See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/263857151

Noncompact and slender rectangular CFT members: Experimental database,

analysis, and design

Article  in  Journal of Constructional Steel Research · October 2014

DOI: 10.1016/j.jcsr.2014.06.004

CITATIONS READS

49 247

3 authors:

Zhichao Lai Amit H Varma

Fuzhou University Purdue University
40 PUBLICATIONS   305 CITATIONS    295 PUBLICATIONS   2,148 CITATIONS   

SEE PROFILE SEE PROFILE

Kai Zhang
Westinghouse Electric Company, LLC
11 PUBLICATIONS   213 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Light Water Reactor Sustainability, US DOE View project

Interface mechanics View project

All content following this page was uploaded by Amit H Varma on 07 June 2020.

The user has requested enhancement of the downloaded file.

 Journal of Constructional Steel Research 101 (2014) 455–468

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Noncompact and slender rectangular CFT members: Experimental

database, analysis, and design
Zhichao Lai ⁎, Amit H. Varma, Kai Zhang
Purdue University, School of Civil Eng., West Lafayette, IN, United States

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

Article history: Rectangular concrete-filled steel tube (CFT) members are categorized as compact, noncompact or slender de-
Received 19 March 2014 pending on the slenderness ratio (width-to-thickness b/t ratio) of the steel tube walls. International design
Accepted 6 June 2014 codes typically focus on the design of compact CFT members with relatively small slenderness (b/t) ratios. The
Available online 10 July 2014
behavior and design of noncompact or slender CFT members is not addressed directly. This paper presents the
basis of the current AISC Specification (AISC 360-10) for the design of noncompact or slender rectangular CFT
Keywords:
Slender
members under axial compression, flexure, and combined axial and flexural loading. The experimental database
Noncompact of tests conducted on noncompact and slender CFT members is reviewed. Design equations are developed based
Composite columns on the experimental results and observations. Detailed 3D finite element method (FEM) models are developed
Finite element analysis for noncompact and slender CFT members, and benchmarked using experimental results. The benchmarked
Design models are used to address gaps in the experimental database, and further verify the conservatism of the design
Local buckling equations.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction
CFT members consist of rectangular or circular steel tubes filled with
concrete. These composite members optimize the use of both steel and
concrete construction materials as compared to steel or reinforced con-
crete structures. The concrete infill delays the local buckling of the steel
tube, while the steel tube provides confinement to the concrete infill.
The behavior of CFT members under axial loading, flexure, and com-
bined axial and flexural loading can be more efficient than that of struc-
tural steel or reinforced concrete members. Moreover, the steel tube
serves as formwork for placing the concrete, which facilitates and expe-
dites construction while reducing labor costs.
CFT members are used widely around the world in various types of
structures. For example, CFT members are used as columns in composite
braced frames in: (i) the Two Union Square building in Seattle,
Washington, (ii) Casselden Place project in Melbourne, Australia, and
(iii) Taipei 101 tower in Taipei, Taiwan. CFT members are also used as
columns in composite moment frames, for example in: (i) 3 Houston
Center in Houston, Texas, and (ii) Postal Office building in Quanzhou,
China. CFT members are used as compression chords in composite brid-
ges, for example, in: (i) the Xialaoxi bridge in Yichang, China, (ii) Jinan
East Railway Station bridge in Jinan, China, and (iii) Pudong Canal bridge
⁎ Corresponding author at: Ph.D. Candidate, Purdue Univ., School of Civil Eng., West
Lafayette, IN 47906, USA.
E-mail addresses: laiz@purdue.edu (Z. Lai), ahvarma@purdue.edu (A.H. Varma),
kai-zh@purdue.edu (K. Zhang).
in Shanghai, China. CFT members are also used as piles, transmission
towers, and bracing members in buckling restrained frames.
2. Background
Since the first documented experimental research on CFT columns in
1957 [1], significant research has been conducted to investigate the be-
havior of CFT members under various loading conditions. For example:
(i) axial compression tests, (ii) flexural tests, and (iii) combined axial
force and flexure (beam-column) tests have been conducted by re-
searchers in various countries. These studies indicate that the strength
of CFT members depends on several parameters, namely, the steel
yield stress Fy, concrete compressive strength f'c, tube wall slenderness
(b/t) ratio, column length to depth ratio L/h and composite interaction
between the steel tube and concrete infill, etc.
Nishiyama et al. [2], Kim [3], Gourley et al. [4], and Hajjar [5] have in-
dependently compiled comprehensive databases of experimental re-
search conducted on rectangular and circular CFTs. The database
compiled by Hajjar [5] (previously, Gourley et al. [4]) is the most com-
prehensive database of experimental and numerical research per-
formed on CFT members, frames, and systems. The database includes
all the tests conducted on compact, non-compact, and slender CFT
members with a wide range of material, geometric, and loading param-
eters. However, a significant portion of the database is comprised of
tests conducted on compact CFT members. There are fewer, but reason-
able number of tests conducted on noncompact and slender CFT mem-
bers, which are the focus of this paper and research.

http://dx.doi.org/10.1016/j.jcsr.2014.06.004
0143-974X/© 2014 Elsevier Ltd. All rights reserved.
456 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468
Most design codes specify steel tube slenderness (b/t) ratio limits for
CFT members. For example, Eurocode 4 (2004) [6] specifies that the
steel tube of rectangular CFT columns in compression should satisfy
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the limit, b=t ≤52 235=F y , where Fy is in MPa, to prevent the local
buckling. AS 4100 (2012) [7] permits the occurrence of steel tube local
buckling, and provides an effective width method to calculate the
axial strength of slender CFT members. The Japanese code (AIJ 2008
[8]) classifies rectangular CFTs into three types, i.e., FA, FC, and FD de-
pending on the steel tube slenderness ratio. CFT columns classified as
FC and FD have larger steel tube slenderness ratios and are susceptible
to local buckling effects. AIJ 2008 [8] provides an axial load capacity fac-
tor to account for the effects of steel tube slenderness (and local buck-
ling) on the axial strength of rectangular CFTs.
Eurocode 4 (2004) [6] specifies that the flexural strength of CFT
members can be calculated as the plastic moment resistance over the
composite cross-section while using: (i) the yield stress (Fy) for steel
in compression or in tension, (ii) the compressive strength (f 'c) for con-
crete in compression, and (iii) neglecting the contribution of concrete in
tension. The Australian and Japanese codes (AS 4100 [7] and AIJ 2008
[8]) specify tube slenderness ratio dependent stress–strain curves for
steel in compression that can be used to calculate the flexural strength
of rectangular CFT members.
None of these international codes specify tube slenderness (b/t)
ratio limits to classify rectangular CFT members into noncompact or
slender CFTs. They also do not have different slenderness (b/t) ratio
limits for rectangular CFTs subjected to different loading conditions
(axial or flexural loading). The AISC 360-05 [9] specification also speci-
fied the tube slenderness (b/t) ratio limits only for compact CFTs, and
did not include any provisions for classifying or calculating the strength
of noncompact or slender CFTs subjected to different loading conditions
(axial or flexural loading). As a result, the design and use of noncompact
or slender CFT members in the US was limited in scope.
This paper presents the development of the AISC 360-10 [10] speci-
fication that includes provisions for classifying and calculating the
strength of noncompact and slender CFTs subjected to different loading
conditions. It is based on the work done earlier by the authors (refer-
enced in AISC 360-10 [10]), and enhanced herein to further confirm
the conservatism of the design provisions. This paper focuses on rectan-
gular CFT members. The development of the AISC 360-10 specification
for (noncompact and slender) circular CFT members and the confirma-
tion of their conservatism is presented elsewhere in [11].
The outline of this paper is as follows. The paper first presents the
AISC 360-10 [10] slenderness (b/t) ratio limits used to classify rectangu-
lar CFT members as compact, noncompact, or slender depending on the
Fig. 1. Effects of concrete infill on the local buckling of hollow tubes.
loading (axial or flexural). It discusses the basis and reasoning for these
limits, and the associated strength equations in AISC 360-10 [10]. The
experimental database of tests conducted on noncompact and slender
rectangular CFT tests subjected to different loading conditions (axial
compression, flexure, combined axial force and flexure) are presented.
The conservatism of the AISC 360-10 [10] design equations are
established by using them to predict the strength of CFT members in
the experimental database. 3D nonlinear inelastic finite element models
are developed and benchmarked for predicting the behavior and
strength of noncompact and slender CFTs in the experimental database.
The benchmarked models are used to address gaps in the experimental
database, and further confirm the conservatism of the AISC 360-10 [10]
design equations.
3. Slenderness limits for rectangular CFT
members—axial compression
The behavior of CFT members is fundamentally different from that of
hollow structural shape (HSS) members. The concrete infill changes the
buckling mode of the steel tube by preventing it from buckling inward,
as shown in Fig. 1. The post-buckling behavior of CFT members is more
ductile than that of equivalent HSS members due to the larger wave-
length of the buckling mode, spreading of plastic deformation, and
slight increase in the moment of inertia of the steel tube due to the
outward buckling shape. The elastic local buckling behavior of the
steel tube walls of rectangular CFT members subjected to axial compres-
sion was investigated analytically by Bradford et al. [12] using the
Rayleigh–Ritz method. The assumed local buckling mode shape
accounted for the effects of concrete infill, i.e., no inward displacements
as shown in Fig. 1.
The resulting equation for local buckling is shown in Eq. (1). In this
equation, Fcr is the critical stress for elastic local buckling, Es is the mod-
ulus of elasticity of the steel tube, ν is the Poisson's ratio for steel, and b/t
is the governing (larger) slenderness ratio. The parameter k accounts for
the local buckling mode. Bradford et al. [12] showed that k was equal to
10.6 for the mode shape shown in Fig. 1. The critical buckling stress Fcr
simplifies to 9.6Es/(b/t)2 after substituting the values of k equal to
10.6, and Poisson's ratio for steel equal to 0.3. The critical buckling stress
(Fcr) reaches the yield stress (Fy) when the slenderness ratio (b/t) be-
pffiffiffiffiffiffiffiffiffiffiffiffi
comes equal to 3:10 Es =F y .
kπ2 Es
F cr ¼   2 ð1Þ
12 1−v2 bt
 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468 457

0
P no ¼ P p ¼ As F y þ 0:85 f c Ac ð2Þ

Previous researchers [13] used the existing experimental database to

show that when the tube wall slenderness (b/t) ratio is less than or
pffiffiffiffiffiffiffiffiffiffiffiffi
equal to 2:26 Es =F y , the axial compressive strength of rectangular
CFT stub (short) columns could be calculated conservatively using
Eq. (2), which consists of superposition of the yield strength of the
steel tube and compressive strength of the concrete infill. Therefore,
pffiffiffiffiffiffiffiffiffiffiffiffi
this slenderness ratio, 2:26 Es =F y, was established as the compactness
limit (λp) in AISC 360-05 [9]. Rectangular CFTs with slenderness (b/t) ra-
tios less than or equal to λp were classified as compact sections, and the
use of CFTs with slenderness ratios greater than λp was not permitted
up to 2010. CFT columns with compact sections can develop yielding
before local buckling and provide adequate confinement of the concrete
infill to develop its compressive strength up to 0.85f'c.
In AISC 360-10 [10], pthe noncompactness limit (λr) was established
ffiffiffiffiffiffiffiffiffiffiffiffi
conservatively as 3:00 Es =F y based on Bradford et al. [12] as explained
above. Rectangular
pffiffiffiffiffiffiffiffiffiffiffiffi CFTs with steel tube slenderness ratio greater than
pffiffiffiffiffiffiffiffiffiffiffiffi
λp (2:26 Es =F y) but less than or equal to λr (3:00 Es =F y) are classified
as noncompact. Noncompact CFT sections can reach the yield stress (Fy)
of the steel tube with local buckling, but cannot provide adequate con-
finement to the concrete infill to reach its full compressive strength.
The concrete infill has significant volumetric dilation after the compres-
sive stress exceeds 0.70f'c [14]. The volumetric dilation of concrete infill
cannot be confined adequately by the noncompact steel tube undergo-
ing local buckling. As a result, the compressive strength of rectangular
CFT stub (short) columns with noncompact steel tubes (slenderness Fig. 2. Variation of: (a) nominal axial compressive strength and (b) nominal flexural
ratio = λr) is limited to that calculated using Eq. (3). strength with respect to tube slenderness ratio.

0
P no ¼ P y ¼ As F y þ 0:70 f c Ac
Rectangular
pffiffiffiffiffiffiffiffiffiffiffiffi CFTs with steel tube slenderness ratio greater than λr ¼
3:00 Es = F y are classified as slender. Slender CFT sections undergo elas-
tic local buckling, and the critical buckling stress Fcr can be calculated
conservatively using Eq. (4), which is based on Bradford et al. [12] as ex-
plained above. The post-local buckling behavior of the steel tube is
constrained by the concrete infill while it is in the elastic range. How-
ever, the concrete has significant volumetric dilation after the compres-
sive stress exceeds 0.70f'c. This volumetric dilation exacerbates the post
local buckling deterioration of the steel tube, which cannot confine it as
well. Therefore, the compressive strength of rectangular CFT stub
(short) columns with slender steel tubes (slenderness ratio N λr) is lim-
ited to that calculated using Eq. (5).
ð3Þ λr), and the maximum permitted slenderness ratio (λlimit) for steel
tubes of rectangular CFTs under axial compression.
Fig. 2(a) shows a graphical representation of the nominal axial com-
pressive strength (Pno) of rectangular CFT stub (short) columns as a
function of the tube slenderness (b/t) ratio. As shown, the nominal
axial compressive strength (Pno) of noncompact rectangular CFT stub
(short) columns (with slenderness ratio λ between λp and λr) can be
calculated using Eq. (6), which has a quadratic variation between Pp
(Eq. (2)) and Py (Eq. (3)) with slenderness ratio (λ = b/t). This nonlin-
ear variation accounts for the fact that steel tubes need to be closer to
the compactness limit λp to confine the concrete, and increase its com-
pressive strength contribution from 0.70f'c at Py to 0.85f'c at Pp. A linear
variation between Py and Pp was considered initially, but found to be
unconservative for some of the experimental results.

9Es
F cr ¼ ð4Þ P p −P y  2
ðb=t Þ2 P no ¼ P p −  2 λ−λp ð6Þ
λr −λp

0
P no ¼ P cr ¼ F cr As þ 0:70 f c Ac
The use of slender rectangular CFTs with tube slenderness ratio
pffiffiffiffiffiffiffiffiffiffiffiffi
greater than 5:00 Es =F y is not permitted by AISC 360-10 [10] due to:
(i) the lack of experimental data in the database for CFTs with such slen-
der steel tubes, and (ii) potential issues with deflections and stresses in
the slender tube walls due to concrete casting pressures and other fab-
rication processes. Table 1 summarizes the slenderness limits (λp and
ð5Þ
Eqs. (7)–(11) are specified by AISC 360-10 [10] to calculate the nom-
inal compressive strength of rectangular CFT columns accounting for
length effects, member slenderness, and residual stresses. In these
equations: Pn is the nominal compressive strength including length
effects; Pno is the nominal compressive strength of the section account-
ing for tube slenderness using Eqs. (2), (3), (5), or (6); Pe is the elastic
(Euler) buckling load of the column calculated using the column length

Table 1
Slenderness limits for rectangular CFT members.

Loading Description of element Width-to-thickness ratio λp compact/noncompact λr noncompact/slender λlimit maximum permitted
qffiffiffiffi qffiffiffiffi qffiffiffiffi
Axial compression Steel tube walls of rectangular CFT Members b/t 2:26 EFsy 3:00 EFsy 5:00 EFsy
qffiffiffiffi qffiffiffiffi qffiffiffiffi
Flexure Flanges of rectangular CFT members b/t 2:26 EFsy 3:00 EFsy 5:00 EFsy
qffiffiffiffi qffiffiffiffi qffiffiffiffi
Webs of rectangular CFT members h/t 3:00 EFsy 5:70 EFsy 5:70 EFsy
458 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468

(KL) and effective flexural stiffness (EIeff). The effective flexural stiffness Previous researchers [13] showed that the nominal flexural strength
includes contributions of both steel and concrete, and accounts for the (Mn) of rectangular CFT members with compact flanges and webs could
effects of concrete cracking. be calculated as the plastic moment (Mp) strength of the cross-section
  using the plastic stress distribution method in AISC 360-10 [10]
P no P no
Section I2.2a. This method assumes rigid-plastic behavior for the steel
When ≤2:25 P n ¼ P no 0:658 Pe : ð7Þ
Pe and concrete materials with the steel yield stress equal to Fy in compres-
sion and tension, and the concrete strength equal to 0.85f'c in compres-
sion and zero in tension. This indicates that the compact steel tube can
P no develop its yield stress Fy in compression and tension, and confine the
When N2:25 P n ¼ 0:877P e : ð8Þ
Pe concrete infill adequately to develop its compressive strength of
0.85f 'c. Fig. 3(a) shows the stress blocks for calculating the plastic mo-
Where, ment (Mp). The distance of the neutral axis (ap) from the compression
face is calculated by establishing axial force equilibrium over the
2 EIeff cross-section, and the plastic moment strength (Mp) is calculated
Pe ¼ π ð9Þ
ðKLÞ2 using this neutral axis location. The resulting Eqs. (12) and (13) for ap
and Mp are given below. The variables in these equations are defined
graphically in Fig. 3(a). In these equations H, b, tw, tf are the tube
EI eff ¼ Es Is þ C 3 Ec Ic ð10Þ depth, width, web thickness, and flange thickness respectively.

  2 F y Ht w þ 0:85f 0c bt f
As ap ¼ ð12Þ
C 3 ¼ 0:6 þ ≤0:9: ð11Þ 4F y t w þ 0:85 f 0c b
Ac þ As

tf tf
4. Slenderness limits for rectangular CFT members—flexure If λ ≤λp ; Mn ¼ M p ¼ F y bt f ap − þ F y bt f H−ap −

2   2

ap H−ap
For rectangular CFT members subjected to flexure, the slenderness þ F y ap 2t w þ F y H−ap 2t w
2 2
ratios are defined by the b/t ratio of the flanges and the h/t ratio of the   a −t

0 p f
webs. Table 1 includes the slenderness limits (λp and λr), and the max- þ0:85f c ap −t f b
2
imum permitted slenderness ratio for the flanges and webs of rectangu-
ð13Þ
lar CFT members subjected to flexure. The local buckling of the steel
tube flanges in compression due to flexure is similar to the local buck-
The flexural strength of noncompact and slender CFT members is
ling of the tube walls of CFTs in compression. Therefore, as shown in
calculated using the principles of the lower bound theorem of plasticity.
Table 1, the slenderness limits for the flanges of rectangular CFT mem-
The lower bound capacity of the composite section is calculated using
bers subjected to flexure are identical to the limits for steel tube walls
admissible stress blocks that satisfy the equations of equilibrium,
of CFTs subjected to axial compression.
boundary conditions, and do not violate the yield criterion
However, the webs are subjected to stress gradients over their depth
(stress N yield stress Fy) anywhere. For noncompact CFT members in
(H). The portion of the web subjected to compressive stresses (above
flexure, this method assumes: (i) elastic-plastic behavior for the steel
the neutral axis) is much shorter than the depth (H), and subjected to
in tension and elastic behavior in compression up to the yield stress
linearly varying compressive strains. As shown in Ziemian [15], the
Fy, (ii) elastic behavior for concrete in compression up to 0.70f'c and
elastic local buckling equation for this case is the same as Eq. (1) but
no contribution in tension. For slender CFT members in flexure, this
with the k factor equal to 23.9. Therefore, the compactness pffiffiffiffiffiffiffiffiffiffiffiffi limit (λp) method assumes: (i) elastic behavior for the steel in tension up to the
for the webs was set conservatively p 3:00 Es =F y , and the
asffiffiffiffiffiffiffiffiffiffiffiffi
yield stress Fy and elastic behavior in compression up to the local buck-
noncompactness limit (λr) was set as 5:70 Esp =Fffiffiffiffiffiffiffiffiffiffiffiffi
y . The maximum slen-
ling stress Fcr, and (ii) elastic behavior for concrete in compression up to
derness ratio for the webs was also set as 5:70 Es = F y due to the lack of
0.70f'c and no contribution in tension. The local buckling critical stress
experimental data and other concerns such as concrete placement.
(Fcr) is given by Eq. (4).
Thus, rectangular CFT members subjected to flexure may have:
Due to their slenderness, noncompact and slender steel tubes
(i) compact, noncompact, or slender flanges, but (ii) only compact or
can neither sustain the compressive stress after local buckling, nor ade-
noncompact webs. However, CFT members with slender flanges and
quately confine the concrete to reach its compressive strength (0.85f'c).
noncompact webs are still classified as slender for flexure. The flexural
As mentioned earlier, concrete has significant volumetric dilation after
resistance is provided by the concrete infill in compression, and the
the compressive stress exceeds 0.70f'c, which exacerbates the post
steel tube in both tension and compression. The concrete contribution
local buckling deterioration of the noncompact steel tube and its inabil-
in tension is ignored due to cracking, which occurs at early stages of
ity to confine the concrete.
loading. The concrete infill prevents the steel tube from buckling in-
The admissible stress blocks for estimating the lower-bound capaci-
wards, and the outward buckling mode slightly increases the section
ty of noncompact sections with tube slenderness equal to λr are shown
moment of inertia and spreads plastic deformations providing more sta-
in Fig. 3(b). As shown, the steel tube is assumed to undergo yielding and
ble post-buckling behavior for CFT beams than HSS tubular beams. The
plasticification in tension and reach the yield stress Fy in compression.
rectangular steel tube provides nominal confinement to the concrete
The concrete is assumed to have maximum compressive stress equal
infill, but this confinement varies significantly with tube slenderness.
to 0.70f'c. The distance of the neutral axis (ay) from the compression
Experimental results indicate that the shear span-to-depth (a/H) ratio
face is calculated by establishing axial force equilibrium over the
and slip between the steel tube and concrete infill do not have a signif-
cross-section, and the moment capacity (Mn = My) is calculated using
icant influence on the moment capacity [16,17]. The flexural behavior
this neutral axis location as shown in Eqs. (14) and (15).
and capacity of rectangular CFT members is governed by yielding of
the steel tube in tension, followed by local buckling in compression 0
2 F y Ht w þ 0:35f c bt f
and concrete crushing failure depending on the material strength and ay ¼ ð14Þ
tube slenderness ratio. 4F y t w þ 0:35f 0c b
 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468 459

tf tf calculated using Eq. (16), which assumes a linear variation between

If λ ¼ λr ; Mn ¼ My ¼ F y bt f ay − þ F y bt f H−ay −

2

2 the moment capacities (Mp corresponding to λp, and My corresponding

2ay 2ay to λr) and tube slenderness. This linear interpolation is adequate be-
þ0:5F y ay 2t w þ 0:5F y ay 2t w
3 3 2  3 ð15Þ
 
  2 ay −t f
cause the concrete contribution to the flexural strength is generally
d 0
þ F y H−2ay 2t w þ 0:35f c ay −t f b4 5: smaller than that of the steel tube. This is in contrast to the axial load
2 3
strength of noncompact CFT columns where the concrete contribution
can be much higher than that of the steel tube, and a nonlinear variation
  was recommended in Eq. (6).
M p −M y   The admissible stress blocks for estimating the lower bound capacity
If λp bλ≤λr ; Mn ¼ Mp −   λ−λp : ð16Þ of slender sections with tube slenderness greater than or equal to λr are
λr −λp shown in Fig. 3(c). As shown, the steel tube is assumed to just reach the
yield stress Fy in tension and the critical buckling stress Fcr in compres-
For noncompact sections with tube slenderness (λ) greater than λp sion. The concrete is assumed to have maximum compressive stress
but less than or equal to λr, the nominal moment capacity (Mn) can be equal to 0.70f 'c. The distance of the neutral axis (acr) from the

Fig. 3. Stress blocks used to calculate the flexural strength of rectangular CFT beams.
460 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468

compression face is calculated by establishing axial force equilibrium

over the cross-section, and the moment capacity (Mn = Mcr) is calculat-
ed using this neutral axis location as shown in Eqs. (17) and (18).
 
F y Ht w þ 0:35f 0c þ F y − F cr bt f
acr ¼   ð17Þ
t w F cr þ F y þ 0:35f 0c b

tf tf
If λ≥λr ; Mn ¼ Mcr ¼ F cr bt f acr − þ F y bt f d−acr −
2 2

 
2acr 2ðd−acr Þ
þ0:5F cr acr 2t w þ 0:5F y ðd−acr Þ2t w
3  3 ð18Þ
2 3
  2 acr −t f
0
þ0:35f c acr −t f b4 5
3

Finally, Fig. 2(b) shows the variation of the nominal flexural strength
with respect to the tube slenderness obtained using Eqs. (15), (16), and
(18).

5. Design of noncompact and slender CFT beam-columns

The behavior of CFT beam-columns subjected to combined axial

compression and flexure depends on three major factors: (i) tube slen-
derness ratio, (ii) axial load ratio, and (iii) member length. The tube
slenderness ratio governs the local buckling behavior of the steel
tubes, and the confinement of the concrete infill. The axial load ratio
(α = Pr/Pc) is the ratio of the required axial strength (Pr) to available
axial strength (Pc) of CFT columns. It governs the axial load-bending
moment (P–M) interaction behavior of the CFT beam-column. When
the axial load ratio (α) is low, i.e., below the balance point on the P–M
interaction curve, flexural behavior dominates the response. When α
is high, i.e., above the balance point, axial compression behavior domi-
nates the response. The member length has greater influence on the
axial strength and the behavior for higher axial load ratios (α) above
the balance point. Additionally, the shape of the P–M interaction curve
for CFT beam-columns is influenced by the relative strength ratio ξ de-
fined in Eq. (19), where As and Ac are the total areas of the steel tube
and concrete infill, respectively. CFT beam-columns with larger ξ values
have P–M interaction curves that are comparable to those for steel col-
umns, while beam-columns with smaller ξ have P–M interaction curves
that are comparable to those for reinforced concrete columns, as shown
in Fig. 4(b).
As F y
ξ¼
Ac f 0c
Fig. 4. (a) AISC 360-10 P–M interaction diagram for CFT beam-columns; and (b) effect of ξ
on the CFT beam-column interaction curve.
interaction for designing noncompact and slender CFT beam-columns.
As shown, only the axial load strength (Pn) and flexural strength (Mn)
are required to develop the P–M interaction equation. For noncompact
and slender CFT beam-columns, these can be calculated using
Eqs. (3)–(11) for Pn and Eqs. (14)–(18) for Mn. In a separate paper
[18], the authors have proposed revisions to these P–M interaction
equations (Eqs. (20) and (21)) to account for the beneficial effects of
axial compression on the flexural strength of noncompact or slender
CFT beam-columns. These revisions were developed by conducting
parametric studies to evaluate the effects of the factors (tube slender-
ness ratio, axial load ratio, and member length) mentioned above. The
revised P–M interaction equations are not discussed in this paper be-
ð19Þ
cause they are not part of the AISC 360-10 [10] specification.

Previous researchers [13] provided approaches for developing P–M Pr Pr 8 Mr

When ≥0:2 þ ≤1:0: ð20Þ
interaction curves and designing compact CFT beam-columns with ϕc P n ϕc P n 9 ϕb Mn
tube slenderness ≤ λp. As explained in AISC 360-05 [9], these ap-
proaches were based on calculating the section P–M interaction curves
using rigid-plastic material properties for both steel and concrete, and Pr Pr Mr
When b 0:2 þ ≤1:0: ð21Þ
then including second-order length effects. The shapes of the resulting ϕc P n 2ϕc P n ϕb Mn
section P–M interaction curves were similar to those for reinforced con-
crete columns, and accounted for the increase in flexural strength due to
axial compression. 6. Experimental database and comparison with design equations
AISC 360-10 [10] provides Eqs. (20) and (21) for developing P–M in-
teraction curves and designing noncompact and slender CFT beam- As part of this research, the authors compiled a comprehensive data-
columns with tube slenderness greater than λp. These equations are base of tests conducted on noncompact and slender CFT members sub-
based on those for steel columns, and do not account for the expected jected to axial compression, flexure, and combined axial compression
beneficial effects of axial compression on the flexural strength, which and flexure. The database included tests from the databases in [4] and
is quite conservative. This conservatism was included to account for: [5], and additional tests from other databases and literature as applica-
(i) the potentially complex behavior of noncompact and slender CFT ble. This section presents the tests included in the noncompact and slen-
beam-columns, which are more susceptible to local buckling effects der CFT database, and the comparisons of experimental results with
and lack of concrete confinement, and (ii) the lack of significant exper- nominal strength predictions based on AISC 360-10 [10] design equa-
imental data. Fig. 4(a) shows an example of the resulting P–M tions presented in Sections 3–5. This discussion does not include tests
 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468 461

on compact CFT specimens, which were the basis of the earlier AISC
360-05 [9] specifications.
6.1. Non-compact and slender CFT members under axial compression
Forty-one noncompact and slender CFT column tests from nine pa-
pers [19–27] were compiled into the database. In these tests, the axial
loads were applied to the specimen using load or displacement control.
Specimens with steel tube loaded only are not included in the database.
Table 2 summarizes the noncompact and slender CFT column tests that
were included. The table includes the relevant parameters for the spec-
imens included in the database. These include the length (L), width (B),
depth (H), flange thickness (tf), web thickness (tw), governing tube
slenderness ratio (B/t and h/tw), and the slenderness coefficient p(λ coeff)
ffiffiffiffiffiffiffiffiffiffiffiffi
obtained by dividing the governing slenderness ratio with Es =F y .
Table 2 also includes the steel yield stress (Fy) and concrete strength
(f 'c) reported by the researchers. The experimental axial load strength
(Pexp) is included in the Table along with the nominal strength (Pn) cal-
culated using Eqs. (2)–(11) as applicable.
Fig. 5(a) shows the comparisons of the calculated axial load capacity
with the experimental results for all specimens in the database. The
ordinate represents the ratio of experimental-to-calculated axial load
strength (Pexp/Pn), while the abscissa represents the normalized
slenderness coefficient (λcoeff). These comparisons indicate that:
(1) the design equations calculate the axial strength conservatively ex-
cept for the data set labeled as “high strength steel.” This data set in-
cludes 11 data points with steel yield stress larger than 525 MPa,
which are not permitted by the AISC specifications and identified in
Table 2 with superscript a. (2) The design equations are more conserva-
tive for columns with slender sections, as compared to columns with
noncompact sections. And, (3) the degree of conservatism of the design
equations varies even for specimens with the same slenderness ratio.
For example, the Pexp/Pn ratio changes from 1.20 to 1.74 for columns
with slenderness coefficient of 3.66.
These observations are explained using CFT behavior mechanics. As
discussed previously, the interaction between the steel tube wall and
concrete infill produces hoop stresses in the steel tube wall and confine-
ment of the concrete infill [28]. These hoop stresses reduce the axial
stress required to cause yielding of the steel, and the confinement of
the concrete infill also increases its compressive strength. For rectangu-
lar CFT members, concrete confinement occurs at the corners and in the
core. AISC 360-10 [10] does not account for these effects directly while
calculating the axial strength of CFT columns. As a result, the strength
contribution of steel tube may be overestimated, while the strength
contribution of the concrete infill may be slightly underestimated. The
degree of conservatism of the design equations is therefore dependent

Table 2
Noncompact and slender rectangular CFT column tests.

Reference Specimen ID L B tf b/tf H tw h/tw λcoeff L/H Fy f 'c Ec Pn Pexp Pexp/Pn

(mm) (mm) (mm) (mm) (mm) (MPa) (MPa) (GPa) (kN) (kN)

Janss and Anslijn [19] 21 1318.3 329.9 4.47 71.8 329.9 4.47 71.8 3.09 4.0 370.3 31.6 26.60 4272.9 4363.3 1.02
22 1328.4 331.0 4.47 72.0 331.0 4.47 72.0 3.10 4.0 370.3 27.4 24.76 3979.2 4411.7 1.11
23 1320.8 331.0 4.50 71.6 331.0 4.50 71.6 3.08 4.0 370.3 27.4 24.76 4013.7 4656.8 1.16
24 1318.3 331.0 4.50 71.6 331.0 4.50 71.6 3.08 4.0 370.3 31.6 26.60 4315.1 4411.7 1.02
25 1318.3 333.0 6.38 50.2 333.0 6.38 50.2 2.37 4.0 444.7 31.6 26.60 6262.6 5862.8 0.94
26 1318.3 331.0 6.30 50.5 331.0 6.30 50.5 2.38 4.0 444.7 31.6 26.60 6150.2 5843.2 0.95
27 1318.3 331.0 6.32 50.3 331.0 6.32 50.3 2.37 4.0 444.7 27.4 24.76 5835.4 5833.4 1.00
28 1320.8 331.0 6.32 50.3 331.0 6.32 50.3 2.37 4.0 444.7 27.4 24.76 5835.1 5637.2 0.97
Lin [20] D10 800.0 150.0 1.40 105.1 150.0 1.40 105.1 3.70 5.3 247.3 22.5 2.93 468.5 711.5 1.52
D12 800.0 150.0 2.10 69.4 150.0 2.10 69.4 2.45 5.3 248.5 22.5 2.95 671.4 792.8 1.18
D16 800.0 200.0 1.40 140.9 150.0 1.40 105.1 4.95 5.3 247.0 22.5 3.88 537.3 881.0 1.64
D18 800.0 200.0 2.10 93.2 150.0 2.10 69.4 3.29 5.3 248.3 22.5 4.34 739.3 843.8 1.14
E10 800.0 150.0 1.40 105.1 150.0 1.40 105.1 3.70 5.3 247.3 35.3 4.06 653.8 973.1 1.49
E15 480.0 200.0 1.40 140.9 150.0 1.40 105.1 4.95 3.2 247.0 33.7 4.14 766.8 1191.7 1.55
E18 800.0 200.0 2.10 93.2 150.0 2.10 69.4 3.29 5.3 248.3 35.3 4.35 984.6 1288.4 1.31
Fujimoto et al. [21] CR4-D-2 485.1 323.1 4.37 72.0 323.1 4.37 72.0 2.60 1.5 261.3 25.4 23.84 3316.6 3365.5 1.01
CR4-D-4-1 485.1 323.1 4.37 72.0 323.1 4.37 72.0 2.60 1.5 261.3 41.0 30.32 4464.6 4949.1 1.11
CR4-D-4-2 485.1 323.1 4.37 72.0 323.1 4.37 72.0 2.60 1.5 261.3 41.0 30.32 4464.6 4828.5 1.08
CR4-D-8 485.1 323.6 4.37 72.1 323.6 4.37 72.1 2.61 1.5 261.3 80.1 42.37 7344.6 7478.4 1.02
CR6-D-2 477.5 319.0 6.35 48.2 319.0 6.35 48.2 2.68 1.5 616.4a 25.4 23.84 6618.9 6318.7 0.95
CR6-D-4-1 477.5 318.5 6.35 48.2 318.5 6.35 48.2 2.67 1.5 616.4a 41.0 30.32 7671.8 7777.3 1.01
CR6-D-4-2 477.5 318.3 6.35 48.1 318.3 6.35 48.1 2.67 1.5 616.4a 41.0 30.32 7664.7 7470.3 0.97
CR6-D-8 477.5 318.5 6.35 48.2 318.5 6.35 48.2 2.67 1.5 616.4a 84.9 43.62 10658.3 10353.7 0.97
CR8-D-2 396.2 264.9 6.48 38.9 264.9 6.48 38.9 2.51 1.5 832.9a 25.4 23.84 6795.9 6544.2 0.96
CR8-D-4-1 396.2 263.9 6.48 38.7 263.9 6.48 38.7 2.50 1.5 832.9a 41.0 30.32 7523.6 7114.5 0.95
CR8-D-4-2 396.2 264.4 6.48 38.8 264.4 6.48 38.8 2.51 1.5 832.9a 41.0 30.32 7539.1 7169.6 0.95
CR8-D-8 396.2 264.9 6.48 38.9 264.9 6.48 38.9 2.51 1.5 832.9a 84.9 43.62 9682.5 8987.2 0.93
Song and Kwon [22] US15 660.4 222.8 3.00 72.3 222.8 3.00 72.3 2.86 3.0 313.7 30.1 25.98 1816.6 2412.7 1.33
Uy [23] NS1 558.0 186.0 3.00 60.0 186.0 3.00 60 2.32 3.0 300 32.0 26.77 1507.1 1555.0 1.03
NS7 738.0 246.0 3.00 80.0 246.0 3.00 80 3.10 3.0 300 38.0 29.17 2341.0 3095.0 1.32
NS13 918.0 306.0 3.00 100.0 306.0 3.00 100 3.87 3.0 300 38.0 29.17 3034.9 4003.0 1.32
NS14 918.0 306.0 3.00 100.0 306.0 3.00 100 3.75 3.0 281 47.0 32.44 3597.5 4253.0 1.18
NS15 918.0 306.0 3.00 100.0 306.0 3.00 100 3.75 3.0 281 47.0 32.44 3597.5 4495.0 1.25
NS16 918.0 306.0 3.00 100.0 306.0 3.00 100 3.75 3.0 281 47.0 32.44 3597.5 4658.0 1.29
Schneider [24] R1 600.0 152.3 3.00 48.8 76.6 3.00 23.5 2.32 7.8 430.0 30.5 26.61 806.4 819.0 1.02
Uy [25] HSS14 630.0 210.0 5.00 40.0 210.0 5.00 40 2.45 3.0 750a 30.0 25.92 3908.8 3710.0 0.95
HSS15 630.0 210.0 5.00 40.0 210.0 5.00 40 2.45 3.0 750a 30.0 25.92 3908.8 3483.0 0.89
Kang et al [26] KOM2001 599.4 199.9 3.20 60.5 199.9 3.20 60.5 2.41 3.0 317.9 24.8 23.55 1530.6 1577.8 1.03
KOM2001 749.3 249.9 3.20 76.1 249.9 3.20 76.1 3.03 3.0 317.9 24.8 23.55 2001.0 2123.1 1.06
KOM2001 899.2 300.0 3.20 91.7 300.0 3.20 91.7 3.66 3.0 317.9 24.8 23.55 2297.4 2749.9 1.20
KOM2001 599.4 199.9 3.20 60.5 199.9 3.20 60.5 2.41 3.0 317.9 30.3 26.07 1695.6 2463.0 1.45
KOM2001 899.2 300.0 3.20 91.7 300.0 3.20 91.7 3.66 3.0 317.9 30.3 26.07 2632.2 4590.6 1.74
Mursi and Uy [27] SH-C210 730.0 220.0 5.00 42.0 220.0 5.00 42 2.59 3.3 761a 20.0 21.16 3818.0 3609.0 0.95
SH-C260 880.0 270.0 5.00 42.0 270.0 5.00 52 2.59 3.3 761a 20.0 21.16 4350.7 3950.0 0.91
a
Steel yield strength exceeds AISC 360-10 limit of 525 MPa.
462 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468

Fig. 5. Comparisons of the nominal and experimental strengths for rectangular noncompact and slender: (a) CFT columns; (b) CFT columns (with normal strength steel only); (c) CFT
beams; (d) CFT beam-columns with normal strength steel; and (e) CFT beam-columns with high strength steel.

on the tube wall slenderness ratio and the relative material strengths of
steel and concrete, i.e., the relative strength ratio ξ defined earlier in
Eq. (19). The tube slenderness ratio governs the local buckling of the
steel tube, the relative area of the steel tube and concrete infill As/Ac,
and the hoop stresses induced in the steel tube. As a result, the AISC
360-10 [10] design equations are more conservative for specimens
with larger slenderness ratio. For the same slenderness ratio, high
strength steel tubes are more susceptible to local buckling than conven-
tional strength steel tubes [28,29]. As a result, AISC 360-10 [10] design
equations are not recommended for specimens with high strength
 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468 463

Fig. 6. (a) Typical 4-point loading scheme for flexural tests; (b) typical loading schemes for beam-column tests.

steel. The variation in Pexp/Pn ratios in Fig. 5(a) for the same normalized
slenderness ratio is due to the fact that the design equation does not ac-
count for the relative strength ratio ξ, which governs the relative areas
and strengths of the steel tube and concrete infill.
Since AISC 360-10 [10] specifies the steel yield stress in CFT columns
to be less than 525 MPa, 11 points with steel yield stress greater than
525 MPa were eliminated from the comparisons. The updated compar-
isons shown in Fig. 5(b) (data points labeled as “EXP”) indicate that the
AISC 360-10 [10] equations are reasonably conservative.
6.2. Noncompact and slender CFT members under flexure
Only four tests in Refs. [17,30] on noncompact or slender CFT beams
were available. The 4-point loading scheme shown in Fig. 6(a) was used
for all these specimens. Table 3 summarizes the noncompact and slen-
der CFT beam tests that were compiled into the experimental database.
The table includes the relevant parameters included in the database,
such as the length (L), width (B), depth (H), flange thickness (tf), web
thickness (tw), governing tube slenderness ratio (b/t and h/tw), and
the coefficient (λcoeff) obtained by dividing the governing slenderness
pffiffiffiffiffiffiffiffiffiffiffiffi
ratio with Es = F y . The shear span to depth ratio (a/H) is also included
in the Table. Table 3(a) also includes the steel yield stress (Fy) and con-
crete strength (f'c) reported by the researchers. The experimental flex-
ural strength (Mexp) is included in the table along with the nominal
flexural strength (Mn) calculated using Eqs. (12)–(18) as applicable.
Comparisons of Mexp/Mn are shown in Fig. 5(c) as the points labeled
“EXP.” These comparisons indicate that the AISC 360-10 [10] design
equations conservatively estimate the flexural strength of noncompact
and slender CFTs. The points labeled “Additional FEM” are presented
later in the paper.
6.3. Noncompact and slender CFT members under combined axial
compression and flexure
Seventeen noncompact and slender CFT beam-column tests from
Refs. [25,27,31,32], were compiled into the experimental database.
There are two types of loading schemes (Type-A and Type-B) for
conducting beam-column tests, as shown in Fig. 6(b). In Type-A loading,
concentric axial loading is applied first and maintained constant. The
bending moment is increased monotonically to failure. In Type-B load-
ing, eccentric axial loading is applied and increased monotonically,
which results in both axial force and bending moment increasing pro-
portionally to failure.
Table 4 summarizes the noncompact and slender CFT beam-column
tests that were compiled in the experimental database. The table in-
cludes the same parameters as Table 2. The experimental axial (Pexp)
and flexural (Mexp) capacities are also included in the table along with
the nominal axial (Pn) and flexural (Mn) capacities calculated using
AISC 360-10 [10] equations. In Fig. 5(d) and (e), the comparisons of
the interaction curve calculated using AISC 360-10 [10] and experimen-
tal results are shown separately for specimens with normal strength
steel and high strength steel. As shown, the AISC 360-10 [10] is conser-
vative for all specimens with normal strength steel. For specimens with
high strength steel, the AISC 360-10 [10] interaction curve is less conser-
vative. This is due to the fact that specimens with high strength steel
have larger relative strength ratio (ξ) as compared to specimens with
normal strength steel, which results in their behavior being comparable
to steel beam-columns.
7. Further verification of design equations using finite
element analysis
The experimental database provided valuable data points to verify
the design equations. However, there were some gaps in the experi-
mental database. For example, for axial compression, there are just
two data points for columns with λcoeff in the range of 4.0 and 5.0. For
flexure, there are no data points for beams with noncompact sections
and only four data points for beams with slender sections. For combined
axial compression and flexure also, additional data points are required
for beam-columns with different tube slenderness ratios and axial
load ratios. It was important to address these gaps in the database to fur-
ther confirm the conservatism of the design equations.
The finite element analysis approach was selected to address these
gaps in the experimental database, and to develop additional data
points to confirm the design equations. Detailed 3D finite element
method (FEM) models were developed and analyzed using ABAQUS
[33]. These models were first benchmarked using results from the ex-
perimental database. The benchmarked models were used to address
gaps in the database. The following sub-sections present: (i) the details
of the finite element models, (ii) benchmarking using experimental re-
sults, and (iii) results of analyses conducted to address gaps in the data-
base and further verify the design equations.

Table 3
Noncompact and slender rectangular CFT beam tests.

Reference Specimen ID L B (mm) tf b/tf H tw h/tw λcoeff a/H Fy f 'c Ec Mn Mexp Mexp/Mn
(mm) (mm) (mm) (mm) (MPa) (MPa) (GPa) (kN-m) (kN-m)

Han et al. [17] SVB-1 1400.0 200.0 1.90 103.3 200.0 1.90 103.3 3.86 1.75 282.0 81.3 42.67 32.9 42.3 1.29
SVB-2 1400.0 200.0 1.90 103.3 200.0 1.90 103.3 3.86 1.75 282.0 81.3 42.67 32.9 54.9 1.67
SSCB-1 1400.0 200.0 1.90 103.3 200.0 1.90 103.3 3.86 1.75 282.0 81.3 42.67 32.9 56.7 1.72
Jiang et al. [30] S-150-2.0 2000.0 150.0 2.00 73.0 150.0 2.00 73.0 3.18 4.67 397.0 56.0 37.42 26.3 31.1 1.18
464 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468

7.1. 3D FEM model details

Mexp/Mn

0.96
1.42
0.66
0.96
1.89
1.89
1.93
1.35
0.99
1.96
2.16
1.99
2.24

0.53
0.31

0.22
0.21
The steel tubes of the CFT members were modeled using a fine mesh
of four-node S4R shell elements. These elements have: (i) six degrees of
(kN-m)
freedom per node, and (ii) five section points to compute the stress and

130.9
201.0
297.0
408.0
593.0
136.0
136.0
139.0
195.0
143.0
Mexp

62.7
69.1
63.5
71.5
77.7

58.8
76.6
strain variations through the thickness, and (iii) reduced integration in
the plan of the elements. The elements model thick shell behavior, but
(kN-m)

209.7

249.1
209.7
614.3
617.6

144.2
144.2

249.1

265.5
368.3
converge to Kirchhoff's thin plate bending theory with reducing thick-
72.0
72.0
72.0

31.9
31.9
31.9
31.9
Mn

ness. The concrete infill of CFT members was modeled using eight-
node solid elements with reduced integration (C3D8R). These elements
Pexp/Pn

0.24
0.73

0.49
0.49

0.24
0.33
0.53
0.25

0.30
0.61
0.24
0.48

0.48

0.64
0.76

0.83
0.74
have three degrees of freedom per node and reduced integration to cal-
culate the stresses and strains in the elements. The C3D8R elements are
computationally effective for modeling concrete cracking. For additional
1049.0
3306.0

2108.0
2108.0
1479.0
4100.0
1967.0

1294.0
2569.0

2617.0
3106.0

2939.0
3062.0
380.0
761.0
380.0
761.0
(kN)
Pexp

details regarding the choice of these finite elements, please refer Varma
et al. [28].
4330.3
4520.1

4330.3
4330.3
4520.1
7725.2
7752.8

4242.7
4242.7
1594.2
1594.2
1594.2
1594.2
4077.3
4077.3
3539.3
4122.9
The contact interactions between the steel tube and concrete infill of
(kN)

CFT members were modeled in both the normal and tangential direc-
Pn

tions. The hard contact pressure-overclosure relationship with penalty

(GPa)

51.62
30.34
30.34
30.34
30.34

51.62
51.62
51.62
51.62
32.65
32.65
32.65
32.65
26.77
26.77
21.16
21.16 constraint method was used for interaction in the normal direction.
Ec

The penalty friction formulation with coulomb friction coefficient

equal to 0.55 and maximum interfacial shear stress equal to 0.41 MPa
(MPa)

119.0
119.0
119.0
119.0
119.0
41.1
41.1
41.1
41.1

47.6
47.6
47.6
47.6
32.0
32.0
20.0
20.0

(60 psi, as suggested by AISC 360-10 [10]) was used for interaction in
f 'c

the tangential direction. There was no additional bond (adhesive or

(MPa)

262.0

310.0
262.0
618.0
618.0

310.0
310.0
781.0
781.0
253.0
253.0
253.0
253.0
750.0

761.0
761.0
750.0

chemical) between the steel tube and the concrete infill in the model.
Fy

The steel tube and concrete infill could slip relative to each other
when the applied interfacial shear stress (τapp) exceeded 0.55 times
11.0
10.4
L/H

3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0

the contact pressure (p).

Geometric imperfections were defined to initiate local buckling in
λcoeff

2.56
2.56
2.63
2.64
2.41
2.41
2.41
3.92
3.92
3.42
3.42
3.42
3.42
2.45

2.59
3.21
2.45

the steel tubes models. As recommended by [34], the shape of the geo-
metric imperfection was developed by conducting eigenvalue buckling
71.7
71.7
48.0
48.2
61.1

96.0

40.0

42.0
52.0
61.1
61.1
62.7
62.7
96.0
96.0

96.0

40.0
h/tw

analysis and the amplitude (magnitude) of the geometric imperfection

was set equal to 0.1 times the tube thickness. Fig. 7(a) shows the first
(mm)

buckling eigenmode shape used to define the geometric imperfection.

5.00
4.38
4.38
6.36
6.36

5.00
3.17

2.04

5.00
3.17
3.17
3.09
3.09
2.04
2.04

2.04

5.00
tw

The boundary conditions and constraints used for the models were de-
signed to simulate those achieved in the experiments, i.e., by using cou-
(mm)

210.0
323.0
323.0
318.0
319.0

220.0
200.0

200.0

270.0
200.0
200.0
200.0
200.0
200.0
200.0

200.0

210.0

pling constrains as shown in Fig. 7(b).

H

The steel material multiaxial behavior was defined using the Von
71.7
71.7
48.0
48.2
61.1

96.0

40.0

42.0
52.0
61.1
61.1
62.7
62.7
96.0
96.0

96.0

40.0

Mises yield surface, associated flow rule, and kinematic hardening. An

b/tf

idealized bilinear curve was used to specify the uniaxial stress–strain

(mm)

behavior of steel. The yield stress (Fy) and elastic modulus (Es) as listed
2.04
4.38
4.38
6.36
6.36
3.17

5.00

5.00
5.00
3.17
3.17
3.09
3.09
2.04
2.04

2.04

5.00

in Table 2 were used for each specimen. The post-yield hardening mod-
tf

ulus was assumed to be equal to Es/100.

(mm)

200.0
323.0
323.0
318.0
319.0

220.0
270.0
200.0
200.0
200.0
200.0
200.0
200.0
200.0

200.0
210.0
210.0

The concrete material multiaxial behavior was modeled using the

B

damaged plasticity material model (CDP) developed by Lee and Fenves

[35]. This model accounts for multiaxial behavior using a special com-
2416.0
2817.0
(mm)

600.0
969.0
954.0
957.0
600.0
600.0
600.0
600.0
600.0
600.0
600.0

600.0
969.0

630.0
630.0

pression yield surface developed earlier by Lubliner et al. [36] and mod-
L

ified by Lee and Fenves [35] to account for different evolution of

BRA4-2-5-02-C
BRA4-2-5-04-C
BR4-3-10-04-1
BR4-3-10-04-2

strength in tension and compression. The empirical stress–strain curve

BRA4-2-5-02
BRA4-2-5-04
BR4-3-10-02

BR8-3-10-02
BR8-3-10-04
Specimen ID

ER4-D-4-20
ER6-D-4-10
ER6-D-4-30
ER4-D-4-06

proposed by Popovics [37] was used to define the compressive behavior

SH-C260
SH-C210
Noncompact and slender rectangular CFT beam-column tests.

of the concrete. The smeared cracking behavior in tension was specified

HSS16
HSS17

using a stress vs. crack opening displacement curve that is based on frac-
ture energy principles and empirical models developed by CEB-FIP [38],
Monotonic, Type-A
Monotonic, Type-B

Monotonic, Type-B

Monotonic, Type-B

as shown in Fig. 8.
The CDP model accounts for the non-associated flow behavior of
Cyclic, Type-A

concrete in compression [14] using: (i) the Drucker–Prager hyperbolic

Load type

function as the flow potential G, (ii) dilation angle ψ, and (iii) eccentric-
ity ratio e. The value of the dilation angle ψ was calibrated by Prabhu
et al. [39] using experimental data for axial and lateral stress–strain be-
Sakino and Nakahara [31]

Nakahara and Sakino [32]

havior reported by other researchers. The resulting value of the dilation

angle for unconfined concrete was 15°. Other researchers have reported
using values between 15° and 30°. Increasing the dilation angle (ψ)
Mursi and Uy [27]

above 15° will result in larger volumetric dilation of the concrete and
potentially better but unrealistic effects of confinement on the strength
Reference

Uy [25]

and ductility. The default value of 0.1 was specified for the eccentricity
Table 4

ratio, e. This default value of e indicates that the dilation angle in the
p–q (hydrostatic pressure–deviatoric stress) plane converges to ψ
 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468
Fig. 7. (a) First buckling eigenmode shape from eigenvalue analysis; and (b) example of boundary conditions, couplings, constraints, and mesh in the FEM model.
reasonably quickly with increasing hydrostatic compression pressure
(p). Other parameters required to define the multiaxial plasticity
model are: the ratio of biaxial compressive strength to uniaxial strength
f 'bc/f 'c, and the ratio of compressive to tensile meridians of the yield sur-
face in Π (deviatoric stress) space Kc. f 'bc/f 'c was assumed to be equal to
1.16 based on [40]. Kc was assumed to be equal to 0.67 based on [14].
The fracture behavior of concrete in tension makes it virtually
impossible to obtain converged results using standard (predictor-
corrector) nonlinear solution strategies like full Newton or modified
Newton–Raphson iteration approaches. Even arc-length-based tech-
niques like the modified-Riks methods cannot provide converged re-
sults due to the brittle fracture behavior of concrete in tension.
Implicit dynamic analysis methods also become unstable and cannot
provide results after significant cracking. Therefore, the explicit dynam-
ic method was used to conduct the analysis. The primary reason for
using this method is that it can find results up to failure, particularly
when brittle materials (like concrete in tension) and failure modes are
involved. The explicit dynamic analysis method was used to perform
quasi-static analyses simulating the experiments.
7.2. Benchmarking the FEM models
The developed FEM models were used to predict the behavior of the
specimens in the experimental database. Some of the specimens in the
experimental database could not be modeled because the correspond-
ing references did not include critical information needed to develop
the finite element models. Thirty-five (noncompact or slender) CFT
member tests from the database were used to benchmark the finite el-
ement models. These included 15 column tests, five beam tests, and 15
beam-column tests. Comparisons of the strengths from the finite
Fig. 8. Tension stress-crack opening displacement curve for concrete ( f 'c = 40 MPa).
465
element analysis and the corresponding tests are shown in Fig. 9,
where the ordinate is the ratio of experimental strength to that predict-
ed by FEM analysis, and the abscissa is the slenderness coefficient.
The strengths from the finite element analyses were defined appro-
priately for different loading conditions (axial compression, flexure, and
combined axial and flexure) as follows. For CFT columns, the axial
strength from the finite element analysis (PFEM) was defined as the max-
imum load value obtained from the analysis and the corresponding
comparisons are shown in Fig. 9(a). For CFT beams, if the method of de-
termining the flexural strength was specified in the corresponding ex-
perimental paper (for example, Han [41] defined flexural strength as
the moment corresponding to 1% strain in the extreme compression fi-
bers), then the same method was used to define the flexural strength
(MFEM) from the finite element analysis. Otherwise, the flexural strength
from finite element analysis (MFEM) was defined as the maximum mo-
ment obtained from the analysis. The corresponding comparisons are
shown in Fig. 9(b).
For beam-columns with Type-A loading (constant axial loading),
only comparisons of the flexural strength were required. The flexural
strength from finite element analysis (MFEM) was defined as the maxi-
mum moment obtained from the analysis, and the resulting compari-
sons are shown in Fig. 9(d). For CFT beam-columns with Type-B
loading (eccentric axial loading), comparisons of both axial and flexural
strength were required. The axial strength (PFEM) was defined as the
maximum value obtained from the analysis, and the flexural strength
(MFEM) was defined as the bending moment corresponding to PFEM in-
cluding second-order effects. The resulting comparisons are shown in
Fig. 9(c) and (d).
Fig. 10 shows typical comparisons of experimental and analytical
load-displacement responses. It includes comparisons of: (i) the mo-
ment vs. midspan deflection curves for beam specimens RB3-1 and
S150-2.0, and (ii) the moment-curvature responses for beam-column
specimens BRA4-2-05-02 and BRA4-2-05-04. The comparisons shown
in Fig. 10 are typical and representative of the comparisons between ex-
perimental and analytical load-displacement responses. As shown in
Figs. 9 and 10, the finite element models predict the behavior and
strengths of CFT columns, beams, and beam-columns reasonably well.
7.3. Further verification of design equations
The benchmarked FEM models were used to perform additional
analysis to address gaps in the experimental database. Prototype speci-
mens were selected from the experimental database for CFT columns,
beams, and beam-columns. Additional analyses were performed using
the benchmarked models of these prototype CFT specimens by varying
the tube slenderness ratios or axial load ratios. The slenderness ratios
were varied by changing the tube wall thickness. The axial load ratios
466 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468

Fig. 9. Comparisons of strengths from the FEM analyses with the corresponding experimental results for: (a) columns; (b) beams; (c) beam-columns with Type-B loading only, and
(d) beam-columns with Type-A and Type-B loadings.

Fig. 10. Comparisons of analytical and experimental load-deflection curves for beam specimens (a) RB3-1 and (b) S150-2.0; and beam-column specimens (c) BRA4-2-5-02, and (d) BRA4-
2-5-04.
 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468 467

(P/Pn) were varied by changing the applied axial load level (P), and
recalculating the ratio using the axial strength Pn.
The prototype selected for CFT columns was specimen E10 tested by
Lin [20]. Thirteen additional analyses were conducted for different tube
slenderness ratios in the range of noncompact and slender (λcoeff =
2.26–5.0) CFT columns. The prototype selected for CFT beams was spec-
imen S-150-2.0 tested by Jiang [30]. Fig. 10(b) showed the comparisons
of the experimental and analytical results for this specimen. Fourteen
additional analyses were conducted for different tube slenderness ratios
in the range of noncompact and slender (λcoeff = 2.26–5.0) CFT beams.
The prototype selected for CFT beam-columns was specimen BRA4-
2-5-04 tested by Nakahara and Sakino [32]. Fig. 10(d) showed the
comparisons of the experimental and analytical results for this speci-
men. Both the tube slenderness ratio and axial load ratio influence the
behavior of CFT beam-columns. Thirty-five additional analyses were
conducted for seven different tube slenderness ratios (λcoeff in the
range of 2.26–5.0) and five different axial load ratios (0, 0.2, 0.4, 0.6,
and 1.0). Beam-columns with axial load ratio of 0 (i.e., beams) were sub- Fig. 12. Comparisons of the critical stress Fcr from FEM analysis and Eq. (4).
jected to bending moments only, and beam-columns with axial load
ratio of 1.0 (i.e., columns) were subjected to axial compressive force
only. Thus, the 35 analyses included 21 beam-column analyses, seven sectional area (As). Fig. 12 shows comparisons of the critical buckling
beam analyses and seven column analyses. stress (FFEM
cr ) obtained from the finite element analyses with respect to
Comparisons of the strength results from the additional finite the values calculated using Eq. (4). The Figure includes (FFEMcr ) data
element analyses and AISC 360-10 [10] equations are included in points from both the benchmarking FEM analyses of tested specimens
Fig. 5(b) for columns and Fig. 5(c) for beams. These data points are la- in Section 7.2 (labeled as “FEM”) and the additional FEM analyses in
beled as “Additional FEM” in the figures. Comparisons of the strength re- Section 7.3 (labeled as “Additional FEM”). These comparisons confirm
sults from the additional analyses of beam-columns and the AISC 360- the conservatism of Eq. (4) for estimating the critical buckling stress
10 [10] nominal strength P–M interaction curve are shown in Fig. 11. of slender CFT members.
These comparisons verify that the AISC 360-10 equations conservatively
estimate the nominal strength of noncompact and slender rectangular 9. Conclusions
CFT members.
This paper presented the basis of the AISC 360-10 [10] Specification
8. Critical buckling stress from FEM for the design of noncompact and slender rectangular CFT members.
Slenderness limits for the classification of rectangular CFT members
As discussed in Section 3, the elastic buckling stress (Fcr) of rectangu- subjected to axial compression or flexure were discussed and presented.
lar CFT member is calculated using Eq. (4), which is based on the The experimental database of noncompact and slender CFT members
research of Bradford [12] using the Rayleigh–Ritz method. The conser- was compiled by reviewing available experimental research. The con-
vatism of Eq. (4) was further verified using the results from the FEM servatism of design equations was verified by using them to calculate
analyses. The stresses in the steel tubes at local buckling (FFEM
cr ) were ex- the strength of all specimens in the experimental database. Detailed
tracted from the results of the FEM analyses as follows. Local buckling of FEM models were developed and benchmarked using experimental re-
the steel tubes occurred very close to the maximum load capacity (PFEM) sults from the database. The benchmarked FEM models were used to
being reached. Therefore, when the maximum load value (PFEM) was address gaps in the experimental database, and further verify the con-
reached in the analysis, the corresponding axial force (SF) in each servatism of the design equations by addressing gaps in the experimen-
steel element of the midspan cross-section (where local buckling oc- tal database. The AISC 360-10 [10] equations are appropriate for
curred) was extracted. The axial force carried by the steel tube (Pcr) classifying rectangular CFT members into compact, noncompact or slen-
was calculated as the sum of the axial forces carried by all steel elements der sections for axial compression or flexure. These equations can be
in the cross-section at the midspan. The steel tube stress at local buck- used to conservatively calculate the strengths of rectangular CFT mem-
ling (FFEM
cr ) was then estimated as Pcr divided by the steel tube cross- bers. The benchmarked FEM models are recommended for predicting
and further evaluating the behavior of noncompact or slender rectangu-
lar CFT columns, beams, and beam-columns.

Nomenclature
Ac Area of the concrete infill
As Area of the steel tube wall
B Width of rectangular sections
EIeff Effective stiffness of CFT members
Es Steel elastic modulus
Fcr Elastic buckling stress
FFEM
cr Critical buckling stress from FEM analyses
Fy Steel yield strength
H Depth of rectangular sections
Ic Moment of inertia of the concrete section
Is Moment of inertia of the steel section
K Effective length factor
L Length of CFT members
Fig. 11. Further verification of the design equations for beam-columns. Mc Available flexural strength CFT members
468 Z. Lai et al. / Journal of Constructional Steel Research 101 (2014) 455–468

Mcr Flexural strength of rectangular CFT members with slender [11] Lai Z. Experimental databases, analysis and design of noncompact and slender
concrete-filled steel tube (CFT) members. [Dissertation] Purdue University; 2014.
sections [12] Bradford MA, Wright HD, Uy B. Local buckling of the steel skin in lightweight com-
Mexp Experimental flexural strength of CFT members posites induced by creep and shrinkage. Adv Struct Eng 1998;2:25–34.
Mn Nominal flexural strength of CFT members [13] Leon RT, Kim DK, Hajjar JF. Limit state response of composite columns and Part 1:
formulation of design provisions for the 2005 AISC specification. Eng J 2007;05:
Mp Flexural strength of compact CFT members 341–58.
Mr Required flexural strength of CFT members [14] Chen WF, Han D. Plasticity for structural engineers. Plantation, FL, USA: J. Ross
My Flexural strength of CFT members with slenderness equal to Publishing; 2007.
[15] Ziemian RD, editor. Guide to stability design criteria for metal structures. 6th ed.
λr Hoboken, NJ: John Wiley & Sons, Inc.; 2010.
Pc Available axial strength of CFT members [16] Lu YQ, Kennedy LDJ. The flexural behaviour of concrete-filled hollow structural sec-
Pcr Cross-section compressive strength of slender CFT members tions. Can J Civ Eng 1994;21:111–30.
[17] Han L-H, Lu H, Yao G-H, Liao F-Y. Further study on the flexural behaviour of
Pe Elastic critical buckling load
concrete-filled steel tubes. J Constr Steel Res 2006;62:554–65.
Pexp Experimental axial strength of CFT members [18] Lai Z, Varma AH. Analysis and design of noncompact and slender concrete-filled
Pn Nominal compressive strength of CFT members steel tube (CFT) beam-columns. Proc. Annu. Stab. Conf. Struct. Stab. Res. Counc.
Pnc Cross-section compressive strength of noncompact CFT Toronto, Canada: SSRC; 2014.
[19] Janss J, Anslijn R. Le Calcul des Charges Ultimes des Colonnes Métalliques Enrobés de
members Béton. Rapp CRIF, MT; 1974 89 (Bélgica).
Pno Nominal cross section compressive strength of CFT members [20] Lin CY. Axial capacity of concrete infilled cold-formed steel columns. In: Yu W-W,
Pp Cross-section compressive strength of compact CFT members Joseph HS, editors. Ninth Int. Spec. Conf. Cold-Formed Steel Struct. St. Louis: Univer-
sity of Missouri - Rolla; 1988. p. 443–57.
Pr Required axial strength of CFT members [21] Fujimoto T, Nishiyama I, Mukai A, Baba T. Test results of eccentrically loaded short
Py Cross-section compressive strength of CFT members with columns – square CFT columns. Proc. Second Jt. Tech. Coord. Comm. Meet. U.S.-
slenderness equal to λr Japan Coop. Res. Program, Phase 5 Compos. Hybrid Struct. Honolulu: National Sci-
ence Foundation; 1995.
a Shear span length [22] Song JY, Kwon Y. Structural behavior of concrete-filled steel box. Compos. Constr.
acr Distance of the neutral axis to the top of the section with Mcr Innov. Int. Conf. Innsbruck, Austria: International Association for Bridge and Struc-
ap Distance of the neutral axis to the top of the section with Mp tural, Engineering; 1997. p. 795–800.
[23] Uy B. Local and post-local buckling of concrete filled steel welded box columns.
ay Distance of the neutral axis to the top of the section with My J Constr Steel Res 1998;47:47–72.
f 'c Concrete uniaxial compressive strength [24] Schneider BSP, Member A. Axially loaded concrete-filled steel tubes, d; 1998
k Plate buckling coefficient 1125–38.
[25] Uy B. Strength of short concrete filled high strength steel box columns. J Constr Steel
t Thickness of the steel tube wall
Res 2001;57:113–34.
tf Thickness of the steel tube flange [26] Kang CH, Oh YS, Moon TS. Strength of axially loaded concrete-filled tubular stub col-
tw Thickness of the steel tube web umn. Int J Steel Struct 2001;13:279–87.
λ Tube wall slenderness ratio, b/t [27] Mursi M, Uy B. Strength of slender concrete filled high strength steel box columns.
pffiffiffiffiffiffiffiffiffiffiffiffi J Constr Steel Res 2004;60:1825–48.
λcoeff Slenderness coefficient, λcoeff ¼ ðb=t Þ= Es =F y [28] Varma AH, Ricles JM, Sause R, Lu L-W. Seismic behavior and modeling of high-
λlimit Maximum permitted slenderness ratio strength composite concrete-filled steel tube (CFT) beam–columns. J Constr Steel
λp Slenderness limit for compact/noncompact sections Res 2002;58:725–58.
[29] Varma AH, Ricles JM, Sause R, Lu L-W. Experimental behavior of high strength
λr Slenderness limit for noncompact/slender sections square concrete-filled steel tube beam-columns. J Struct Eng 2002;128:309–18.
ξ Relative strength ratio, ξ = (AsFy)/(Ac f 'c) [30] Jiang A, Chen J, Jin W. Experimental investigation and design of thin-walled
υ Poisson's ratio concrete-filled steel tubes subject to bending. Thin-Walled Struct 2013;63:44–50.
[31] Sakino K, Nakahara H. Practical analysis for high-strength CFT columns under eccen-
tric compression. In: Xiao Y, Mahin S, editors. Proc. Sixth ASCCS Int. Conf. Steel-
Concrete Compos. Struct., Los Angeles, California; 2000. p. 473–80.
References [32] Nakahara H, Sakino K. Flexural behavior of concrete filled square steel tubular beam-
columns. Proc. Twelfth World Conf. Earthq. Eng. Auckland, New Zealand: New
[1] Klöppel K, Goder W. Traglastversuche mit ausbetonierten Stahlrohen und Zealand Society for, Earthquake Engineering; 2000. p. 441–8.
Aufstellung einer Bemessungsformel. Berlin: Der Stahlbau; 1957 26. [33] ABAQUS. ABAQUS version 6.12 analysis user's manuals. Providence, RI, USA:
[2] Nishiyama I, Morino S, Sakino K, Nakahara H, Fujimoto T, Mukai A, et al. Summary of Dassault Systemes Simulia Corporation; 2012.
research on concrete-filled structural steel tube column system carried out under [34] Varma AH, Sause R, Ricles JM, Li Q. Development and validation of fiber model for
the U.S.-Japan cooperative research on composite and hybrid structures. Japan: high-strength square concrete-filled steel tube. ACI Struct J 2005;102:73–84.
Ibaraki Prefecture; 2002. [35] Lee J, Fenves GL. Plastic-damage model for cyclic loading of concrete structures. J Eng
[3] Kim DK. A database for composite columns. [Thesis] Georgia Institute of Technology; Mech 1998;124:892–900.
2005. [36] Lubliner J, Oliver J, Oller S, Onate E. A plastic-damage model for concrete. Int J Solids
[4] Gourley BC. A synopsis of studies of the monotonic and cyclic behavior of concrete- Struct 1988;25:299–326.
filled steel tube members, connections, and frames. Report No. NSEL-008. Depart- [37] Popovics S. A numerical approach to the complete stress–strain curve of concrete.
ment of Civil and Enviromental Enginnering, University of Illinois at Urbana- Cem Concr Res 1973;3:583–99.
Champaign; 2008. [38] Comité Euro International du Béton (CEB), Fédération Internation alede la
[5] Hajjar JF, Gourley BC, Tort C, Denavit MD, Schiller PH. Steel-concrete composite Précontrainte (FIP). Model Code for concrete structures. (CEB-FIP MC 2010)
structural systems. http://www.northeastern.edu/compositesystems Northeastern London, U.K.: Thomas Telford; 2010
University; 2013. [39] Prabhu M, Varma A, Buch N. Analytical investigation of the effects of dowel misalign-
[6] Eurocode. Eurocode 4: design of composite steel and concrete structures Part 1-1: ment on concrete pavement joint opening behaviour. Int J Pavement Eng
general rules and rules for buildings. Brussels, Belgium: CEN; 2004. 2009;10:49–62.
[7] AS. Australian standard for steel structures. Canberra, Australia: ABCB; 2012. [40] Kupfer HB, Gerstle KH. Behavior of concrete under biaxial stresses. J Eng Mech Div
[8] AIJ. Recommendations for design and construction of concrete filled steel tubular 1973;99:853–66.
structures. Tokyo, Japan: Architectural Institute of Japan; 2008. [41] Han L-H. Flexural behaviour of concrete-filled steel tubes. J Constr Steel Res
[9] AISC 360-05. Specification for structural steel buildings. Chicago, IL, USA: AISC; 2005. 2004;60:313–37.
[10] AISC 360-10. Specification for structural steel buildings. Chicago, IL, USA: AISC; 2010.

View publication stats