for the plate and link are simu- 5.1. Ultimate strength and failure behaviour
lated with a trilinear elasto-plastic model that includes strain hard-
ening, using the measured material test data for the steel plate The values of experimental and numerical peak loads and peak
listed in Table 2. The subscripts y, sh, and u in the table signify strains, along with their ratios, for the seven PEC columns with
the yield, onset of strain hardening, and ultimate stresses and high strength concrete are shown in Table 3. The mean value of
strains, respectively. The value of Poisson’s ratio for steel used in the experimental-to-numerical peak load ratio, Pexp/Pnum, is 0.99,
the numerical analysis is 0.3. The damage plasticity model in ABA- with a standard deviation of 0.03. This indicates the excellent per-
QUS is used to simulate the concrete material behaviour in the formance of the finite element model in predicting the ultimate
composite columns. The uniaxial compressive stress–strain re- capacity of PEC columns with high strength concrete in both con-
sponse of high strength concrete is assumed to be linear up to centrically and eccentrically loaded conditions. As shown in
30% of its compressive strength. In the plastic regime, the effective Table 3, the ratio of the experimental-to-numerical average axial
stress-plastic strain compressive response is described through a strain at peak load, eexp/enum, ranged from 0.97 (specimen H11) to
stress–strain function using the models described in Section 2.1. 1.28 (specimen H9), with a mean value of 1.09 and a standard devi-
The material properties used to generate the curve, listed in Table 2, ation of 0.12. The ratios for specimens H4, H5, H10 and H11 are all
were obtained from standard cylinder tests performed at the test very close to 1.0, indicating an accurate representation of ductility
day on the concrete used in each test specimen. The uniaxial ten- up to the peak load. However, the numerical model is observed to
sile strength of concrete, ftu, was set at 5% of the uniaxial compres- underestimate the axial deformation at peak load significantly for
sive strength for high strength concrete, as recommended by specimens H3, H8 and H9. The possible reason behind this could be
Marzouk and Chen [30]. In the damage model, the stress–strain re- the presence of larger confining pressures on the concrete, as com-
sponse under uniaxial tension follows a linear elastic relationship pared to the other specimens, exerted by the closer link spacing in
until the tensile strength is reached. The post-cracking tensile specimen H3 and by the compression flange in specimens H8 and
properties for the concrete model are defined as a stress-displace- H9, which had bending about the strong axis. In general, the failure
ment curve following the relationships described in Section 2.2. of the short PEC column specimens in the numerical analyses oc-
curred due to concrete crushing combined with local buckling of
4.3. Load application and solution technique the steel flanges between two transverse links. Since the column
cross-section in the numerical model is perfectly symmetric, local
In the finite element model the axial load is applied through the buckling in a concentrically loaded column is observed to occur on
top rigid body reference node, which is defined at the centre of the all four sides of the unsupported flange in the middle segment of
column cross-section for concentrically loaded columns and at the the column. On the other hand, local buckling is observed in the
eccentric point for the eccentrically loaded columns, using a dis- flange plates on the compression side only of the eccentrically
placement control technique. An explicit solution technique is loaded columns. The failure mode obtained from the numerical
implemented in the numerical analysis since the problem exhibits analysis, along with the stress contour for eccentrically loaded test
highly nonlinear behaviour containing contact. In this type of prob- specimen H8, is shown in Fig. 4(a). This specimen is observed to
lem, convergence may not be possible using the implicit method, experience local buckling at the compression flange accompanied
especially after the limit point. The explicit solution method, by steel yielding and concrete crushing at failure. Similar behav-
originally developed for dynamic problems, can be applied to iour was observed in the experiments [6], as shown in Fig. 4(b).

Table 3
Comparison of numerical and experimental results.

Specimen Eccentricity Peak axial load Pexp/Pnum Avg. axial strain at peak load eexp/enum
Num. Exp. Num. Exp.
ex ey Pnum Pexp enum eexp
(mm) (mm) (kN) (kN) (le) (le)
H3 – – 12,450 12,340 0.99 2890 3420 1.18
H4 – – 12,150 11,860 0.98 2810 2835 1.01
H5 – – 12,160 12,390 1.02 2890 2905 1.00
H8 23 – 10,700 10,920 1.02 2290 2650 1.16
H9 100 – 7160 7260 1.01 1590 2040 1.28
H10 – 25 10,100 9740 0.96 1820 1895 1.04
H11 – 74 6840 6370 0.93 1330 1290 0.97
Mean 0.99 1.09
SD 0.03 0.12
M. Begum et al. / Engineering Structures 56 (2013) 1718–1727 1723

Local buckling of flange

(a) (b)
Fig. 4. Comparison of deformed shapes at failure for eccentrically loaded PEC column H8. (a) FEM Model and (b) experimental.

14000 14000
12000 12000
10000
Load (kN)
Load (kN)

10000
8000 Test 8000
FEM Test
6000 6000
4000 4000 FEM

2000 2000
0 0
0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000
Strain ( με ) Strain (με)
(a) Specimen H3 (b) Specimen H4
14000
FEM
12000

10000
Load (kN)

8000
6000
Test
4000

2000
0
0 2000 4000 6000 8000 10000 12000
Strain (με)
(c) Specimen H5
Fig. 5. Numerical and experimental load versus strain behaviour for concentrically loaded PEC columns with HSC.

5.2. Load versus strain response s = 0.5d and 15.9 mm diameter spaced at s = 1.0d, respectively.
The numerical curve for specimen H3, which had 12.8 mm diame-
The experimental and numerical load versus strain responses ter links spaced at s = 0.3d, did not match as well with the experi-
for specimens H3, H4 and H5 are shown in Fig. 5. The numerical mental curve around and after the peak. Among all the tests
models for specimens H4 and H5 predicted the experimental re- performed on PEC columns published in the literature, speci-
sponse with good accuracy, both in the pre-peak and post-peak re- men H3 was the only column that had a link spacing of s < 0.5d.
gions of the load versus average axial strain curves. These two More tests on PEC columns with link spacing less than 0.5d
specimens had transverse links of 12.8 mm diameter spaced at are, therefore, required to propose a proper explanation for the
1724 M. Begum et al. / Engineering Structures 56 (2013) 1718–1727

Fig. 6. Numerical and experimental load versus strain behaviour for eccentrically loaded PEC columns with HSC.

Test H8
12000 12000
Test H10
10000 10000
Test H9
Load (kN)

8000 8000 Test H11


Load (kN)

FEM H8 FEM H10


6000 6000

4000 4000
FEM H9 FEM H11
2000 2000

0 0
0 200 400 600 800 1000 0 200 400 600
Moment (kN-m) Moment (kN-m)
(a) Specimens H8 and H9 (b) Specimens H10 and H11
Fig. 7. Numerical and experimental load versus moment curves for PEC columns with HSC.

disagreement between the numerical and experimental behav- the ability of the numerical model to predict the behaviour of
iours for specimen H3. PEC columns with high strength concrete accurately when they
Fig. 6 presents the performance of the finite element model in are subjected to simultaneous axial compression and flexure.
predicting the load versus strain behaviour of specimens H8 to
H11, which were subjected to axial compression and bending. 6. Effect of concrete strength on PEC column behaviour
The figure shows the load versus strain responses at both the ten-
sile and compressive extreme fibres of the cross-section, as well as The compressive strength of concrete plays an important role in
the average values. For all specimens, good agreement is in general increasing the load-carrying capacity of concrete, thereby reducing
obtained between the numerical and the experimental load versus the required column size for a particular design load. However,
strain behaviours, despite the complexity of the behaviour. limited experimental investigations have been performed, to date,
on PEC columns with high strength concrete. This research pre-
5.3. Load versus moment response sents a parametric study of PEC column behaviour conducted using
the validated finite element model. The study investigates the
Fig. 7 shows the numerical and experimental load versus mo- influence of high strength concrete in combination with other geo-
ment responses for specimens H8 to H11 tested by Prickett and metric parameters of the composite system. Concrete strengths of
Driver [6] for both strong and weak axis bending. For all four spec- 30 MPa (‘‘normal’’ strength) and 60 MPa (‘‘high’’ strength) were
imens, excellent agreement is observed between the numerical used in the study. Three reference columns, designated ‘SN1’,
and the experimental load versus moment curves over the entire ‘IN1’ and ‘LN1’ (as shown in Table 4), were designed using
loading histories (the responses are essentially coincident and 30 MPa concrete, where ‘‘S’’, ‘‘I’’ and ‘‘L’’ indicate short (L/d = 5),
can scarcely be distinguished in the figures). This demonstrates intermediate (L/d = 10) and long (L/d = 15) columns, respectively.
M. Begum et al. / Engineering Structures 56 (2013) 1718–1727 1725

The reference columns were designed to incorporate the effects of In Eqs. (7) and (8), fcu is in units of MPa. The Poisson’s ratio for con-
the influential geometric parameters. These are the initial load crete was taken as 0.20, as reported by Rashid et al. [36] for concrete
eccentricity-to-column depth ratio, e/d, flange slenderness ratio, with strengths ranging from 20 to 120 MPa.The material property
b/t, and link spacing-to-column depth ratio, s/d. The details are gi- curves for normal and high strength concrete used in the current
ven in Table 4. Two eccentricity ratios, three flange slenderness ra- parametric study are shown in Fig. 8(a) and (b).
tios and two link spacing ratios are considered in the investigation. Table 4 also shows the key results for normal and high strength
To differentiate between the normal strength and high strength concrete with nominal strengths of 30 and 60 MPa. The ultimate
concrete columns, the letters ‘‘N’’ and ‘‘H’’ are included in the col- axial load, Pu, of column SN1 (e/d = 0.15, b/t = 25, s/d = 0.5), IN1
umn designations. (e/d = 0.15, b/t = 30, s/d = 0.7) and LN1 (e/d = 0.30, b/t = 35,
All of the columns have a square cross-section, with outer s/d = 0.7), which were constructed with 30 MPa concrete, are in-
dimensions of 450 mm  450 mm. This is a moderate size for com- creased by 57%, 60% and 48%, respectively, when 60 MPa is used in-
posite columns and might be suitable in the construction of mid- stead. The moment, Mu, sustained at the peak compressive load is
rise buildings. The transverse links had a 12.7 mm diameter, which increased by 55% for columns SN1 and LN1 and 69% for column
meets the requirements of CSA standard S16-09 [27]. The steel sec- IN1. The average increases in the peak load and corresponding mo-
tion of the column and the transverse links were assumed to be ment are 55% and 57%, respectively.
fabricated from CSA-G40.21 grade 350W steel. The nominal yield As expected, the overall behaviour of the PEC columns is greatly
strength, Fy, of 350 MPa, was used in the analyses with an assumed affected by the compressive strength of concrete for all three sets
modulus of elasticity of 200 GPa. The yield strain, ey, was taken as of analysis, as shown in Fig. 9. The axial load versus average strain
0.00175 mm/mm. The ultimate strength of the steel plate is taken responses (Fig. 9(a)) for PEC columns with high strength concrete
as 450 MPa and the corresponding strain is assumed to be 100 show steeper slopes in the initial portions of the curves because
times the yield strain, i.e., 0.175 mm/mm. The point delineating of the higher modulus of elasticity of high strength concrete. Col-
the onset of strain hardening is defined at a stress value of umns SH1, IH1 and LH1 also demonstrate sharp post-peak strength
350 MPa, with a strain of 10ey. The Poisson’s ratio used for the steel declines as compared to columns SN1, IN1 and LN1, respectively.
is 0.30. Two types of concrete, with nominal strengths of 30 MPa For all geometries examined, the residual capacities of the columns
and 60 MPa, were used in the parametric study. The strain at the after failure are similar for the two concrete strengths.
ultimate compressive strength was calculated using the expression The moment versus curvature plots shown in Fig. 9(b) depict
proposed by Almusallam and Alsayed [34]: the ductile nature of the plastic hinge that forms at mid-height
in columns SN1, IN1, and LN1 as compared to those in columns
ecu ¼ ð0:2f cu þ 13:06Þ  104 ð7Þ SH1, IH1 and LH1, respectively. This effect appears to become less
which was developed based on experimental investigations of both pronounced as the overall column slenderness increases, although
normal and high strength concrete (up to 100 MPa). The elastic additional research is required to isolate the effects of the individ-
moduli for the normal and high strength concrete were calculated ual variables. This behaviour would be expected because as the col-
using the following expression according to ACI 363R-92 [35]: umn slenderness increases, global buckling behaviour gradually
pffiffiffiffiffi dominates over local flange buckling, irrespective of the strength
Ec ¼ 3320 fcu þ 6900 ð8Þ of the encased concrete.

70 3.0
High strength concrete
60 2.5
Tensile stress (MPa)

50 Normal strength concrete


Stress (MPa)

2.0
40 High strength concrete
1.5 Normal strength concrete
30
1.0
20

10 0.5

0 0.0
0 2000 4000 6000 8000 10000 12000 0 0.5 1 1.5 2
Strain (με) Crack width (mm)

(a) (b)
Fig. 8. (a) Stress–strain curves for concrete in uniaxial compression, and (b) stress–crack width curves for concrete in uniaxial tension.

Table 4
Effect of concrete compressive strength on strength and failure mode of PEC columns.

Column designation Column properties Load and Moment at the ultimate point %Difference Occurrence of local buckling
L/d e/d b/t s/d fcu Pu Mu Pu Mu
(MPa) (kN) (kN-m) (%) (%)
SN1 5 0.15 25 0.5 30 6428 458 – – After peak at 0.99Pu
SH1 5 0.15 25 0.5 60 10,100 712 57 55 After peak at 0.89Pu
IN1 10 0.15 30 0.7 30 5724 460 – – After peak at 0.99Pu
IH1 10 0.15 30 0.7 60 9160 776 60 69 After peak at 0.94Pu
LN1 15 0.3 35 0.7 30 3736 667 – – After peak at 0.92Pu
LH1 15 0.3 35 0.7 60 5521 1036 48 55 After peak at 0.92Pu
1726 M. Begum et al. / Engineering Structures 56 (2013) 1718–1727

12000 800
10000 30 MPa (SN1) 30 MPa (SN1)

Moment (kN-m)
60 MPa (SH1) 600 60 MPa (SH1)
8000
Load (kN)

6000 400
4000
200
2000

0 0
0 1000 2000 3000 4000 5000 0 1 2 3 4 5 6
Average axial strain (με) Curvature (10-5/mm)

12000
1200
10000 30 MPa (IN1)
1000 30 MPa (IN1)
60 MPa (IH1)

Moment (kN-m)
8000
Load (kN)

60 MPa (IH1)
800
6000
600
4000 400
2000 200

0 0
0 1000 2000 3000 4000 5000 0 1 2 3 4 5 6
Average axial strain (με) Curvature (10-5/mm)

8000 1200
30 MPa (LN1)
30 MPa (LN1) 1000
60 MPa (LH1)
6000
Moment (kN-m)

60 MPa (LH1)
800
Load (kN)

4000 600

400
2000
200

0 0
0 500 1000 1500 2000 2500 3000 0 1 2 3 4 5 6
Average axial strain (με) Curvature (10-5/mm)

(a) Load versus Average Strain (b) Moment versus Curvature


Fig. 9. Effect of high strength concrete on the behaviour of PEC columns.

The local flange buckling in column SN1 (e/d = 0.15, b/t = 25, friction-type simple master-slave contact was assumed at the
s/d = 0.5) and IN1 (e/d = 0.15, b/t = 30, s/d = 0.7) occurred shortly interface between the steel flange and concrete infill. The use of
after the onset of concrete crushing at a load of 0.99Pu, as indicated a dynamic explicit formulation and a concrete damage plasticity
in Table 4. When high strength concrete is used in these columns model permitted good predictions of the capacities of both uniax-
(i.e., columns SH1 and IH1) local buckling is delayed significantly. ially- and eccentrically-loaded PEC column tests reported in the lit-
The loads corresponding to the first occurrence of local buckling erature.The model provides good representations of the axial
in the flanges for these columns were 0.89Pu and 0.94Pu, respec- deformation at the peak load, the post-peak behaviour and the fail-
tively. However, the concrete compressive strength did not affect ure mode observed in those tests.
the initiation of local buckling in the flanges of the long column The validated numerical model for PEC columns was imple-
(LH1), since its behaviour is governed by global bending. Both long mented to identify the effects of high strength concrete on the
columns (LN1 and LH1) experienced local buckling at 0.92Pu after strength and failure behaviour of this new composite system. The
reaching the peak axial load. axial capacity of the PEC column, with a variety of L/d, e/d, b/t
and s/d ratios, is greatly increased (average increase for the cases
7. Summary and conclusions considered is 55%) by the use of high strength (60 MPa) concrete
instead of normal strength (30 MPa) concrete, at a relatively small
Finite element analyses were conducted to study the behaviour increase in associated cost. However, the load-deformation re-
of thin-walled PEC columns constructed with high strength con- sponse of high strength concrete PEC columns exhibited a some-
crete. Both nonlinear material behaviour and geometric nonlinear- what more brittle failure as compared to the normal strength
ities caused by large deformations were accounted for in the concrete columns, and similar residual capacities after failure.
numerical model. To simulate the nonlinear material behaviour
of high strength concrete, a constitutive model capable of tracing Acknowledgements
the overall stress–strain behaviour of high strength concrete,
including the post-peak softening branch and residual strength, Funding for this research program was provided by the Canam
was selected. The steel-concrete interface in the composite column Group, Saint-Georges, Québec and the Natural Sciences and Engi-
was modelled using the contact pair algorithm in ABAQUS. A neering Research Council of Canada (NSERC).
M. Begum et al. / Engineering Structures 56 (2013) 1718–1727 1727

References [15] Wang PT, Shah SP, Naaman AE. Stress–strain curves of normal and lightweight
concrete in compression. ACI J 1978;75(11):603–11.
[16] Carreira DJ, Chu KM. Stress–strain relationship for plain concrete in
[1] Filion I. Étude Expérimentale des Poteaux Mixtes Avec Section d’acier de classe
compression. ACI J 1985;82(6):797–804.
4’’. Report no. EPM/GCS-1998-06, Dept of Civil, Geological and Mining
[17] Tsai WT. Uniaxial compression stress–strain relation of concrete. J Struct Eng,
Engineering, Ecole Polytechnique, Montreal, Canada; 1998.
ASCE 1988;114(9):2133–6.
[2] Tremblay R, Massicotte B, Filion I, Maranda R. Experimental study on the
[18] CEB. CEB-FIP Model Code. Bulletin d’information, No. 203, Paris, France; 1990.
behaviour of partially encased composite columns made with light welded H
[19] Hsu LS, Hsu TC-T. Complete stress–strain behaviour of high strength concrete
steel shapes under compressive axial loads. In: Proc., SSRC annual technical
under compression. Mag Concr Res 1994;46(169):301–12.
session and meeting, Atlanta; 1998. p. 195–204.
[20] Almusallam TH, Alsayed SH. Stress–strain relationship of normal, high
[3] Chicoine T, Tremblay R, Massicotte B, Yalcin M, Ricles J, Lu L-W. Test
strength and lightweight concrete. Mag Concr Res 1995;47(170):39–44.
programme on partially-encased built up three-plate composite columns.
[21] Gysel VA, Taerwe L. analytical formulation of the complete stress–strain curve
Joint Report EPM/GCS No. 00-06, February, Dept of civil, geological and mining
for high strength concrete. Mater Struct 1996;29:529–33.
engineering, ecole polytechnique, Montreal, Canada – ATLSS Engineering
[22] Wee TH, Chin MS, Mansur MA. Stress–strain relationship of high-strength
Research Centre, No. 00-04, Lehigh University, Bethlehem, Pennsylvania, USA;
concrete in compression. J Mater Civil Eng, ASCE 1996;8(2):70–6.
2000.
[23] Barr B, Lee MK. Modelling the strain-softening behaviour of plain concrete
[4] Chicoine T, Massicotte B, Tremblay R. Long-term behavior and strength of
using a double-exponential model. Mag Concr Res 2003;55(4):343–53.
partially-encased composite columns with built up shapes. J Struct Eng, ASCE
[24] Hillerborg A, Modeer A, Peterson PE. Analysis of crack formation and crack
2003;129(2):141–50.
growth in concrete by means of fracture mechanics and finite elements. Cem
[5] Bouchereau R, Toupin J-D. Étude du Comportement en Compression-Flexion
Concr Res 1976;6:773–82.
des Poteaux Mixtes Partiellement Enrobés. Report EPM/GCS-2003-03. Dept of
[25] Foote RML, Mai YW, Cotterell B. Crack growth resistance curves in strain–
civil, geological and mining engineering. Ecole Polytechnique, Montreal,
softening material. J Mech Phys Solids 1986;34(6):593–607.
Canada; 2003.
[26] Gopalaratnam VS, Shah SP. Softening response of plain concrete in direct
[6] Prickett BS, Driver RG. Behaviour of partially encased composite columns made
tension. ACI J 1985;82(3):310–23.
with high performance concrete. Structural engineering report No 262. Dept of
[27] Du J, Yon JH, Hawkins NM, Arakawa K, Kobayashi AS. Fracture process zone for
civil and environmental engineering, University of Alberta, AB, Canada; 2006.
concrete for dynamic loading. ACI Mater J 1992;89(3):252–8.
[7] Chen Y, Wang T, Yang J, Zhao X. Test and numerical simulation of partially
[28] Li Q, Ansari F. High strength concrete in uniaxial tension. ACI Mater J
encased composite columns subject to axial and cyclic horizontal loads. Int J
2000;97(1):49–57.
Steel Struct 2010;10(4):385–93.
[29] Li Q, Duan Y, Wang G. Behaviour of large concrete specimens in uniaxial
[8] Muise J. Behaviour of simple framing connections to partially concrete encased
tension. Mag Concr Res 2002;54(5):385–91.
H section columns. Master’s Thesis, Dept of civil engineering, University of
[30] Marzouk H, Chen Z. Fracture energy and tension properties of high-strength
Toronto, Toronto, Canada; 2000.
concrete. J Mater Civil Eng 1995;7(2):108–16.
[9] Maranda R. Analyses par Éléments Finis de PoteauxMixtes Avec Sections
[31] CSA S16-09. Design of steel structures. Canadian Standards Association,
d’acier En I de Classe 4 Report no. EPM/GCS-1998-11. Dept of civil, geological
Mississauga, ON; 2001.
and mining engineering, Ecole Polytechnique, Montreal, Canada; 1998.
[32] CSA G40.21-04. Structural quality steel. Canadian standards association,
[10] Chicoine T, Tremblay R, Massicotte B. Finite element modelling and design of
Mississauga, ON; 2004.
partially encased composite columns. Steel Compos Struct 2002;2(3):171–94.
[33] Hibbitt, Karlsson and Sorensen, Inc. ABAQUS/Explicit User’s Manual, Version
[11] Begum M, Driver RG, Elwi AE. Finite element modeling of partially encased
6.3; 2003.
composite columns using the dynamic explicit solution method. J Struct Eng,
[34] Almusallam TH, Alsayed SH. Stress–strain relationship of normal, high
ASCE 2007;133(3):326–34.
strength and lightweight concrete. Mag Concr Res 1995;47(170):39–44.
[12] Ellobody E, Young B, Lam D. Eccentrically loaded concrete encased steel
[35] ACI Committee 363. State-of-the-art report on high-strength concrete, ACI
composite columns. Thin-Walled Struct 2011;49(1):53–65.
363R-92. Farmington Hills, MI.; 1992. p. 1–55.
[13] Desayi P, Krishnan S. Equation for stress–strain curve of concrete. ACI J
[36] Rashid MA, Mansur MA, Paramasivum P. Correlation between mechanical
1964;61(3):345–50.
properties of high strength concrete. J Mater Civil Eng, ASCE
[14] Popovics S. A numerical approach to the complete stress–strain curve of
2002;14(3):230–8.
concrete. Cem Concr Res 1973;3(4):583–99.

Anda mungkin juga menyukai