Journal of Constructional Steel Research: Binglin Lai, J.Y. Richard Liew
Journal of Constructional Steel Research: Binglin Lai, J.Y. Richard Liew
Journal of Constructional Steel Research: Binglin Lai, J.Y. Richard Liew
a r t i c l e i n f o a b s t r a c t
Article history: This paper investigates the axial force-moment (N-M) interaction behaviour of high strength concrete encased
Received 27 February 2020 steel (CES) composite columns. A numerical method is developed to simulate the nonlinear inelastic behaviour
Received in revised form 20 April 2020 of CES composite beam-columns, which includes concrete cover spalling, concrete confinement effect, buckling
Accepted 30 April 2020
of longitudinal reinforcing bars, and stain-hardening of the encased steel section. The numerical model is imple-
Available online xxxx
mented in a self-developed computer program to generate the load-deflection response, moment-curvature re-
Keywords:
lation, as well as the N-M strength interaction diagrams. The accuracy of the proposed analysis method is
Axial force-moment interaction validated by comparison with the experimental results report in the companion paper and the test data available
Concrete encased in literatures. To further understand the resistance of CES columns consisting of high strength concrete and high
Composite columns strength steel, parametric study is performed to evaluate the effect of concrete compressive strength, steel yield
High strength concrete strength, load eccentricity, as well as steel area ratio. Finally, a simplified method is proposed to construct the N-
High strength steel M strength interaction curve of CES composite columns. The proposed method can be used to design CES beam
Nonlinear inelastic behaviour columns for concrete grade up to C100 and steel grade up to S960.
© 2020 Elsevier Ltd. All rights reserved.
1. Introduction compressed high strength concrete columns may suffer from premature
cover spalling due to the different lateral expansion between concrete
Concrete Encased Steel (CES) composite columns have attracted cover and core concrete [3,16], special attention should be paid in the
substantial research attentions. Experimental work has been carried modelling of concrete core-cover interface. Current research work
out to investigate the axial performance [1–3], seismic performance indicted that the steel fiber reinforced concrete or engineered cementi-
[4–6], bond slippage performance [7,8] and the behaviour under axial tious composite (ECC) materials may be used to mitigate or prevent
force and single-curvature bending moment [9–11]. Aside from experi- concrete cover spalling [49], but these materials require more stringent
mental work, a considerable volume of numerical investigations has casting sequence and quality control on top of higher cost of such
also been conducted using finite element analysis [12,13] and fiber ele- materials.
ment analysis [14,15]. Foster et al. [16] set a lower stiffness to the concrete cover element
Ellobody et al. [12,13] employed ABAQUS software package to inves- once the predefined threshold tensile strain was reached, and it was
tigate the behaviour of CES columns under concentrically and eccentri- found that the premature cover spalling occurs when the tensile stress
cally compression. In the modelling of concrete component, the at the core-cover interface exceeded 1/4 of the uniaxial tensile resis-
confinement effect was taken into account by dividing it into highly tance at compression force corresponding to 0.88 times in-situ com-
confined, partially confined and unconfined region, and assigning re- pressive strength [16]. Another strategy to incorporate premature
spective constitutive relations in the finite element model. Good agree- spalling in finite element analysis is proposed by Piscesa et al. [17].
ment was achieved between numerical and experimental results in This strategy quantified this special issue at constitutive level and
terms of both failure mode and load-carrying capacity. Since all the the model was developed under the plasticity-fracture framework.
test data reported in [12,13] only covered concrete compressive The proposed cover spalling properties were implemented in a self-
strength within 13.7 to 35.1 MPa, and yield strength of the steel section developed in-house program (3D-NLFEA) [17], which predicts well
between 232 and 380 MPa, it remains unknown whether the model not only the first and second peak load, but also the softening
works for high strength CES columns. In addition, as the purely- behaviour.
As discussed above, finite element analysis requires massive compu-
tational effort due to the potential convergence issues and the sophisti-
⁎ Corresponding author. cated modelling at core-cover interface. As an alternative solution, fiber
E-mail address: ceeljy@nus.edu.sg (J.Y.R. Liew). element method (also termed as “Strip Element Method” [18]) is more
https://doi.org/10.1016/j.jcsr.2020.106136
0143-974X/© 2020 Elsevier Ltd. All rights reserved.
2 B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136
programmable and able to simulate the overall behaviour of composite with large eccentricity. Although the proposed method can predict the
columns with sufficient accuracy [19,20]. In the fiber element approach, test results well, it may not be applicable for specimens loaded with
the composite section is discretized as numerous fiber elements and small eccentricity or concentric compression, which may exhibit more
assigned with different material properties [10,11,14,15,19–24]. Liang brittle damage patterns.
et al. [19,20] numerically studied the fundamental behaviour of eccen- To fill the research gap, a numerical model capable of representing
trically loaded Concrete Filled Steel Tubular (CFST) beam-columns. In the characteristics of high strength CES columns is required. A nonlinear
the numerical model, the local buckling of steel tube and concrete con- fiber element analysis is employed in this paper, and effects of concrete
finement effect is incorporated, and the influences of parameters inclu- cover spalling, reinforcing bar buckling, concrete confinement, as well
sive of concrete compressive strength, steel yield strength, slenderness as strain-hardening of steel section are considered in the model. Firstly,
ratio, and width to thickness ratio were discussed in detail. CES column the experimental program is briefly reviewed and followed by the intro-
behaviour were numerically investigated by El et al. [14,15] and Kim duction of numerical procedure. After the calibration of numerical
et al. [10,11]. According to El et al. [15], the steel flange can be consid- model with the benchmark experiments, parametric study is carried
ered as a cantilever beam that deflect outward freely, so that the bend- out to investigate the effect of concrete compressive strength, steel sec-
ing moment generated at the intersection of steel flange and web can be tion yield strength, load eccentricity ratio, and steel area ratio on the ca-
computed by assuming the attainment of yield strength. Applying the pacity of CES columns. Finally, a simplified design method is developed
model developed by Mander et al. [25], the confinement effect induced to construct the N-M curve of CES columns with high strength materials.
by steel section can be computed accordingly. Kim et al. [11] extended
this model to high strength CES columns and proposed a formulae of 2. Summary of test results
equivalent ultimate strain, εcue, to account for the improvement of con-
crete deformation capacity offered by confinement effect. It appears The test results and phenomenon were discussed in detail in the
that concrete cover was included in deriving “εcue”, which may conflict companion paper [44]. The test setup and specimen configurations are
with the cover spalling phenomenon, since concrete cover would not shown in Fig. 1. In each series, one concentric compression test, two ec-
resist load after spalling off from the specimen. In addition, most of centric compression tests (e = 20/50 mm), and one four-point bending
the tests reported by Kim et al. [10,11] only covered compression load test were conducted. All specimens have the same cross-section
Fig. 1. Test setup and specimen configurations: (a). Concentric compression; (b). Eccentric compression; (c) Specimen configuration; (d). Four-point bending.
B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136 3
dimension, and fabricated by C90 concrete and S500 steel section. For The deflections of eccentrically loaded specimens at different load-
identification, each specimen is designated a string of characters and ing stages are plotted in Fig. 3. It can be seen that the deflection shape
numbers. For example, E20-SF0.5 refers to the specimens subjected to along the specimen height could be approximated as half-sine wave.
20 mm eccentric compression load and having 0.5% steel fiber. When the axial force increased from 0.8 Nu to 1.0 Nu, the specimens
Fig. 2 shows the load-displacement responses of the test specimens deflected more significantly than the previous loading stages, indicating
and their corresponding behaviour at different stages of loading. For the that when approaching the maximum load, the flexural stiffness de-
eccentrically loaded specimens, the reinforcing bars at the compression creased more rapidly due to material nonlinearities.
side reached the yield strain before the first peak load, followed by con- As shown in Fig. 4, the strain distribution of concrete is almost line-
crete cover spalling (in plain concrete specimens) or concrete longitudi- arly varied along the section, satisfying the hypothesis of plane-section
nal cracking (in fiber-reinforced specimens). In the post-peak stage, remains plain after deformation. The specimens subject to 20 mm and
yielding occurred at the compression flange of the steel section and at 50 mm load eccentricity experienced typical compression failure. As
the tensioned reinforcing bars. For the specimens under four-point indicated from Figs. 4(a) & (b), the mid-height section of specimen
bending, the rebar and steel flange at tension side yielded at early E20-SF0 was completely under compression before the failure, with
stage, but the steel flange at compression side only yielded after the the neutral axis locating outside the composite section. Very small ten-
attainment of peak load. Since compact S500 steel section was used in sile strain was observed for Specimen E50-SF0 when the maximum load
the test, all specimens exhibited excellent residual strength and post- was reached. For the beam specimen B-SF0, tensile strain generated fast
peak ductility. and the neutral axis position also shifted upward rapidly. The strain
Fig. 2. Load-deflection curves and response behaviour of high strength CES specimens.
4 B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136
gauges attached at the bottom side were damaged early due to widen- concrete, steel section and reinforcing bars. Each material component
ing of tensile cracks, so they are not presented in Fig. 4(c). contains numerous fiber elements assigned with respective material
properties. The stress of each element can be derived according to the
3. Numerical model predefined stress-strain relation discussed below, and the internal
force generated by the composite section can be computed as the stress
3.1. Fiber element method resultant of fiber elements. Several hypotheses are commonly made:
(1). Plane section remains plane after deformation, so that strain distri-
Nonlinear fiber element analysis is conducted by discretizing the bution is linear variable along the section; (2) Perfect bond between
section into numerous fiber elements. As sketched in Fig. 5 (a), the steel and concrete; (3) The tensile resistance of concrete is neglected
CES composite section consists of unconfined concrete, confined and shear deformation is also neglected.
Fig. 3. Deflection of specimens at different loading stages: (a) Eccentrically loaded specimens; (b) Beam specimens.
B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136 5
Fig. 4. Concrete strain distribution along the section height of mid-height (span) section.
From the experimental measurements reported in Fig. 3, the deflec- enclosed by the stirrups. The confinement model proposed by Razvi
tion along the column height can be assumed to shape as a half-sine and Saatcioglu [26] is adopted in analysis, which is applicable for con-
wave as depicted in Fig. 5(b), and the maximum deflection occurs at crete strength ranging from 30 to 130 MPa, and also covers both con-
the mid-height section. The curvature of section can be approximately centric and eccentric compression. However, there is still an argument
taken as the second-order derivative of deflection [19], so the curvature with regard to the concrete confinement effect offered by steel section
at mid-height section can be written as: [27,28]. Despite the availability of a few steel-induced confinement
models [27,28], all of these are derived from pure compression scenario,
π2 and the concrete grade is also limited to C70 [27]. Moreover, to the best
ϕ¼δ 2 ð1Þ
L knowledge of the authors, the steel-induced confinement effect under
eccentric compression has not been fully investigated by experiment.
If the curvature and neutral axis position are determined, the inter- Therefore, the steel induced confinement is ignored for conservative
nal axial force generated on the composite section can be computed, reason, and only the stirrup-induced confinement effect is incorporated
and the load-deflection response can also obtained once the force equi- in the numerical analysis.
librium is satisfied. The detailed algorithm is demonstrated in Fig. 6. In According to Razvi and Saatcioglu [26], the confined concrete has a
view of the fine mesh size and iterations involved in calculation, the al- bilinear descending branch that remains constant when the axial strain
gorithm is implemented in a self-developed computer program using exceeds ε20. As listed in Eqs. (2)–(8), both the strain and strength en-
MATLAB. hancement can be formulated as functions of effective confining pres-
sure fle. For the unconfined concrete, the stress-strain relation can be
3.2. Constitutive models of materials predicted by assuming the confining pressure as 0.
f cc ¼ f co þ k1 f le ð2Þ
3.2.1. Concrete materials
As sketched in Fig. 5 (a), the concrete is divided into unconfined re- k1 ¼ 6:7ð f le Þ
−0:17
ð3Þ
gion (concrete cover) outside the stirrups and the confined region
Fig. 5. Discretization of CES composite section and assumed deflection shape of a pin-pin ended column.
6 B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136
2
ε 085 ¼ ε01 þ 0:0018k3 ð8Þ
f le ¼ k2 f l ð4Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X¼ 1−½1−ðε l −εbl Þ2 ð10Þ
Pq
i¼1 ðAs f s sinα Þi
fl ¼ ð5Þ where dl is the diameter of longitudinal reinforcing bars. εbl is the buck-
sbc
ling strain equalling to concrete spalling strain εsp.
where fcc, fco refer to the confined and unconfined concrete strength; s High strength steel always presents a rounded stress-strain relation
and bc refers to stirrup spacing and concrete core dimension; k1 and k2 with no marked yield plateau [29], so the model proposed by Rasmus-
are strength enhancement coefficient and confinement effectiveness sen [36] is adopted. It should be noted that the buckling of steel section
coefficient, respectively; fs is the actual stress of lateral reinforcing bars in CES columns is not considered in this paper as it occurs at very large
when fcc is achieved. fs is calculated as follow: deformation, which may not have much practical significance unless it
sffiffiffiffiffiffiffiffiffiffi! is used in seismic region. The stress-strain formulas are represented
3 k 2 ρc by Eqs. (11) & (12).
f s ¼ Es 0:0025 þ 0:04 ≤ f yt ð6Þ
f co !n
σ σ
ε¼ þ 0:002 if σ ≤ f ys ð11Þ
where ρc is volumetric ratio of stirrups. The strains at peak stress (ε01) Es f ys
and 85% peak stress in the post-peak branch (ε085) of unconfined con-
!m
crete are estimated below. It should be noted the Eq. (7) follows the σ −f ys σ− f ys
model proposed by Légeron & Paultre [29], as it matches the material ε¼ þ εus þ εys if f ys b σ ≤ f us ð12Þ
E0:2 f us −f ys
test result most accurately.
0:4
ε 01 ¼ 0:0005ð f co Þ ð7Þ
B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136 7
where E0.2 is the tangent modulus at 0.2% proof stress. n and m are coef- To evaluate the effect of concrete cover spalling and concrete
ficients controlling the shape of stress-strain curve. The detailed equa- confinement, several numerical results of E50-SF0 are comparatively
tions can be found in [36]. plotted in Fig. 9 (a). As demonstrated in Fig. 9 (a), the black dash line
gives the best representation of the experimental result. Without
consideration of concrete confinement effect, the post-peak resis-
3.3. Calibration of numerical model tance is much lower than the experimental result. On the other
hand, ignoring concrete cover spalling would result in overestima-
3.3.1. Load-deformation response tion of the load carrying capacity. To properly simulate the sudden
Since the constitutive law of fiber-reinforced concrete varies widely unloading after reaching the first peak load and the post-peak be-
depending on the specific fiber type, and it appears that the small haviour, both the cover spalling and confinement effect should be
straight fiber does not work well when the volume content reaches incorporated.
1.0%, therefore only the plain concrete specimens is investigated in the The contribution of each material component in the load-deflection
numerical analysis. Assigning the material properties to respective curve is extracted from the numerical analysis and plotted in Fig. 9 (b). It
fiber elements, the full range load-deflection curve can be generated fol- can be seen that at the first peak load, the axial force carried by concrete
lowing the algorithm in Fig. 6. Mesh sensitivity study indicates that 0.5 cover and reinforcing bars reaches the maximum while for the core con-
mm element size is optimal to achieve the desired accuracy and compu- crete and steel section still keeps increasing. After the cover spalling, the
tational time. Reducing the mesh size does not improve the accuracy of post-peak resistance is mainly provided by the steel section and the core
prediction but require more computation effort. concrete.
8000 8000
CES column
Concrete cover
Core concrete
6000 6000 Steel section
Reinforcing bars
Load (kN)
Load (kN)
4000
4000 Cover spalling
Table 1
Specimen details in Refs. [10, 11].
Ref Specimen Cross section B × D (mm) Steel section b × h × tf × tw (mm) fc (MPa) fys (MPa) fyl (MPa) s (mm) Lo (mm) e (mm) N1st (kN) Nmax (kN)
[10] C1 260 × 260 100 × 150 × 17.6 × 17.6 94 913 525 50 2620 120 2120 2203
C2 260 × 260 100 × 150 × 17.6 × 17.6 94 913 525 50 2620 60 3752 3752
C3 260 × 260 100 × 150 × 17.6 × 17.6 94 913 525 130 2620 120 2020 2020
C4 260 × 260 100 × 150 × 17.6 × 17.6 94 913 525 50 2620 120 2072 2211
[11] C10 260 × 260 150 × 150 × 15 × 15 104 812 512 65 2620 120 2023 2329
C11 260 × 260 150 × 150 × 15 × 15 104 812 512 65 3520 120 1986 2013
Where s refers to stirrup spacing. Lo is effective length and e is load eccentricity. N1st and Nmax refer to the first peak load and maximum load obtained from test.
As plotted in Fig. 10, good agreement is achieved between numerical and the ultimate compressive strain is found to be higher [37,38]. As dis-
results and experimental results for other test specimens. For the beam covered by Hadhood et al. [38], the ultimate strain of beams generally
specimens, using spalling strain 3000 με seems to underestimate the far exceeds the ultimate strain (spalling strain) of concentrically loaded
moment-carrying capacity. It should be noted that concrete under members. To reflect this issue in the moment-curvature relation, a sen-
bending exhibits better deformability than when it is pure compression, sitively study is presented in Fig. 10 (d). Increasing the spalling strain
Fig. 11. Comparison between numerical results and experimental results reported by Kim et al. [10,11].
10 B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136
Table 2
Comparison of load-carrying capacity between numerical and experimental results.
Axial load Bending moment NNum/Ntest NNum/Ntest Axial load Flexural moment NNum/Ntest NNum/Ntest
N1st (kN) M1st (kNm) (%) (%) NMax (kN) MMax (kNm) (%) (%)
Authors E20-SF0 8026 183 105.7 103.6 8026 183 105.7 103.6
E50-SF0 6582 355 98.6 96.7 6582 355 98.6 96.7
B-SF0 – 473 – 96.8 – 473 – 96.8
Kim et al. [10] C1 1980 269 93.4 93.4 2244 376 101.9 112.0
C2 3484 251 92.8 95.0 3484 251 92.8 95.0
C3 1951 265 96.6 98.0 1951 265 96.6 98.0
C4 1970 268 95.1 95.3 2225 373 100.6 114.4
Kim et al. [11] C10 2211 301 109.3 109.1 2263 361 97.2 98.5
C11 2033 308 102.4 107.0 2050 390 101.9 107.2
Mean 99.2 99.4 99.4 102.5
Fig. 13. Evaluation of N-M curve obtained from numerical analysis and design code EN 1994-1-1 and AISC 360–16.
B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136 11
Table 3
Variation range of parameters investigated in the parametric study.
7000 7000
6000 C60 C60 C60
C80 6000 C80 6000 C80
C100 C100 C100
5000
5000 5000
Load (kN)
Load (kN)
Load (kN)
4000
4000 4000
3000
3000 3000
Section A Section A
2000 Section A 2000 2000
fys=550MPa fys=690MPa
fys=355MPa
L=2000mm L=2000mm
1000 L=2000mm 1000
e=0.2D
1000 e=0.2D
e=0.2D
0 0 0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
Deflection (mm) Deflection (mm) Deflection (mm)
Load (kN)
Load (kN)
Load (kN)
4000
3000 3000
3000
2000 Section A 2000 Section A Section A
fc=60MPa 2000
fc=80MPa fc=100MPa
1000
L=2000mm 1000 L=2000mm L=2000mm
e=0.2D e=0.2D 1000
e=0.2D
0 0 0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
Deflection (mm) Deflection (mm) Deflection (mm)
strength at the tension side. Therefore, when high strength steel section 4. Parametric study
is used, the stress development may be limited because of concrete
crushing, and the full plastic resistance of the steel section cannot be The fiber element approach has been shown to be capable of simu-
achieved. A detailed discussion on this aspect will be presented later. lating the nonlinear inelastic behaviour of high strength CES columns
9000
Section A e=0.1D e=0.1D
6000
fc=60MPa e=0.2D e=0.2D
e=0.3D 7500 e=0.3D
fys=355MPa
5000 e=0.4D e=0.4D
L=2000mm
e=0.5D e=0.5D
6000
Load (kN)
Load (kN)
4000
4500
3000
3000
2000
1000 1500
0 0
0 5 10 15 20 25 30 35 40 45 0 100 200 300 400 500
Deflection (mm) Moment (kN m)
(a) (b)
Fig. 18. Effect of load eccentricity on the load-deflection curve.
B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136 13
MPa [10], the maximum steel grade is set to be S960 in parametric the peak load is more pronounced due to the increased brittleness
study. The load eccentricity ratio ranges from 0.1 to 0.5, and the effective of high strength concrete. For S355 and S550 steels, only one peak
buckling length of column varies from 1000 mm to 5000 mm, with the load is observed, which may be attributed to the ignorance of strain-
corresponding non-dimensional slenderness ratio (λ) ranges from 0.16 hardening effect of steel section and weak confinement effect of con-
to 0.82 [39]. The three steel section sizes result in steel area ratio of crete. In the case of S690 steel, the second peak load is observed and
9.82%, 7.27% and 4.71%, and the concrete depth (ac) of 50 mm, 75 mm, increases as concrete strength increasing. Fig. 16 (c) shows that the
and 100 mm, respectively. For comparison, the reference specimen is second peak load tends to be higher than the first one when C60 and
defined to have C60 concrete, S355 steel, 0.2 load eccentricity ratio, C80 concrete are used. It is believed that lower concrete strength
2000 mm column height, and steel Section A. leads to stronger confinement effect under the same confining configu-
Since normal strength steel and high strength steel differ consider- ration [29].
ably in the strain-hardening behaviour, the elastic-perfectly plastic
stress-strain model is adopted, for simplicity, in the parametric study. 4.1.2. Effect of steel yield strength
Different from concrete, steel materials always possess constant
Young's modulus regardless of steel grade. Consequently, all the speci-
4.1. Evaluation of load-deflection relation mens exhibit the same stiffness as evident in Fig. 17. By changing the
steel section grade from S355 to S550, both the peak strength and resid-
4.1.1. Effect of concrete compressive strength ual strength increase, but the load-deflection curves of specimens with
As presented in Fig. 16, regardless of steel grade, increasing concrete S550 and S690 steel almost coincide with each other until the strength
compressive strength improves the column flexural stiffness, peak recovery after first peak load. Since the yield strength of S690 is
strength, as well as post-peak behaviour, but the strength decay after achieved after cover spalling, it plays a more important role in the
post-peak performance. In summary, using high strength steel section 4.1.4. Effect of steel area ratio
in concrete encased steel composite column may not be beneficial in The steel area ratio is defined as the ratio of steel section area to the
term of higher load carrying capacity, but it does improve the post gross area of the composite section. It is an important parameter to
peak ductility. judge whether the composite column behaves more like concrete col-
umns or steel columns. According to AISC 360–16, steel section area
shall comprise at least 1.0% of the entire CES section [40]. As shown in
4.1.3. Effect of load eccentricity ratio Fig. 19, all the specimen displayed similar load-deflection response,
All the column specimens in the parametric study are bending about but higher steel area ratio results in higher stiffness, higher load-
the major axis. As shown in Fig. 18 (a), at 0.1D eccentricity, the speci- carrying capacity, as well as better post-peak ductility. The effect of
men fails at a very small lateral deflection, followed by a sudden drop steel area ratio on the design of CES columns will be discussed in detail
of load. With the increase of load eccentricity, the strength and stiffness in the following subsection.
decrease while the deformability and ductility increase. The load-
moment relations under each load eccentricity are plotted with the N- 4.2. Evaluation of N-M curve
M curve in Fig. 18 (b). It can be seen that larger load eccentricity results
in lower ultimate load and higher bending moment. Since all the col- As mentioned before, the cross-sectional N-M curve of CES columns
umns are 2000 mm in length, with a low λ of 0.33, the second-order ef- can be constructed using the fiber element approach. To provide a com-
fect is not obvious as shown in Fig. 18 (b). prehensive assessment of existing design method in EN 1994-1-1 and
Fig. 23. Change of N-M curve shape with different steel contribution ratio.
AISC 360–16, and to develop a new design method if necessary, the ef- steel grade from S355 to S550 does not really enhance the overall resis-
fect of concrete compressive strength, steel yield strength, and steel tance in the region of N-M curve that represents small axial force and
area ratio are discussed below. large bending moment, and even no enhancement is observed when
small size steel section is used. It is also observed that the N-M curve
4.2.1. Effect of concrete compressive strength of CES Section with S690 steel section is almost coincide with that
As tabulated in Table 3, both the concrete and steel material with S550 section, suggesting that the yield strength of S690 is not effec-
have three different grades in the parametric study. Each combination tive utilized due to the failure of concrete, which is also demonstrated in
of material strength and steel area ratio is comparatively plotted in Fig. 17. Although S690 steel section offers excellent post-peak resistance
Fig. 20. It can be seen that regardless of steel grade, increasing concrete and ductility to the CES column, it may not be a preferable option when
grade from C60 to C100 results in enhanced cross-section capacity. It strength rather than ductility is of priority in design, especially in non-
is interesting to discover that when S335 steel is used, the axial strength seismic regions.
enhancement is significant while the flexural strength enhan-
cement seems limited, which is consistent with the findings by
Ahmed et al. [21] in CFST columns, but when higher steel grade is 4.2.3. Effect of steel area ratio
used, the flexural strength enhancement is also significant as indicated The effect of steel area ratio on the N-M curve is evident in Fig. 22.
from Fig. 20 (c), (f), (i). Regardless of concrete and steel grade, increasing the steel area ratio
tends to expand the N-M curve outward with more improvement of
flexural strength instead of axial strength. The shape of N-M curve
4.2.2. Effect of steel yield strength also changes accordingly. When small size steel section is used, the
Different from the effect of concrete compressive strength, higher CES columns behave more like RC columns, and the N-M curve comes
steel yield strength may not have major contribute to the cross- with a marked balance point (see Fig. 23) characterized by maximum
section capacity. The comparison study in Fig. 21 shows that increasing bending capacity. However, as the steel area increases, the steel section
contributes more to the column capacity, and the shape of N-M curve is
consequently more close to that of pure steel column with the maxi-
mum moment capacity occurring at pure bending scenario [40]. Similar
discoveries have been reported by Lai et al. [41,42], Choi et al. [45,46],
and Hang & Yang [47] for Concrete Filled Steel Tubes (CFST) columns.
Fig. 24. Proposed method to construct N-M interaction curve. Fig. 25. Plastic stress distribution of CES section at pure bending.
B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136 17
RM (Numerical)
RM (Numerical)
RM (Numerical)
fc=100MPa fc=100MPa fc=100MPa
5. Design recommendation generate N-M curve similar to RC column and steel column, respectively,
but they cannot provide a unified estimation of the N-M curve shape. To
5.1. General shape of N-M curve this end, it is necessary to propose a new method that gives proper rep-
resentation of N-M curve regardless of the steel contribution ratio.
As concluded from the parametric study, the cross-section capacity It should be noted that a unified approach has been proposed by the
of high strength CES column is highly associate with concrete compres- authors [18] previously to evaluate the N-M curve of CES columns,
sive strength and steel area ratio, and it is less affected by the section's which is based on elastic-plastic stress distribution over the steel sec-
yield strength. Consequently, the shape of N-M curve also varies de- tion, and appropriate stress block parameters are used to simplify the
pending on the specific geometric and material properties. Steel contri- concrete contribution. Due to the irregular geometry of the concrete
bution ratio (δ) combines the effects of concrete compressive strength, net area in CES section, this method needs some computational effort al-
steel yield strength, and steel area ratio, so it is selected here for the though it predicts the section resistance with good accuracy. In view of
evaluation of N-M curve shape. As formulated in Eq. (13), δ is taken as this, a simplified method is proposed herein, which is schematically il-
the ratio of axial resistance of the steel section to the squash load of lustrated in Fig. 24.
the entire composite section [39]. The exact N-M curve can be approximated as a bilinear curve A-C-B
in Fig. 24. Point A and B represent the axial compression capacity and
As f ys pure bending capacity, while Point C determines the shape of N-M
δ¼ ð13Þ
0:85Ac f c þ As f ys þ Al f yl curve and moves with the change of steel contribution ratio
[41,42,45,46]. Therefore, the N-M curve equations can be formulated
below. Substituting α = 0.9 and β = 0.2 in Eqs. (14) & (15) results in
The shapes of N-M curves generated in parametric study are sum-
the same equation given by AISC 360–16, which is now considered as
marized in Fig. 23 (a), since all specimens have different cross-
a special case of the proposed method.
section capacity, only the normalized N-M curve is presented. Nn
and M n refers to the capacity corresponding to pure compression N N 1−β M
When ≥β þ ¼1 ð14Þ
and pure bending, respectively, which will be discussed in detail in Nn Nn α Mn
Section 5.2. As demonstrated in Fig. 23 (a), when the steel contribu-
tion ratio δ increases from 0.16 to 0.45, the N-M curve shrinks in-
ward with less pronounced balance point. When CES column is
reinforced with δ = 0.16, the moment capacity at balance point
even reaches more than 1.4 times of that at pure bending. But
when δ increases to 0.45, the N-M curve comes with no balance 1.1
point and behaves like a steel column.
+10%
To summarize, the shape of N-M curve is pertinent to the steel con- 1.0
tribution ratio δ. When δ is increasing, the N-M curve of CES column is
similar to that of pure steel members, and maximum moment resis- 0.9
RM (Proposed)
tance occurs at the point of pure bending. On the other hand, as concrete
contribution becomes dominant, it behaves more like RC members. For 0.8
RC members, in the presence of certain compression force, concrete -10%
cracking can be delayed and moment capacity improves accordingly,
0.7
and thus the maximum moment resistance occurs when the section is
pre-compressed by a certain amount of axial force, which is termed as
balance point in Fig. 23 (b). 0.6 RM(Proposed) Mean=0.978
RM(Numerical) St.dev=0.031
N 1−α N M
When bβ þ ¼1 ð15Þ 1:0−ð f c −50Þ=200 50 N=mm2 bf c ≤90 N=mm2
Nn β Nn Mn η¼ ð18Þ
0:8 f c N90 N=mm2
5.2.1. Point A In addition, material compatibility between steel and concrete also
For CES columns with high strength concrete, the axial compression needs to be satisfied. Based on the criterion given by Lai et al. [43], if
capacity is affected by premature cover spalling [2,3,43]. The equations the steel yield strength is not achieved before concrete crushing, the ef-
proposed by Lai et al. [43] have been confirmed to predict the axial ca- fective strength of steel, instead of yield strength, should be used in de-
pacity with good accuracy. As listed in Eqs. (16)–(18), the concrete sign. fys′ and fyl′ refer to the effective strength of high strength steel and
strength reduction factor αcc is incorporated as a product of 0.85 and rebar, which can be determined based on the peak strain of concrete
η, which accounts for the reduction of concrete strength due to prema- and Young's modulus of steel as below [43]:
ture spalling. As the premature cover spalling effect would gradually re- 0
duce as the axial force reduces and moment increases. Connecting point f ys ¼ Es ε co ≤ f ys ð19Þ
A and Point C in Fig. 24 using straight line reflects this gradual reduction.
0
0 0 f yl ¼ El εco ≤ f yl ð20Þ
Nn ¼ α cc Ac f c þ As f ys þ Al f yl ð16Þ
α cc ¼ 0:85η ð17Þ where Es and El refer to the Young's modulus of steel section and
longitudinal reinforcing bars.
Fig. 28. Calibration of proposed method against the experimental and numerical results.
B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136 19
where Wps, Wpc and Wpl refers to the plastic modulus of steel section,
concrete and longitudinal reinforcing bars. Wpsn, Wpcn and Wpln refers
to the plastic modulus of steel section, concrete and longitudinal rein-
Calculate flexural capacity
forcing bars within the region of 2hn as illustrated in Fig. 25. MActual (Point C) using Eq. (21)
equals to Mn in Fig. 24.
RM can be expressed as a function of several parameters. All
the values of R M generated in parametric study are plotted in
Fig. 26. It can be seen that RM experiences significant reduction Calculate shape factor α, β to
as steel yield strength increases from 355 MPa to 960 MPa. For get Point B using Eqs. (25-26)
S355 steel, RM is almost 1.0, implying that full plastic strength
can be achieved. It is noteworthy that RM sometimes even ex-
ceeds 1.0, which may be attributed to the usage of 0.85fc in de-
sign code for conservative purpose (see Fig. 25), while 1.0fc is
Construct simplified N-M
used in fiber element analysis. For S960 steel, R M drops to
around 0.6, suggesting that full plastic design overestimates
interaction curve (A-C-B)
the actual moment capacity. Except for steel yield strength, con-
crete compressive strength fc and concrete thickness ratio ac /h Fig. 29. Flowchart of generating N-M interaction curve using the proposed method.
also affect RM. It can be seen from Fig. 26 that RM tends to in-
crease with the increase of fc and ac/h. Fig. 28, the proposed method gives conservative and reasonable pre-
Based on the regression analysis of both numerical and experi- diction of the actual response of high strength CES columns. High
mental data, a semi-empirical formula of RM is derived in Eqs. (23) level accuracy is achieved in terms of axial capacity, flexural capac-
& (24), which is taken as a function of fys, fc and ac/h. Among these ity, as well as the capacity with coupled compression and bending.
three variables, fys is the most predominant one affecting the value Therefore, this method can be employed as an efficient tool for prac-
of RM. titioners to approximate the N-M interaction curve of CES columns
with concrete grade up to C100, and steel grade up to S960. How-
−0:54
RM ¼ 1−kð f c Þ f ys −355 ð23Þ ever, given the fact that high strength steel may not be able to de-
velop the full plastic strength due to concrete failure, the
ac maximum steel grade is recommended to be capped at S550, beyond
k ¼ 0:0063−0:0013 ð24Þ
h which the capacity may not be enhanced as concluded from the
parametric study.
The accuracy of the proposed equation is verified in Fig. 27. It A step-by-step flowchart is given in Fig. 29 to summarize the proce-
can be seen that the predicted RM agrees well with the numerical dure of generating N-M curve using the proposed method. It should be
results. noted that this is a unified method that works for concrete Class up to
C100 and steel Grade up to S960 of any combinations. However, for
5.2.3. Point C the plastic design method that uses plastic stress blocks to evaluate
Point C can be related to Point A and Point B once the ordinate (β) the cross section resistance, e.g. EN 1994-1-1, the material compatibility
and abscissa (α) are determined. As mentioned before, steel contribu- of high-strength steel and high-strength concrete should be satisfied.
tion ratio δ controls the overall shape of N-M curve, so α and β can be
also considered as a function of δ. Based on the parametric study, the
N-M curve of high strength CES columns resembles that of steel column
6. Conclusions
when δ exceeds 0.45. In this case, α and β is recommended to keep as 0.9
and 0.2 as recommended in current AISC 360–16. Below 0.45, α and β
A nonlinear inelastic analysis method using a fiber element
are derived as Eq. (26) based on regression analysis.
model is proposed to study the axial force-moment interaction be-
When δN0:45 α ¼ 0:9 haviour of high strength concrete encased composite (CES) sec-
ð25Þ
β ¼ 0:2 tions. The proposed numerical model is capable of capturing the
important attributes that affect the behaviour of high strength
When δ≤0:45 α ¼ 0:82δ−0:28 ð26Þ
CES column including concrete cover spalling, concrete confine-
β ¼ 0:13δ−0:61 ment, longitudinal bars buckling, and strain-hardening of the em-
bedded steel section. Good agreement was observed between
The predicted bilinear N-M curve is compared with the test re- numerical and experimental results in terms of the flexural stiff-
sults reported in [10,11,44] and the N-M curve generating using ness, first peak load, second peak load, and load displacement
existing EN 1994-1-1 and AISC 360–16 method. As plotted in the behaviour.
20 B. Lai, J.Y.R. Liew / Journal of Constructional Steel Research 170 (2020) 106136
From the parametric study conducted using the fiber element [18] B.L. Lai, J.Y.R. Liew, A. Le Hoang, M. Xiong, A unified approach to evaluate axial force-
moment interaction curves of concrete encased steel composite columns, Eng.
analysis, the use of higher strength steel section, such as grades Struct. 201 (2019) 109841.
S690 and S960, is not effective in enhancing the cross-section ca- [19] Q.Q. Liang, High strength circular concrete-filled steel tubular slender beam–col-
pacity of CES composite section. In addition, the shape of N-M umns, part I: numerical analysis, J. Constr. Steel Res. 67 (2) (2011) 164–171.
curve varies depending on the contribution of steel section. [20] Q.Q. Liang, High strength circular concrete-filled steel tubular slender beam–columns,
part II: fundamental behaviour, J. Constr. Steel Res. 67 (2) (2011) 172–180.
A simplified design method is proposed to construct N-M strength [21] M. Ahmed, Q.Q. Liang, V.I. Patel, M.N. Hadi, Behaviour of eccentrically loaded double
interaction curve for high strength CES columns. Good agreement is circular steel tubular short columns filled with concrete, Eng. Struct. 201 (2019)
achieved between predicted result and test results. The proposed 109790.
[22] V.I. Patel, Q.Q. Liang, M.N. Hadi, Numerical simulations of circular high strength con-
method requires three feature points to construct the N-M interaction
crete-filled aluminum tubular short columns incorporating new concrete confine-
curve, i.e., one representing pure compression capacity, one represen- ment model, Thin-Walled Struct. 147 (2020) 106492.
ting pure bending capacity, and an additional point controlling the [23] G.M. Kamil, Q.Q. Liang, M.N. Hadi, Fiber element simulation of interaction behaviour
shape of N-M curve. When the steel contribution ratio reaches 0.45 of local and global buckling in axially loaded rectangular concrete-filled steel tubular
slender columns under fire exposure, Thin-Walled Struct. 145 (2019) 106403.
and above, the N-M shape can be approximated by the equations sug-
[24] C.S. Kim, H.J. Hwang, Numerical investigation on load-carrying capacity of high-
gested in AISC 360–16. Otherwise, the proposed equation should strength concrete-encased steel angle columns, Int. J. Concr. Struct. Mater. 12 (1)
be used. (2018) 11.
Although the proposed method is applicable for CES columns with [25] J.B. Mander, M.J. Priestley, R. Park, Theoretical stress-strain model for confined con-
crete, J. Struct. Eng. 114 (8) (1988) 1804–1826.
concrete grade up to C100 and steel grade up to S960, the steel section [26] S. Razvi, M. Saatcioglu, Confinement model for high-strength concrete, J. Struct. Eng.
grade should be capped at S550 to ensure optimum use of the concrete 125 (3) (1999) 281–289.
and steel material strengths. [27] X. Zhao, F. Wen, T.M. Chan, S. Cao, Theoretical stress–strain model for concrete in
steel-reinforced concrete columns, J. Struct. Eng. 145 (4) (2019), 04019009.
[28] S. Chen, P. Wu, Analytical model for predicting axial compressive behaviour of steel
Acknowledgement reinforced concrete column, J. Constr. Steel Res. 128 (2017) 649–660.
[29] F. Légeron, P. Paultre, Uniaxial confinement model for normal-and high-strength
The present research work is funded by Building & Construction concrete columns, J. Struct. Eng. 129 (2) (2003) 241–252.
Authority in Singapore under project grant R-302-000-168-490. JFE [30] S.J. Foster, On behaviour of high-strength concrete columns: cover spalling, steel fi-
bers, and ductility, Struct. J. 98 (4) (2001) 583–589.
Steel Corporation and Yongnam Engineering & Construction are also [31] S. Bae, O. Bayrak, Early cover spalling in high-strength concrete columns, J. Struct.
gratefully acknowledged for supply and fabrication of the test Eng. 129 (3) (2003) 314–323.
specimens. [32] S.A. Sheikh, S.M. Uzumeri, Strength and ductility of tied concrete columns, J. Struct.
Div. 106 (1980)(ASCE 15388 Proceeding).
[33] A.H. Zubydan, A.I. ElSabbagh, Monotonic and cyclic behaviour of concrete-filled
References steel-tube beam-columns considering local buckling effect, Thin-Walled Struct. 49
(4) (2011) 465–481.
[1] W.Q. Zhu, G. Meng, J.Q. Jia, Experimental studies on axial load performance of high-
[34] C.C. Chen, N.J. Lin, Analytical model for predicting axial capacity and behaviour of
strength concrete short columns, Proc. Inst. Civil Eng. Struct. Build. 167 (9) (2014)
concrete encased steel composite stub columns, J. Constr. Steel Res. 62 (5) (2006)
509–519.
424–433.
[2] B.L. Lai, J.Y.R. Liew, A. Le Hoang, Behaviour of high strength concrete encased steel
[35] S. Morino, C. Matsui, S. Yoshikai, Local buckling of steel elements in concrete
composite stub columns with C130 concrete and S690 steel, Eng. Struct. 200
encased columns, Proceedings of the Pacific Structural Steel Conference. PSSC, Auck-
(2019) 109743.
land, New Zealand, 2, 1986, pp. 319–335.
[3] B.L. Lai, J.Y.R. Liew, M. Xiong, Experimental study on high strength concrete encased
steel composite short columns, Constr. Build. Mater. 228 (2019) 116640. [36] K.J. Rasmussen, Full-range stress–strain curves for stainless steel alloys, J. Constr.
[4] Y. Yang, Y. Xue, Y. Yu, F. Gao, Experimental study on seismic performance of partially Steel Res. 59 (1) (2003) 47–61.
precast steel reinforced concrete columns, Eng. Struct. 175 (2018) 63–75. [37] R.B. Khadiranaikar, M.M. Awati, Concrete stress distribution factors for high-perfor-
[5] W. Zhu, J. Jia, J. Gao, F. Zhang, Experimental study on steel reinforced high-strength mance concrete, J. Struct. Eng. 138 (3) (2011) 402–415.
concrete columns under cyclic lateral force and constant axial load, Eng. Struct. 125 [38] A. Hadhood, H.M. Mohamed, B. Benmokrane, Experimental study of circular high-
(2016) 191–204. strength concrete columns reinforced with GFRP bars and spirals under concentric
[6] Y. Xue, Y. Yang, Y. Yu, Cyclic behaviour of partially precast steel reinforced concrete and eccentric loading, J. Compos. Constr. 21 (2) (2016), 04016078.
short columns: experiment and theoretical analysis, Eng. Struct. 199 (2019) 109658. [39] . EN 1994-1-1, Eurocode 4: Design of Composite Steel and Concrete Structures-Part
[7] X. Wang, Y. Liu, F. Yang, Y. Lu, X. Li, Effect of concrete cover on the bond-slip behav- 1–1, General Rules and Rules for Buildings, 2004.
iour between steel section and concrete in SRC structures, Constr. Build. Mater. 229 [40] ANSI/AISC 360–-16, Specification for Structural Steel Buildings, American Institute
(2019) 116855. of Steel Construction (AISC), Chicago, USA, 2016.
[8] K. Wu, F. Chen, C. Chen, H. Zheng, J. Xu, Load-transfer mechanism and bond-stress [41] Z. Lai, A.H. Varma, K. Zhang, Noncompact and slender rectangular CFT members: ex-
components in steel and steel Fiber–reinforced concrete structure, J. Struct. Eng. perimental database, analysis, and design, J. Constr. Steel Res. 101 (2014) 455–468.
145 (12) (2019), 04019160. [42] Z. Lai, A.H. Varma, L.G. Griffis, Analysis and design of noncompact and slender CFT
[9] Y. Yang, Y. Chen, W. Zhang, S. Feng, Behaviour of partially precast steel reinforced beam-columns, J. Struct. Eng. 142 (1) (2015), 04015097.
concrete columns under eccentric loading, Eng. Struct. 197 (2019) 109429. [43] B.L. Lai, J.Y.R. Liew, A. Venkateshwaran, S. Li, M. Xiong, Assessment of high-strength
[10] C.S. Kim, H.G. Park, K.S. Chung, I.R. Choi, Eccentric axial load testing for concrete- concrete encased steel composite columns subject to axial compression, J. Constr.
encased steel columns using 800 MPa steel and 100 MPa concrete, J. Struct. Eng. Steel Res. 164 (2020) 105765.
138 (8) (2012) 1019–1031.
[44] B.L. Lai, J.Y.R. Liew, Axial-moment interaction of high strength concrete encased
[11] C.S. Kim, H.G. Park, K.S. Chung, I.R. Choi, Eccentric axial load capacity of high-
steel composite columns: experimental investigation, J. Constr. Steel Res. (2020)
strength steel-concrete composite columns of various sectional shapes, J. Struct.
(Submitted for review).
Eng. 140 (4) (2014), 04013091.
[45] Y.H. Choi, D.A. Foutch, J.M. LaFave, New approach to AISC P–M interaction curve for
[12] E. Ellobody, B. Young, Numerical simulation of concrete encased steel composite col-
square concrete filled tube (CFT) beam–columns, Eng. Struct. 28 (11) (2006)
umns, J. Constr. Steel Res. 67 (2) (2011) 211–222.
1586–1598.
[13] E. Ellobody, B. Young, D. Lam, Eccentrically loaded concrete encased steel composite
columns, Thin-Walled Struct. 49 (1) (2011) 53–65. [46] Y.H. Choi, K.S. Kim, S.M. Choi, Simplified P–M interaction curve for square steel tube
[14] S. El-Tawil, C.F. Sanz-Picon, G.G. Deierlein, Evaluation of ACI 318 and AISC (LRFD) filled with high-strength concrete, Thin-Walled Struct. 46 (5) (2008) 506–515.
strength provisions for composite beam-columns, J. Constr. Steel Res. 34 (1) [47] L.H. Han, Y. Yang, The State-of-the-Art-Technology of Concrete Filled Steel Tubes,
(1995) 103–123. China Architecture and Press, Beijing, 2007.
[15] S. El-Tawil, G.G. Deierlein, Strength and ductility of concrete encased composite col- [48] H.J. Lee, H.G. Park, I.R. Choi, Eccentric compression behaviour of concrete-encased-
umns, J. Struct. Eng. 125 (9) (1999) 1009–1019. and-filled steel tube columns with high-strength circular steel tube, Thin-Walled
[16] S.J. Foster, J. Liu, S.A. Sheikh, Cover spalling in HSC columns loaded in concentric Struct. 144 (2019) 106339.
compression, J. Struct. Eng. 124 (12) (1998) 1431–1437. [49] M.K. Khan, M.M. Rana, Y.X. Zhang, C.K. Lee, Compressive behaviour of engineered
[17] B. Piscesa, M.M. Attard, D. Prasetya, A.K. Samani, Modeling cover spalling behaviour cementitious composites and concrete encased steel composite columns, J. Constr.
in high strength reinforced concrete columns using a plasticity-fracture model, Eng. Steel Res. 167 (2020) 105967.
Struct. 196 (2019) 109336.