Stability Analysis and Design of Composite Structures

Mark D. Denavit, [Link] 1; Jerome F. Hajjar, [Link] 2; Tiziano Perea, [Link] 3;

and Roberto T. Leon, [Link] 4

Abstract: The direct analysis method is the primary means of assessing system stability within a standard specification. This method, and in
particular its use of reduced stiffness, has been thoroughly validated for use in frames consisting of structural steel members. However,
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

appropriate stiffness reductions have not yet been established nor has the method as a whole been validated for frames with steel-concrete
composite columns. Through comparisons between second-order inelastic analysis results and results from the design methodology on a
parametric suite of small frames, the current design provisions are evaluated in this paper. The results indicate that while the current design
provisions are safe and accurate for the majority of common cases, there exist cases in which the current provisions result in high levels of
unconservative error. Modifications to the current design provisions are proposed to address these issues. DOI: 10.1061/(ASCE)ST.1943-
541X.0001434. © 2015 American Society of Civil Engineers.
Author keywords: Steel-concrete composite; Beam columns; Structural stability; Design; Metal and composite structures.

Introduction frames focusing on the development and validation of direct analy-

sis recommendations for composite systems was conducted and is
The direct analysis method as defined in the American Institute presented in this paper.
of Steel Construction (AISC) “Specification for Structural Steel
Buildings” (AISC 2010b) provides an accurate and straightforward
way of addressing in-plane frame stability in the design process. In Benchmark Frames
this method, required strengths are determined from a second-order
elastic analysis in which members are modeled with a reduced ri- This paper includes a parametric study in which results from sec-
gidity and initial imperfections are either directly modeled or rep- ond-order inelastic analyses of benchmark composite structures are
resented with notional lateral loads. Available strengths are then compared with results from current and proposed design method-
computed based on the unsupported length of the member, elimi- ologies based on second-order elastic analysis. The results of both
nating the need to compute an effective length factor. are dependent on specific cross section properties and frame con-
The validity of the direct analysis approach has been established figuration. Thus, to ensure broad applicability of the findings, a
through comparisons between second-order elastic analyses and wide variety of material and geometric properties are examined.
second-order inelastic analyses that had been deemed sufficiently Prior benchmark studies to calibrate stability procedures for struc-
accurate to provide a basis for design provisions (Surovek-Maleck tural steel systems made use of a set of small nonredundant frames
and White 2004a, b; Deierlein 2003; Martinez-Garcia 2002). Sim- constructed with a W200 × 46.1ðW8 × 31Þ section in either strong
ilar methodologies were utilized to establish the validity of the or weak axis bending (Kanchanalai 1977; Surovek-Maleck and
effective length method (Kanchanalai 1977) as well as the elastic White 2004b). This set of frames is expanded and generalized
second-order approach within the American Concrete Institute for this paper and a variety of composite cross sections are selected
(ACI) code (ACI 2011; Hage and MacGregor 1974). (each cross section is used within each benchmark frame to provide
However, to date no appropriate reduced elastic rigidity values a comprehensive set of results). The composite frames and cross
have been developed nor has the direct analysis methodology in sections used in this paper were also used to evaluate and compare
general been thoroughly validated for steel-concrete composite current design methodologies according to the AISC specification
columns. To address this current design need, a large parametric (2010b) and ACI code (2011) in prior research (Denavit et al. 2014).
study investigating the stability behavior of small, nonredundant
1
Cross Sections
Design Engineer, Stanley D. Lindsey and Associates, Ltd., 2300
Windy Ridge Pkwy SE, Suite 675S, Atlanta, GA 30339 (corresponding The cross sections chosen for investigation in this paper are cat-
author). E-mail: mdenavit@[Link] egorized into four groups: (1) circular concrete-filled steel tubes
2
CDM Smith Professor and Chair, Dept. of Civil and Environmental (CCFTs), (2) rectangular concrete-filled steel tubes (RCFTs),
Engineering, Northeastern Univ., Boston, MA 02115. E-mail: [Link]@ (3) steel reinforced concrete (SRC) subjected to strong axis bend-
[Link] ing, and (4) SRC subjected to weak axis bending. Within these
3
Professor, Departamento de Materiales, Universidad Autónoma Metro- groups, sections were selected to span practical ranges of concrete
politana, Mexico City 02200, Mexico. E-mail: tperea@[Link] strength, steel ratio (ρs ¼ As =Ag , where As is the area of the steel
4
D.H. Burrows Professor, Via Dept. of Civil and Environmental
section and Ag is the gross area of the composite section), and
Engineering, Virginia Tech, Blacksburg, VA 24061. E-mail: rleon@[Link]
Note. This manuscript was submitted on January 21, 2015; approved on reinforcing ratio (ρsr ¼ Asr =Ag , where Asr is the area of reinforcing
September 4, 2015; published online on October 30, 2015. Discussion per- bars) for the SRC sections (CFTs with longitudinal reinforcing bars
iod open until March 30, 2016; separate discussions must be submitted for are excluded in this paper). Other section properties (e.g., steel
individual papers. This paper is part of the Journal of Structural Engineer- yield strength) were taken as typical values. Steel yield strengths
ing, © ASCE, ISSN 0733-9445. were selected as Fy ¼ 345 MPa (50 ksi) for wide-flange shapes,

© ASCE 04015157-1 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 Table 1. Selected Steel Shapes and Reinforcing Configurations (Adapted
from Denavit et al. 2014, © ASCE)
Section ρs or
type Index Steel shape or reinforcing configuration ρsr (%)
CCFT A HSS177.8 × 12.7ðHSS7.000 × 0.500Þ 24.8
B HSS254 × 12.7ðHSS10.000 × 0.500Þ 17.7
C HSS323.9 × 9.5ðHSS12.750 × 0.375Þ 10.6
D HSS406.40 × 6.4ðHSS16.000 × 0.250Þ 5.7
E HSS609.6 × 3.2ðHSS24.000 × 0.125Þa 1.9
RCFT A HSS152.4 × 152.4 × 12.7ðHSS6 × 6 × 1=2Þ 27.6
B HSS228.6 × 228.6 × 12.7ðHSS9 × 9 × 1=2Þ 19.1
C HSS203.2 × 203.2 × 6.4ðHSS8 × 8 × 1=4Þ 11.1
HSS228.6 × 228.6 × 3.2ðHSS9 × 9 × 1=8Þ
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

D 5.0
E HSS355.6 × 355.6 × 3.2ðHSS14 × 14 × 1=8Þa 3.3
SRC A W360 × 463ðW14 × 311Þ 11.7
(shape) B W360 × 347ðW14 × 233Þ 8.7
C W360 × 179ðW12 × 120Þ 4.5 Fig. 1. Schematic of the benchmark frames (reprinted from Denavit
D W200 × 46.1ðW8 × 31Þ 1.2 et al. 2014, © ASCE)
SRC A 20 #36 (#11) 4.0
(reinforcing) B 12 #32 (#10) 1.9
C 4 #25 (#8) 0.4 and slender sections. However, local buckling is neglected in this
a study, both by not modeling it in the inelastic analyses and by not
Not a standard section.
including the strength reductions in the design strength calculations.
Thus, the results of this study are only strictly applicable to com-
Fy ¼ 290 MPa (42 ksi) for round hollow structural section (HSS)
pact sections. Nonetheless, it is expected that the recommendations
shapes, Fy ¼ 317 MPa (46 ksi) for rectangular HSS shapes, and
developed in this study can be extended to noncompact and slender
Fysr ¼ 414 MPa (60 ksi) for reinforcing bars. The AISC specifica-
sections when the appropriate local buckling strength reductions
tion (2010b) prescribes a lower limit of 21 MPa (3 ksi) and an upper
are applied. This is consistent with the development of the direct
limit of 70 MPa (10 ksi) on the concrete strength. Three concrete
analysis method for steel structures where local buckling was also
strengths were selected: f c0 ¼ 27.6, 55.2, and 110.3 MPa (4, 8, and
neglected (Surovek-Maleck and White 2004b).
16 ksi). The highest concrete strength exceeds the AISC limit but
was included to broaden the comparison and to ensure applicability
of the results should the limit be altered in the future. Frames
There is no prescribed upper limit of steel ratio for composite A set of small nonredundant frames were described and used in
sections within the AISC specification (2010b); however, practical previous stability studies on structural steel members (Kanchanalai
considerations and the dimensions of commonly produced steel 1977; Surovek-Maleck and White 2004b). The set includes both
shapes impose an upper limit of approximately 25% for CFT sidesway-inhibited and sidesway-uninhibited frames and a range
and 12% for SRC, although higher ratios can be obtained with of slenderness, end constraints, and leaning column loads. The
built-up sections. The AISC specification (2010b) sets a lower limit set of frames was expanded and the frame parameters were gener-
of steel ratio for composite sections of 1%. However, a stricter limit alized for use with composite sections in this study. The frames are
is imposed by the maximum permitted width-to-thickness ratios for shown schematically in Fig. 1. The sidesway-uninhibited frame is
CFT members. For the steel yield strengths listed previously, the described by a slenderness value (λoe1g ), which defines the length
width-to-thickness limits correspond to steel ratio limits of 1.86% of the column, a pair of end restraint parameters (Gg;top and Gg;bot ),
for CCFT and 3.16% for RCFT. For SRC members, the AISC which define the stiffness of the rotational spring at the top and
specification (2010b) prescribes a minimum reinforcing ratio of bottom of the column, and the leaning column load ratio (γ), which
0.4% and no maximum. defines the amount of load allocated to the leaning column. The
Noting these limitations, five round HSS shapes were selected sidesway-inhibited frame is defined by the same slenderness value
for the CCFT sections, five rectangular HSS shapes were selected (λoe1g ) and the end moment ratio (β), which defines the relative
for the RCFT sections, and for the SRC sections with outside di- values of moment applied at each end. The values of these param-
mensions of 711 × 711 mm (28 × 28 in:), four wide-flange shapes eters selected for the frames are described in Tables 2 and 3; a total
and three reinforcing configurations were selected (Table 1). For of 80 frames were selected. The subscript g in the end restraint
the HSS shapes, the design thickness, equal to 0.93 times the nomi- parameters and slenderness value denotes that these values are de-
nal thickness, was used for all calculations. The reinforcing steel fined with respect to gross section properties.
was assumed to have a cover from the edge of the concrete to
the edge of the bar of 48 mm (1-7=8 in:) and was placed symmet-
Second-Order Elastic Analysis of Benchmark Frames
rically within the section grouped in the corners with a center-to-
center spacing between the bars of 2.5 times the diameter of the The second-order elastic analysis results described in this paper
bars. Altogether, five steel shapes with three concrete strengths were obtained from the solution of the governing differential equa-
means 15 total sections were selected each for RCFT and CCFT, tion [Eq. (1)] using the appropriate boundary conditions (Table 4)
and four steel shapes with three reinforcing configurations and and the coordinate system as shown in Fig. 1. Closed-form solu-
three concrete strengths means 36 total sections were selected each tions were obtained for displacement and moment along the length
for strong and weak axis bending of SRC. of the column using a computer algebra system. This approach is
With the selected CFT sections, the full range of permitted steel computationally expeditious and accurate for moderate displace-
ratios is examined, including those associated with noncompact ments; however, only flexural deformations are included. Where

© ASCE 04015157-2 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 Table 2. Benchmark Frame Variations (Reprinted from Denavit et al. 2014, © ASCE)
Frame Slenderness End restraint Leaning column load ratio End moment ratio Number of frames
Sidesway-uninhibited 4 values 4 value pairs 4 values N/A 64 (¼ 4 × 4 × 4)
λoe1g ¼ f0.22; 0.45; 0.67; 0.90g (Table 3) γ ¼ f0,1; 2,3g
Sidesway-inhibited 4 values N/A N/A 4 values 16 (¼ 4 × 4)
λoe1g ¼ f0.45; 0.90; 1.35; 1.90g β ¼ f−0.5; 0.0; 0.5; 1.0g

Table 3. End Restraint Value Pairs (Reprinted from Denavit et al. 2014, in moment frames will include some bearing components or through
© ASCE) bolts internal to the composite member that will limit slip in the con-
nection region. In addition, Hajjar et al. (1998) have shown that con-
Pair Gg;top Gg;bot
nection slip, where it is allowed, rarely impacts the global response
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

A 0 0 of composite structural systems. As noted previously, local buckling

B 1 or 3 a
1 or 3a of the steel tube and other steel components was also neglected.
Elastic–perfectly plastic constitutive relations and the Lehigh
C 0 ∞ residual stress pattern (Galambos and Ketter 1959) were used to
D 1 or 3 a
∞ model the wide-flange steel shapes. Reinforcing steel was also
modeled as elastic–perfectly plastic, but assumed to have negligible
a
3 when γ ¼ 0; 1 otherwise.
residual stress. Residual stresses in cold-formed steel tubes vary
through thickness. To allow a reasonable fiber discretization of the
CFT sections, residual stresses were included in the constitutive
Table 4. Benchmark Frame Boundary Conditions
relation. A multilinear constitutive relation was used in which the
Boundary stiffness decreases at 75, 87.5, and 100% of the yield stress to
condition Sidesway-uninhibited Sidesway-inhibited approximate the gradual transition into plasticity observed in cold-
1 vð0Þ ¼ 0 vð0Þ ¼ 0 formed steel (Abdel-Rahman and Sivakumaran 1997). In addition,
2 −EI elastic v 0 0 ð0Þ ¼ −kθ;bot v 0 ð0Þ −EI elastic v 0 0 ð0Þ ¼ M the yield stress in the corner region of the rectangular members was
3 −EI elastic v 0 0 0 ðLÞ − Pv 0 ðLÞ vðLÞ ¼ 0 increased to account for the additional work hardening in that region.
¼ H þ γP L vðLÞ The Popovics concrete model was selected, with the peak com-
4 −EI elastic v 0 0 ðLÞ ¼ kθ;top v 0 ðLÞ −EI elastic v 0 0 ðLÞ ¼ βM pressive stress taken as fc0 or greater to account for confinement
(Denavit and Hajjar 2014). Spalling behavior was incorporated into
the model for the cover concrete of SRC sections by overriding the
necessary, the effective length factor (K) for the benchmark frames stress-strain response with linear degradation to zero stress starting
was back-calculated from the critical load obtained using the same at two times the strain at peak stress. The modulus of elasticity, Ec ,
differential equation and boundary conditions used in the analysis was calculated by Eq. (2) taken from the ACI
code (2011) for normal-weight concrete. Eq. (2) is equivalent to the
P expression in the AISC specification (2010b) when the weight of
v 0 0 0 0 ðxÞ þ v 0 0 ðxÞ ¼ 0 ð1Þ
EI elastic concrete per unit volume is 2,372 kg=m3 (148.1 lbs=ft3 ). Other ex-
pressions for the modulus of elasticity have been proposed as more
where v = lateral deflection; P = axial compression; and EI elastic = accurate for high-strength concrete and could have been used in the
elastic flexural rigidity of the column. analysis. However, it is not the intention of this paper to develop
design recommendations that implicitly include a correction for any
potential inaccuracies in the code specified formula for the concrete
Second-Order Inelastic Analysis of Benchmark Frames modulus; thus the modulus used in the analysis and in the design
The second-order inelastic analysis results described in this paper calculations was chosen to be the same
were obtained from finite-element analyses. Key aspects of the pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
model are summarized here and a full description is available else- Ec ½MPa ¼ 4,733 fc0 ½MPa ð2Þ
where (Denavit and Hajjar 2014). The columns were modeled using
a distributed plasticity mixed beam finite-element formulation imple- Nominal geometric imperfections equal to the fabrication and
mented in the OpenSees framework (McKenna et al. 2000). Geomet- erection tolerations in the AISC “Code of Standard Practice for
ric nonlinearity is captured using a total Lagrangian formulation Steel Buildings and Bridges” (AISC 2010a) were modeled explic-
assuming small strains and moderate rotations in the corotational itly. An out-of-plumbness of L=500 was included for the sidesway-
frame and is coupled with an accurate geometric transformation to uninhibited frames and a half-sine wave out-of-straightness with
the global frame. With multiple elements along the length of the col- maximum amplitude of L=1,000 was included for all frames. The
umn, large displacement and rotation behavior is captured accurately. initial out-of-plumbness and initial out-of-straightness were applied
Material nonlinearity is captured using a number of fiber cross in the same direction because this produced the greatest destabiliz-
sections along the length of each element. The uniaxial constitutive ing effect for these frames.
relations assigned to each fiber were selected to correspond to All frame analyses were performed with six elements along the
assumptions common in the development of design recommenda- length of the composite column, each with three integration points.
tions (e.g., neglecting steel hardening and concrete tension strength) Because the analyses were two-dimensional, strips were used for
and are comparable to those used in commensurate studies (Surovek- the fiber section; the nominal height of the strips was 1=30 the sec-
Maleck and White 2004b). Slip between the steel and concrete tion depth (e.g., for a CCFT section, approximately 30 steel and 30
elements was neglected. Although slip may be important in under- concrete strips of near-equal height were used). The leaning column
standing the local behavior of composite columns, most connections was modeled with a stiff truss element. The lateral deflections at the

© ASCE 04015157-3 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 top of the leaning column and composite column were set equal cross sections makes these differences more pronounced under
with a numerical constraint. Cross section analysis was performed weak-axis bending. These errors do not occur when the strain com-
in lieu of frame analysis for the case of zero applied axial load. patibility approach defined as an alternative approach in the AISC
Determination of the limit point was often the objective of the specification (2010b) and in the ACI code (2011) is used because in
analysis. The limit point was defined as the point in the analysis at these methods the compatibility of the steel and concrete is ad-
which the lowest eigenvalue of the system reached zero or when dressed explicitly. Overall, these unconservative errors are small;
the maximum longitudinal strain in either tension or compression however, they will be observed again in subsequent results exam-
at any point within any cross section along the length of the ining interaction strength.
member reached 0.05, whichever occurred first. The strain limit
was imposed because for some cases such as low or zero applied
axial load, the eigenvalue limit may only be reached at very high Axial Strength
deformations. In the AISC specification (2010b), the nominal axial compressive
The formulation has been validated against hundreds of exper- strength [Pn , Eq. (3)] is determined from a column curve based on
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

imental results from composite members under a variety of loading the nominal compressive strength of the column without length ef-
conditions and with a wide range of material and geometric prop- fects [Pno , Eq. (4) for SRC and Eq. (5) for compact CFT without
erties. These comparisons are presented elsewhere (Denavit and reinforcement] and the slenderness [λoe , Eq. (6)]. The slenderness
Hajjar 2014) along with more accurate constitutive relations that is a function of Pno , the effective length (KL), and the effective
include behavior (e.g., steel strain hardening and concrete tensile rigidity [EI eff , Eqs. (7) and (8) for SRC and Eqs. (9) and (10) for
strength) that was neglected for this study. CFT]. However, when utilized within the direct analysis method,
the axial compressive strength is not necessarily representative of
the maximum axial load permitted by the design methodology be-
Evaluation of Current Design Procedures cause required accounting of initial geometric imperfections can
impart bending moments that reduce the axial strength. To assess
Flexural Strength the maximum permitted axial load, a second-order elastic analysis
with reduced elastic rigidity and accounting of initial geometric im-
The AISC specification (2010b) allows for the flexural strength of
perfections (e.g., either through the use of notional loads or through
compact composite columns to be computed by the plastic stress
direct modeling of the imperfections) must be run to determine the
distribution method. In this method, the steel components are as-
maximum applied downward vertical load that results in required
sumed to have reached a stress of Fy in either tension or compres-
axial compression and bending moment that remain within the
sion and the concrete components are assumed to have reached a
strength interaction diagram
stress of 0.85fc0 in compression (or 0.95fc0 for CCFT to account

for confinement). Pn 2
0.658λoe when λoe ≤ 1.5
The flexural strength obtained from the inelastic analyses ¼ ð3Þ
Pno 0.877=λ2oe when λoe > 1.5
(Manalysis ) for each section is compared with the nominal strength
from AISC (2010b) (M nðAISC 2010Þ ) in Fig. 2 (noting that in this
study local buckling is neglected). A maximum of 5% unconser- Pno ¼ Fy As þ Fysr Asr þ 0.85f c0 Ac ðSRCÞ ð4Þ
vative error is desired for beam-column design methodologies
(ASCE 1997); this limit is shown in the figure as a dashed line. Pno ¼ Fy As þ C2 fc0 Ac ðCFTÞ ð5Þ
For CFT and SRC members in strong-axis bending, the nominal
strength from AISC (2010b) is either accurate or conservative. For sffiffiffiffiffiffiffiffiffiffi
SRC members in weak-axis bending, the nominal strength from KL Pno
λoe ¼ ð6Þ
AISC (2010b) is mildly unconservative for some sections, particu- π EI eff
larly steel dominant sections, overpredicting the strength by up to
8%. The primary cause of error is the deformation compatibility of
the steel and concrete, and the fact that the curvature required to EI eff ¼ Es I s þ 0.5Es I sr þ C1 Ec I c ðSRCÞ ð7Þ
activate the plastic stress distribution in the steel section assumed  
in design is large compared with the curvature at which the concrete As
C1 ¼ 0.1 þ 2 ≤ 0.3 ð8Þ
reaches its peak moment. The distribution of steel within SRC Ac þ As

1.15
Manalysis/Mn (AISC 2010)

1.10

1.05

1.00

0.95
0.90
0 0.10 0.20 0.30 0 0.10 0.20 0.30 0 0.05 0.10 0.15 0 0.05 0.10 0.15
ρs ρs ρs+ρsr ρs+ρsr
(a) (b) (c) (d)

Fig. 2. Flexural strength comparison: (a) CCFT; (b) RCFT; (c) SRC (strong axis); (d) SRC (weak axis)

© ASCE 04015157-4 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 EI eff ¼ Es I s þ Es I sr þ C3 Ec I c ðCFTÞ ð9Þ
 
As
C3 ¼ 0.6 þ 2 ≤ 0.9 ð10Þ
Ac þ As

where Ac = area of concrete; C2 = 0.85 for RCFT and 0.95 for

CCFT; I s = moment of inertia of the steel section; I sr = moment
of inertia of reinforcing bars; I c = moment of inertia of concrete;
and Es = modulus of elasticity of steel = 200,000 MPa (29,000 ksi).
The commentary of the AISC specification (2010b) recom-
mends the reduced rigidity of composite columns for determining
the required strength in the direct analysis method be computed by
applying the 0.8τ b reduction (as for structural steel) to EI eff
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

[Eq. (11)]. The stiffness reduction parameter (τ b ) was developed

for structural steel members and depends on the required axial
strength (Pr ) and the axial yield strength (Py ), a parameter that
is undefined for composite members. For the purposes of this study,
Py is taken as Pno , resulting in τ b given by Eq. (12)
EI DA ¼ 0.8τ b EI eff ð11Þ
Fig. 3. Example interaction diagram and force trace (adapted from
 Denavit et al. 2014, © ASCE)
1.0 for Pr =Pno ≤ 0.5
τb ¼ ð12Þ
4ðPr =Pno Þð1 − Pr =Pno Þ for Pr =Pno > 0.5

In the direct analysis method, initial imperfections must be ac- “A” and “B”). The frames (Fig. 1, sidesway-uninhibited) from
counted for either through direct modeling or with representative which the two force traces were derived have the same cross section
notional loads. The imperfections were directly modeled in the sec- (RCFT with HSS228.6 × 228.6 × 12.7 and f c0 ¼ 55.2 MPa) and
ond-order inelastic analyses (because it is a more exact approach), column length (λoe1g ¼ 0.67), thus they have the same beam-
while they were represented with notional loads in the second-order column strength interaction diagram. The frames differ only in
elastic analyses (because this approach is far more common in the stiffness of the boundary conditions (Gg;top ¼ Gg;bot ¼ 0 for
design). In each analysis in which the notional load was used, a A and Gg;top ¼ Gg;bot ¼ 1 for B) and the magnitude of the leaning
notional lateral load equal to 0.2% of the vertical load was included. column load (γ ¼ 0 for A and γ ¼ 3 for B). Within the direct analy-
According to Section C2.2b(4) of the AISC specification (2010b), sis method, the effects of boundary conditions and leaning columns
the notional load was taken as a minimum lateral load when the are expected to be captured primarily by the second-order elastic
ratio of maximum second-order drift to maximum first-order analysis in the form of required moment. This is in contrast to the
drift was less than or equal to 1.7 and as an additive lateral load
effective length method in which the effective length factor and
otherwise.
thus the axial strength plays a more direct role in capturing the ef-
The commentary of the AISC specification (2010b) describes
fects. One consequence of this difference is the manner in which a
several methods of determining the beam-column interaction
particular column appears to fail. The maximum permitted axial
strength for composite columns. One method in particular, the plas-
load (Pmax ) occurs where the internal force point trace intersects
tic stress distribution approach, is used in this study. In this method,
the nominal beam-column strength interaction diagram and the cor-
a set of points are computed based on the cross-section strength:
Point A is the pure axial strength, Point B is the flexural strength, responding value including resistance factors (Pmax;ϕ ) occurs where
Point C corresponds to a plastic neural axis location that results in the internal force point trace intersects the design beam-column
the same flexural strength as Point B, and Point D corresponds to strength interaction diagram. As will be shown subsequently, this
the plastic neutral axis location that results in an axial compressive second value is useful when comparing it with strengths that have
strength one-half of that determined for Point C and represents the also had resistance factors applied. Despite both columns having
maximum moment capacity. The axial component of each of the only vertical load applied, the column with softer boundary
points is factored down by the ratio Pn =Pno to obtain the nominal conditions and higher leaning column load (B) intersects the inter-
beam-column strength. The points are further factored down by the action diagram with high required moment. Additionally, the differ-
resistance factors to obtain the available beam-column strength. ence between Pmax and Pmax;ϕ is much smaller for B due to the high
These factoring rules and the shape of typical interaction curves level of geometric nonlinearity. This would not be the case for
can lead to an illogical and potentially unsafe situation in which either of these observations using the effective length method
the factored Point D lies outside the original section strength inter- because the beam-column strength interaction diagram would be
action curve. Because of this, as recommended in the commentary reduced significantly for B.
of the AISC specification (2010b), Point D is neglected, resulting in The critical axial load obtained from the second-order inelastic
a bilinear interaction diagram defined by Points A, C, and B. This analyses (Panalysis ) for each section and frame pair is compared
interaction diagram is termed the A-C-B interaction in this paper. with the maximum permitted axial load from the AISC method
More consistent methods of applying stability reductions and resis- (PmaxðAISC 2010Þ ) as a function of slenderness (λoe ) in Fig. 4. In ad-
tance factors to avoid the potentially large conservative errors from dition to this comparison with nominal strengths, a comparison
neglecting Point D are recommended for future research. with available strengths is also presented in Fig. 4. In the compar-
An example beam-column interaction diagram is shown in Fig. 3 isons with available strengths, Panalysis is multiplied by a resistance
along with two examples of internal force point traces from elastic factor (ϕc ¼ 0.75) and Pmax;ϕ is used in lieu of Pmax . The resistance
analyses as described previously (the lines denoted as Force Trace factor for compression is applied to Panalysis because only axial

© ASCE 04015157-5 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 /Pmax (AISC 2010)
1.6

1.4

1.2
analysis

1.0

0.8
P
φPanalysis/Pmax,φ (AISC 2010)

1.6

1.4
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

1.2

1.0

0.8

0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4
λoe (AISC 2010) λoe (AISC 2010) λoe (AISC 2010) λoe (AISC 2010)

(a) (b) (c) (d)

Fig. 4. AISC (2010b) axial strength comparison (reprinted from Denavit et al. 2014, © ASCE): (a) CCFT; (b) RCFT; (c) SRC (strong axis); (d) SRC
(weak axis)

loads were applied to the frame in the analysis, with bending This fact indicates that the resistance factor applied to the interac-
moments due only to initial imperfections. From a historical per- tion curve is not effective at reducing the maximum permitted
spective, prior to the requirement for notional loads, the required applied loads. Stiffness reduction factors are effective for these
strength for these cases would not include any moment and the cases; however, the current stiffness reduction [Eq. (11)] was cali-
axial load in the column would be compared with the design brated for structural steel, which has a higher resistance factor for
compression strength (ϕc Pn ). Thus, the comparison between compression than composite (ϕc ¼ 0.9 versus ϕc ¼ 0.75). To alle-
ϕc Panalysis and Pmax;ϕ in Fig. 4 is meant to ensure that the design viate this error, a stiffness reduction on the order of 0.65 [≈0.877ϕc
strength for the sidesway-uninhibited frames is not excessivly (Surovek-Maleck and White 2004b)] would be more appropriate.
liberal when compared with historical approaches or the simply
supported columns for which the current resistance factor was
Interaction Strength
originally developed (Leon et al. 2007). Similar comparisons were
made in the validaion of the direct analysis method for structural Axial compression-bending moment interaction strength is repre-
steel (Surovek-Maleck and White 2004b). The dashed line indi- sented not with single values, but a curve that identifies the strength
cates 5% unconservative error. of the beam-column under combined loading varying from pure
A wide range of behavior can be seen in the results of Fig. 4. bending (typically plotted on the horizontal axis) to pure axial (typ-
First, within the nominal strength comparison, most cases are ically plotted on the vertical axis). To construct this curve, several
shown to be conservative, with the inelastic analysis indicating that points on the curve at varying axial loads are determined and
the strength is higher than that allowed by the design methodology. straight lines are assumed between the points. The process to de-
The results are generally accurate for the CFT beam columns, with termine the individual points is similar to that used to determine the
some slight unconservative error for intermediate slenderness axial strength as described in the previous section. Typical results
CCFTs and for both RCFTs and CCFTs of very high slenderness. are presented in Fig. 5 for an example benchmark frame (RCFT
For all but the stockiest SRC beam columns, the results are very with HSS355.6 × 355.6 × 3.2, f c0 ¼ 27.6 MPa, sidesway-uninhib-
conservative, indicating a significant underprediction in strength ited, λoe1g ¼ 0.67, γ ¼ 1, and Gg;top ¼ Gg;bot ¼ 0).
by the design methodology. This is due to the effective stiffness The interaction strength according to the second-order inelastic
[Eq. (7)] and in particular the C1 value [Eq. (8)], which are both analysis is constructed by selecting a number of axial load values
likely lower than necessary because they were based on the limited between Panaysis and zero. At each of these values a separate analy-
range of experimental test data available without the benefit of sis is performed using a nonproportional loading pattern in which
analyses, such as presented here, which place the experimental re- the axial compression is applied then held constant, while the
sults in a broader context. lateral load is applied until a limit point is determined. Both the
The strength ratios at the available strength level are lower than applied loads (Curve 1 in Fig. 5) and internal forces (Curve 2 in
those at the nominal strength level. The reason for this can be seen Fig. 5) are recorded at the limit point. Only slight differences are
in Fig. 3. While a constant reduction is applied to Panalysis , the dif- noted between limit points obtained from nonproportional analyses
ference between Pmax;ϕ and Pmax is not constant, as discussed pre- and those from proportional analyses. A similar procedure was
viously, and for frames in which the geometric nonlinear effects are used to experimentally determine the interaction strength of con-
dominant (such as the case of Force Trace B in Fig. 3), Pmax;ϕ and crete-filled steel tube beam columns by Perea et al. (2014).
Pmax can be similar in value (although the corresponding available The second-order interaction strength according to the design
axial compression and bending moment strengths are reduced). methodology is determined from design equations. In Fig. 5 the

© ASCE 04015157-6 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 results of Fig. 6 highlight the worst-case maximum unconservative
0.8 error, which is useful for the following discussion, but hide the
Normalized Axial Compression (P/P )
majority of cases, in which no unconservative error is evident.
no 0.7 The comparisons of Fig. 6 are performed at the nominal strength
level, not including any resistance factors in either the inelastic
0.6 analysis or elastic design methodology.
The most striking results in Fig. 6 are the large unconservative
0.5 errors for very slender concrete dominant frames with CFT cross
sections [lower right corner of Figs. 6(a and c)]. These errors can be
0.4 attributed to changes in shape of the strength interaction curve that
are not taken into account in the design equations. Interaction
0.3 strength curves of composite cross sections and short composite
beam column are quite convex, particularly for concrete-dominant
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

0.2 members. With increases in length, the interaction strength curves

become much less convex and often concave. This change is caused
0.1 by reductions in the flexural rigidity that occur due to material
nonlinearity (primarily concrete cracking but also concrete crush-
0
0 0.5 1 1.5 ing and steel yielding) that occur prior to obtaining the peak
Normalized Bending Moment (M/M ) strength. The effect is greater for more slender columns because
n
the second-order effects are greater, but also because the ratio of
Fig. 5. Comparison of interaction diagram results (adapted from bending moment to axial load is greater, a condition that leads
Denavit et al. 2014, © ASCE) to greater reductions in effective stiffness. This behavior has also
been observed experimentally (Perea et al. 2014). The slenderness
used to categorize results in Fig. 6 is based on the effective length
(KL), so high slenderness can either be caused by long unsupported
current nominal A-C-B interaction diagram is denoted as Curve 3. lengths (L) or by high effective length factors (K) from either soft
Curve 4 in Fig. 5 represents the maximum permitted applied loads boundary conditions or high leaning column load. The larger errors
using the current design methodology. Under these applied are not observed for the SRC columns due to comparatively low
loads, the required axial and flexural strengths determined from EI eff values, indicating that reducing the flexural rigidity in the
a second-order elastic analysis with stiffness reductions and no- elastic analysis would be an effective way to reduce these errors.
tional loads (where appropriate) lie on the nominal strength inter- Other unconservative errors seen in Figs. 6(a and c) are smaller
action diagram. and are typically from either the unconservative error in the axial
In Fig. 5, Curve 4 is the envelope of applied loads that are strength of intermediate slenderness steel-dominant CCFT columns
deemed safe by the design methodology. Curve 1 is the envelope as seen previously or sidesway-inhibited single-curvature cases.
of applied loads that are demonstrated to be safe by the inelastic The latter cases are challenging due to the fact that cracking and
analysis. Thus, regions in which Curve 4 is outside of Curve 1 the accompanying stiffness reduction will occur along the entire
are considered unconservative. length of the column as opposed to just the member ends.
To quantify error in the design methodology (ε), a radial mea- Again, little unconservative error is noted for SRC columns
sure defined by Eq. (13) is selected, where rinelastic is the distance [Figs. 6(e and g)], with the exception of the moment strength
from the origin to the maximum applied loads determined by sec- of steel-dominant SRC columns bent about the weak axis, as dis-
ond-order inelastic analysis and rdesign is the distance along the cussed previously.
same line to the maximum applied loads permitted by the design
methodology. The error varies along the curves, making it neces-
sary to compute its value at many points (i.e., for many angles be- Proposed Changes to the AISC Specification
tween the horizontal and vertical axis). Unconservative error by
Eq. (13) is represented with negative values To address the most salient unconservative errors identified in the
rinelastic − rdesign previous section, two changes are proposed to the AISC specifica-
ε¼ ð13Þ tion. The first change is to adopt the stiffness reduction recom-
rinelastic mended in the commentary to the AISC specification [Eq. (11)],
but define the parameter τ b to equal 0.8 in all cases for composite
The interaction strength comparison results are shown in
columns [Eq. (14)]. The value 0.8 was selected in part because it
Figs. 6(a, c, e, and g) for the various section types. For these results
results in a total stiffness reduction approximately equal to 0.877ϕc ,
the entire suite of benchmark frames has been sorted into a number
just as with the reduction for structural steel (Surovek-Maleck and
of bins based on their steel ratio and slenderness [Eq. (6)]. For each
White 2004a)
of these bins, the maximum unconservative error in the high axial
load–low bending moment range, high bending moment–low axial τ b ¼ 0.8 ð14Þ
load range, and intermediate range has been determined and dis-
played. The unconservative errors are shown as positive values The second change is to modify the formulas for EI eff for both
in this figure. For example, in Fig. 6(c), for benchmark frames with SRC and CFT columns [Eqs. (15)–(18)]. This change eliminates
RCFT cross sections with steel ratio ρs ¼ 0.28 and slenderness the 0.5 factor on the contribution from the steel reinforcing and
(λoe ) between 2.0 and 3.0, the maximum unconservative error is alters the concrete contribution factors C1 and C3 . Analyses have
0% for cases of high axial load and low bending moment (leftmost shown a constant concrete contribution factor (i.e., C1 ¼ 0.7 and
number), indicating that no unconservative error was found, 0.5% C3 ¼ 0.9) can accurately capture axial strength (Denavit and Hajjar
for cases of high bending moment and low axial load (rightmost 2014). However, the poorer behavior of more concrete dominant
number), and 3.5% for intermediate cases (center number). The columns under axial compression and bending moment justifies

© ASCE 04015157-7 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

© ASCE
ρs = 0.01
ρs = 0.04
ρs = 0.09
ρs = 0.12
ρs = 0.01
ρs = 0.04
ρs = 0.09
ρ = 0.12
ρs = 0.03
ρs = 0.05
ρs = 0.11
ρs = 0.19
ρs = 0.28
ρs = 0.02
ρs = 0.06
ρs = 0.11
ρs = 0.18
ρs = 0.25

axis), proposed
λ oe
0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%

≤0
.5
0% 2.7% 12.1% 13.6% 0% 0% 3.3% 6.2% 3.1% 4.9% 6.2% 6.2% 3.7% 5.0% 7.1% 3.8% 4.9% 5.7%
2.1% 2.6% 7.6% 11.3% 2.1% 0.7% 1.5% 1.4% 0% 0% 0.0% 1.5% 0% 0% 0% 0% 2.5% 2.0%

0.5
<λ o
0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 1.4% 2.0% 1.4% 0% 5.1% 8.7% 14.3% 15.5%

e
≤1
0% 7.5% 10.7% 14.8% 0% 0% 4.2% 6.4% 1.5% 8.5% 8.9% 8.5% 8.4% 3.7% 9.3% 13.7% 15.5% 15.4%

.0
2.1% 2.6% 7.5% 11.9% 2.1% 0.7% 1.5% 1.4% 0.6% 2.9% 6.3% 6.0% 3.4% 0% 7.7% 9.9% 10.9% 10.8%

1.0
<λ
0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0.3% 0.5% 1.6% 0% 0% 5.1% 9.3% 10.2%

oe
≤1
0% 0% 5.6% 6.8% 0% 0% 0% 0.3% 9.6% 9.2% 9.6% 8.3% 5.6% 1.3% 13.7% 14.9% 15.2% 15.1%

.5
2.1% 2.6% 8.5% 10.3% 2.1% 0.7% 1.5% 1.4% 12.1% 10.5% 9.1% 8.0% 5.2% 4.0% 14.9% 14.4% 14.5% 14.7%

1.5
<
(c)

(g)
(e)
(a)

λ
0% 0% 0% 0% 0% 0% 0% 0% 0% 1.3% 0.2% 0% 0% 0% 0% 3.4% 4.6% 1.5%

oe
≤2
0% 0% 0.4% 3.7% 0% 0% 0% 0% 10.8% 16.3% 12.7% 10.8% 8.0% 16.2% 11.3% 18.9% 18.9% 17.9%

.0
2.1% 5.0% 6.3% 12.5% 2.1% 0.7% 1.8% 1.4% 8.9% 17.3% 13.4% 10.2% 7.2% 16.3% 9.9% 20.2% 19.0% 17.5%

2.0
<λ o
0% 0% 0% 0% 0% 0% 0% 0% 11.6% 6.4% 6.1% 1.1% 0% 21.1% 6.5% 7.0% 5.5% 2.2%

e
≤3
0% 0% 0% 2.5% 0% 0% 0% 0% 23.0% 17.8% 13.4% 7.8% 3.5% 35.2% 20.4% 17.3% 15.5% 12.6%

.0
3.2% 7.5% 11.5% 12.9% 3.0% 2.7% 5.4% 3.8% 21.3% 14.7% 9.1% 4.3% 0.5% 29.5% 21.0% 14.3% 12.5% 10.2%

0% 0% 0% 0% 0% 0% 0% 0% 27.1% 19.7% 9.2% 2.6% 0% 39.7% 18.5% 12.1% 7.2% 3.0%

3.0
0% 0% 0% 2.3% 0% 0% 0% 0% 33.4% 25.6% 15.2% 8.7% 4.0% 54.5% 27.5% 21.0% 16.5% 13.2%

<λ oe
5.2% 8.0% 6.8% 7.8% 5.0% 6.6% 1.5% 1.4% 26.1% 18.8% 10.0% 4.8% 0.8% 44.4% 21.8% 16.5% 13.1% 10.6%

04015157-8
λ oe
0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 1.4% 0% 0% 0.4%

≤0
.5
0% 5.6% 13.3% 17.3% 0% 1.1% 5.5% 7.2% 0.2% 2.7% 3.9% 5.0% 2.6% 5.4% 6.0% 2.8% 2.4% 3.0%
2.1% 2.6% 8.5% 12.8% 2.1% 0.7% 1.5% 1.4% 0% 0% 0% 0.3% 0% 0% 0% 0% 1.0% 0.5%

0.5

J. Struct. Eng., 2016, 142(3): 04015157

<λ o
0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0.8% 3.9% 6.8% 12.4% 14.6%

e
≤1
0% 7.5% 14.7% 15.8% 0% 0% 3.9% 6.0% 0% 1.6% 3.2% 3.9% 1.7% 6.3% 5.7% 4.8% 6.9% 9.5%

.0
2.1% 2.6% 10.8% 15.2% 2.1% 0.7% 1.5% 1.4% 0% 0% 2.1% 2.4% 0% 0% 3.6% 5.5% 6.2% 6.3%

1.0
<λ
0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%

oe
≤1
0% 0% 10.3% 15.7% 0% 0% 0% 0% 0% 0.5% 2.6% 1.8% 0% 0% 4.4% 6.8% 7.6% 7.9%

.5
2.1% 6.2% 9.9% 14.9% 2.1% 0.7% 1.5% 1.4% 0% 4.6% 4.6% 3.2% 0.4% 0% 8.5% 9.3% 10.1% 10.0%

1.5
(f)

<
(h)
(d)
(b)

λ
0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%

oe
≤2
0% 2.0% 10.3% 14.1% 0% 0% 0% 0% 0.6% 0% 4.9% 3.6% 0.9% 0% 2.9% 9.7% 10.7% 9.7%

.0
4.5% 10.2% 12.4% 14.1% 4.2% 3.6% 1.9% 2.4% 4.7% 0.9% 8.1% 5.6% 2.4% 6.4% 6.3% 13.9% 13.7% 12.2%

2.0
<λ o
0% 0% 0% 0.5% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%

e
≤3
0% 5.4% 10.4% 13.8% 0.5% 0% 0.5% 0% 9.1% 6.8% 4.7% 0.3% 0% 12.2% 9.3% 8.3% 7.3% 4.7%

.0
7.9% 11.3% 13.0% 14.3% 12.2% 9.7% 6.4% 4.4% 11.5% 9.7% 3.9% 0% 0% 17.1% 13.0% 8.0% 6.9% 4.6%

19.4% 6.7% 2.8% 4.7% 16.9% 0% 0% 0% 3.0% 0% 0% 0% 0% 8.4% 0% 0% 0% 0%

3.0
24.0% 10.6% 6.5% 9.3% 20.9% 11.7% 3.0% 1.3% 17.7% 12.8% 6.0% 1.0% 0% 28.1% 14.8% 11.1% 8.1% 5.2%

<λ oe
24.3% 10.3% 6.5% 8.9% 25.1% 11.7% 3.0% 1.4% 16.2% 11.1% 4.8% 0.2% 0% 28.8% 13.6% 10.2% 7.5% 5.0%

Fig. 6. Maximum unconservative error of interaction strength: (a) CCFT, AISC (2010b); (b) CCFT, proposed; (c) RCFT, AISC (2010b);
(d) RCFT, proposed; (e) SRC (strong axis), AISC (2010b); (f) SRC (strong axis), proposed; (g) SRC (weak axis), AISC (2010b); (h) SRC (weak

J. Struct. Eng.
 the further reduction for these sections. The specific values in C1 methodology, however, cases of such high slenderness are rare
and C3 were selected to obtain good results with little unconserva- in practice.
tive error in the following comparison. The proposed formulas re- The key difference between CFT and SRC columns that neces-
present a significant increase in the EI eff for SRC columns and a sitates the difference in EI eff [i.e., Eq. (15) versus Eq. (17)] is the
slight decrease in the EI eff for some CFT columns. Also, the steel distribution of steel within the composite cross sections. In SRC
ratio variable in C1 and C3 has been modified such that the total columns, the steel is typically more centrally located and there ex-
steel area (including both the steel shape and the reinforcing) is in ists cover concrete. The selected sections are representative of a
the numerator and the gross composite area is in the denominator wide variety of typical columns, but all have cover equal to the
minimum allowed (ACI 2011). If cover in excess of the minimum
EI eff ¼ Es I s þ Es I sr þ C1 Ec I c ðSRCÞ ð15Þ be used, the strength of the column may decrease and a smaller
value of C1 may be appropriate.
 
As þ Asr
C1 ¼ 0.25 þ 3 ≤ 0.7 ð16Þ
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

Ag Interaction Strength
Interaction strength comparison results are shown in Figs. 6(b, d, f,
EI eff ¼ Es I s þ Es I sr þ C3 Ec I c ðCFTÞ ð17Þ and h) for the various section types. The proposed changes are ef-
fective at reducing the maximum unconservative errors for CFT.
  Similar to the axial strength results, the proposed changes introduce
As þ Asr
C3 ¼ 0.45 þ 3 ≤ 0.9 ð18Þ greater maximum unconservative errors for the SRC columns in
Ag some of the ranges of steel ratio and slenderness. The error in
the weak axis flexural strength of steel-dominant SRC columns re-
mains. The largest unconservative errors for all section types occur
Axial Strength with the most slender frames, specifically those frames where
The proposed changes do not alter the flexural strength, so axial Pmax =Pno < 0.15, in other words, under gravity-only loading, the
strength will be examined first. Axial strength comparison results columns in these frames can support less than 15% of their squash
are shown in Fig. 7, which was constructed in the same manner as load. This ratio can be used by engineers to identify stability-
Fig. 4 but using the two proposed changes in the design method- sensitive structures. When the maximum permitted axial loads
ology. These results show that the proposed changes are effective at are this low, it would be advisable to use further stiffness reduc-
both reducing the discrepancy between CFT and SRC columns and tions to avoid excessive unconservative error. One option
reducing the unconservative error seen in the available strength is to use τ b ¼ 0.4 in place of Eq. (14). Another option is to cal-
comparison for high slenderness CFT columns. The error for in- culate the stiffness for the composite member as though it
termediate slenderness steel-dominant CCFT columns remains were bare steel. Both of these options were found to reduce the
because the proposed changes were not specifically designed to unconservative error to acceptable levels or eliminate it
address that error. The proposed changes introduce some unconser- altogether. However, care should be taken when selecting an
vative error to high slenderness SRC columns for the available elastic stiffness that differs significantly from the expected stiff-
strength comparison (i.e., with reduction factors applied); this is ness because unrealistic distributions of moment in the beams
an unfortunate consequence of reducing the overall error in the and columns can arise.
/Pmax (Proposed)

1.6

1.4

1.2
analysis

1.0
P

0.8
φPanalysis/Pmax,φ (Proposed)

1.6

1.4

1.2

1.0

0.8

0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4
λoe (Proposed) λoe (Proposed) λoe (Proposed) λoe (Proposed)

(a) (b) (c) (d)

Fig. 7. Proposed axial strength comparison: (a) CCFT; (b) RCFT; (c) SRC (strong axis); (d) SRC (weak axis)

© ASCE 04015157-9 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 13.2% 13.1%
8.2% 11.8%

13.5% 14.7%
10.5% 14.4%

11.8%
13.1%

13.0%

12.3% 13.2%
Alternative Stiffness Reduction

9.6%

6.2% 8.7%

8.4%
8.3%
ρs =

0%

0%

0%

0%

0%

0%
Due to the cumbersome nature of a stiffness reduction that varies 0.12
with axial load, provisions in the AISC specification (AISC 2010b)

9.2%
9.3%

6.3%
6.7%

6.2%
7.5%
ρs =
allow the engineer to use τ b ¼ 1.0 if an additional 0.1% of the

0%

0%

0%

0%

0%

0%
gravity load is applied as a notional load. The proposed τ b for 0.09
composite columns [Eq. (14)], however, does not vary with load,

3.2%
2.6%

7.5%
2.6%

2.6%

5.4%

8.2%

1.9%
9.0%
ρs =

0%

0%

0%
0%

0%
0%

0%
0%

0%
making such a substitution both less necessary and less accurate. 0.04
The additional notional load required to reduce the strength by an

2.1%

2.1%

2.1%

2.1%

3.4%

6.9%
amount equivalent to the effect of using τ b ¼ 0.8 is on the order of ρs =

0%
0%

0%
0%

0%
0%

0%
0%

0%
0%

0%
0%
1% for cases in which elastic buckling controls. Such high addi- 0.01
tional notional loads would be overly conservative in other frames. (a)
It is recommended that the use of τ b ¼ 1.0 in exchange for a higher

15.8% 19.6%
10.7% 15.7%

21.3% 23.6%
15.5% 21.9%

16.0% 27.7% 29.1%

14.9% 23.3% 27.3%

6.6% 11.9%
24.1% 29.8% 31.4%
22.3% 27.1% 29.5%

17.2% 21.1% 23.8%

13.3% 28.5% 30.1% 32.3%
18.0% 26.8% 28.2% 29.2%

38.6% 31.9% 26.0% 26.6%

49.1% 27.1% 16.8% 18.6%
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

notional load not be permitted for composite columns. For the case

1.5% 7.8%
ρs =

0%

0%

0%
of a structure containing both composite columns and highly 0.12
loaded (Pr > 0.5Py ) steel columns, a conservative approach to
avoid a variable stiffness would be to apply the additional notional ρs =

0%

0%

0%
load so that τ b ¼ 1.0 can be used for the steel columns and main- 0.09
tain τ b ¼ 0.8 for the composite columns.

7.6%
2.6%

8.3%
5.1%
ρs =

0%

0%

0%

0%

19.2% 0%
0.04
Effective Length Method

2.1%

1.0%
2.1%

0.0%
2.1%

6.6%
7.6%
ρs =

0%
0%

0%

0%

0%

0%
The effective length method was the primary method of design in
the AISC specification (AISC 2010b) prior to the direct analysis 0.01
method, and is still allowed as an option to the engineer for cases (b)
in which the ratio of maximum second-order drift to maximum

13.0%
13.0%

10.3% 15.4%
11.9% 15.4%

10.4% 14.8%
12.1% 15.2%
2.7%
7.8%

8.4%
8.9%

1.7%

4.2%

2.4% 8.6%

9.7%
9.3%
first-order drift is less than or equal to 1.5. The effective length ρs =

0%

0%

0%
method differs from the direct analysis method in that no stiffness 0.12
reduction is used, notional loads are minimum lateral loads for all

6.3%

2.8%
6.3%

8.6%
9.7%

7.4%
7.3%
ρs =
cases (because the effective length method is limited to cases in 0%
0%

0%

0%

0%

0%
which the ratio of maximum second-order drift to maximum 0.09
first-order drift is less than 1.5), and the compressive strength is 2.6%

2.6%

4.5%

1.3%
7.0%

3.5%
8.3%

23.0% 9.8%
25.9% 9.7%
ρ =
based on the effective length, KL. The use of this method was stud-
0%
0%

0%
0%

0%
0%

0%

0%

0%
s
0.04
ied in a similar fashion to what has been shown here for the direct
analysis method. The proposed EI eff [Eqs. (15)–(18)] was used for
2.1%

2.1%

2.1%

2.1%

4.8%
ρs =
0%
0%

0%
0%

0%
0%

0%
0%

0%
0%

0%
the elastic stiffness in the second-order elastic analysis, and the K 0.01
factor was determined from the governing differential equation (c)
[Eq. (1)] as noted previously. Minimal unconservative error was
.0 .5 .0 .0
found within the axial strength results, however, the maximum un- ≤0
.5 ≤1 ≤1 ≤2 ≤3 <λ oe
λ oe λ oe λ oe λ e λ e 3.0
.5< .0< < <
o o
conservative error in the intermediate and high moment ranges was 0 1 1.5 2.0
found to be greater than for the direct analysis method, particularly
for the high slenderness cases. Interaction strength comparison re- Fig. 8. Maximum unconservative error of interaction strength for the
sults for SRC columns bent about the weak axis are presented in effective length method: (a) SRC (weak axis), AISC (2010b), A-C-B
Fig. 8. The data in Fig. 8 are presented in the same manner as in interaction; (b) SRC (weak axis), proposed, A-C-B interaction; (c) SRC
Fig. 6, with the exception that cases in which the ratio of maximum (weak axis), proposed, Section H1.1 interaction
second-order drift to maximum first-order drift is greater than 1.5
are excluded. Had these cases not been excluded, the maximum
errors would be greater. Other section types showed similar trends Illustrative Example
in the results.
These increased errors can be attributed to the change in shape With the aim of broad applicability, this study has focused on ab-
of the interaction strength as the column slenderness increases. stract frames and results have been presented in normalized terms.
Without stiffness reductions to account for the inelasticity To aid in the understanding of the results, this section presents prac-
(e.g., concrete cracking and partial steel yielding) that occurs prior tical examples. The column cross section for all the examples are
to the ultimate load being reached under axial compression plus the same, a rectangular CFT column constructed of an HSS203.2 ×
bending moment, an interaction strength diagram based on section 203.2 × 9.5 (HSS8 × 8 × 3=8) of typical strength [Fy ¼ 317 MPa
strength is too convex. To alleviate this error, it is recommended (46 ksi)] and filled with normal-strength concrete [f c0 ¼ 34.5 MPa
that when using the effective length method, the interaction dia- (5 ksi)]. The steel ratio of this section is 16.3%, an intermediate
gram should be taken as described in Section H1.1 of the AISC value for CFT members. The effective stiffness is EI eff ¼
specification (AISC 2010b) instead of the A-C-B interaction dia- 10,786 kN · m2 ð3.76 × 106 kip · in:2 ) according to Eq. (17) and
gram described and used previously. As seen in Fig. 8, using the the nominal section compressive strength is Pno ¼ 3,127 kN
interaction diagram from Section H1.1, the maximum unconserva- (703 kips). If this member was used as a leaning column or a col-
tive errors are approximately equal to those for the direct analysis umn in a braced frame (K ¼ 1) with a floor height of 4.27 m (14 ft),
method. Alternatively, in lieu of using the interaction diagram from then the slenderness would be λoe ¼ 0.731 [Eq. (6)]. While not
Section H1.1, the elastic stiffness could be taken as a value less than stocky, this slenderness is far from the range in which the highest
EI eff , although an appropriate value would need to be calibrated. unconservative errors were determined (e.g., λoe > 2.5 as seen in

© ASCE 04015157-10 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 that would allow the inclusion of the balance point would be
beneficial, especially for short concrete dominant columns. Inclu-
sion of the balance point in the interaction diagram would require a
reevaluation of the stiffness reduction and potential reduction
of EI DA .
The contribution of creep and shrinkage to structural instability
was not addressed in this study. The ACI code (ACI 2011)
combines and addresses these effects with the use of beta factors
when determining the elastic flexural rigidities. A detailed study is
warranted to examine how long-term effects on composite columns
Fig. 9. Multibay industrial frame can be addressed within the AISC specification (AISC 2010b).
Only flexural deformations were considered in the elastic analy-
ses performed in this study, thus only the flexural rigidity (EI) was
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

directly addressed. However, in the general case other rigidities

Fig. 6). To achieve a very high slenderness, λoe ¼ 2.5, the effective
contribute to the stability of the structure and their values should
length of the column would need to be KL ¼ 14.59 m (47.85 ft).
be carefully selected.
At this effective length the slenderness reduction is Pn =Pno ¼ 0.14
In this study, an expression for the elastic flexural rigidity was
according to the column curve in the AISC specification (AISC
validated for use in the execution of strength checks. However, this
2010b) [Eq. (3)]. This is an impractical height if K ¼ 1; however,
is only one use for EI in the design process. Expressions for EI for
with leaning column loads the effective length could be achieved
other uses, including drift checks and determination of natural
with much smaller unbraced lengths. For example, consider a
frequencies and mode shapes, need to be thoroughly validated.
cantilever column with a perfectly rigid connection, two leaning
columns, and all three columns supporting the same axial load.
For this case, K ¼ 3.25 determined using the solution of the gov- Conclusions
erning differential equation [Eq. (1)] as was typically done to de-
termine K in this paper. Thus, λoe ¼ 2.5 is achieved when the This paper presents the results of a large parametric study under-
column height is 4.49 m (14.7 ft). At this length the maximum taken to assess the in-plane stability behavior of steel-concrete
permitted applied load under a gravity-only load case is Pmax ¼ composite columns, evaluate current second-order elastic design
305.6 kN (68.7 kips), resulting in the ratio Pmax =Pno ¼ 0.098, be- provisions within the AISC specification, and propose changes to
low the 0.15 recommended limit at which unconservative errors those provisions where necessary and justified. Comparisons were
may be significant. As an alternative example, consider a variation made between second-order inelastic analysis results, deemed suf-
of the 11-bay industrial frame (Fig. 9) previously analyzed by ficiently accurate to form the basis of design recommendations, and
others (Surovek-Maleck and White 2004a; Deierlein 2003; second-order elastic analysis results, representative of the analyses
Martinez-Garcia 2002), where λoe ¼ 2.5 is achieved when H ¼ an engineer performs as part of the methodology.
5.59 m (18.4 ft). The effective length for this case was determined In general, the design methodology was found to be safe and
as K ¼ 2.61 using the story buckling approach described in the accurate for the majority of cases. However, some unconservative
commentary of the AISC specification (AISC 2010b) (the differ- errors were identified, particularly for concrete-dominant members
ential equation approach used elsewhere in this paper is not appli- with high slenderness effects. Some significant conservative errors
cable to this case). At this length the maximum permitted applied were also identified in which the effective flexural rigidity of SRC
load under a gravity-only load case is Pmax ¼ 441.4 kN (99.2 kips), members was underestimated in the design provisions.
resulting in the ratio Pmax =Pno ¼ 0.141, again below the 0.15 rec- New effective flexural rigidities for calculating the axial
ommended limit at which unconservative errors may be significant. compressive strength and new direct analysis stiffness reductions
While these are not typical cases, they are feasible. It is recom- were proposed to eliminate the most salient unconservative and
mended that should a case such as this arise, further stiffness re- conservative errors observed in the current design provisions.
ductions should be applied as discussed previously. The proposed beam-column design methodology is safe and accu-
rate for the vast majority of cases of composite member behavior,
although further research is recommended to continue to investi-
Further Work gate the axial compressive strength of steel-dominant intermediate
slenderness CCFTs, the weak-axis flexural strength of steel-dominant
A number of aspects were not addressed in this study and further SRCs, and the long-term behavior of composite columns.
work on these will be necessary to bring the level of comprehen-
siveness of the design provisions for steel-concrete composite
members to that of either structural steel or reinforced concrete. Acknowledgments
First, the proposed changes described in this paper have been cali-
brated to the results of a second-order inelastic analysis model that The authors are grateful for the advice received from AISC
was validated against experimental results. Validation directly of Committee on Specifications Task Committee 5 on Composite
experiments would increase confidence in the proposals. Addition- Construction and Task Committee 10 on Stability, as well as Pro-
ally, a reliability study is necessary to verify that the current resis- fessors Donald White, Ronald Ziemian, Gregory Deierlein, and
tance factors are adequate for the new proposals. Andrea Surovek regarding this work. This material is based on
The use of the A-C-B interaction diagram, which neglects the work supported by the National Science Foundation under Grant
balance point, is recommended in the commentary of the AISC Nos. CMMI-0530756 and CMMI-0619047 as part of the George
specification (AISC 2010b) because of inconsistencies that can E. Brown, Jr. Network for Earthquake Engineering Simulation
arise when applying a stability reduction and resistance factors. (NEES), the American Institute of Steel Construction, Georgia In-
However, the A-C-B interaction diagram is very conservative in stitute of Technology, University of Illinois at Urbana-Champaign,
certain cases. More consistent methods for applying the reductions and Northeastern University.

© ASCE 04015157-11 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157

 References Hajjar, J. F., Schiller, P. H., and Molodan, A. (1998). “A distributed plas-
ticity model for concrete-filled steel tube beam-columns with interlayer
Abdel-Rahman, N., and Sivakumaran, K. S. (1997). “Material properties slip.” Eng. Struct., 20(8), 663–676.
models for analysis of cold-formed steel members.” J. Struct. Eng., Kanchanalai, T. (1977). “The design and behavior of beam-columns in
10.1061/(ASCE)0733-9445(1997)123:9(1135), 1135–1143. unbraced steel frames.” CESRL Rep. No. 77-2, Structures Research
ACI (American Concrete Institute). (2011). “Building code requirements Laboratory, Dept. of Civil Engineering, Univ. of Texas at Austin,
for structural concrete and commentary.” Farmington Hills, MI. Austin, TX.
AISC. (2010a). “Code of standard practice for steel buildings and bridges.” Leon, R. T., Kim, D. K., and Hajjar, J. F. (2007). “Limit state response of
Chicago. composite columns and beam-columns. Part 1: Formulation of design
AISC. (2010b). “Specification for structural steel buildings.” Chicago. provisions for the 2005 AISC specification.” Eng. J., 44(4), 341–358.
ASCE. (1997). “Effective length and notional load approaches for assessing Martinez-Garcia, J. M. (2002). “Benchmark studies to evaluate new pro-
frame stability: Implications for American steel design.” Reston, VA. visions for frame stability using second-order analysis.” M.S. thesis,
Deierlein, G. G. (2003). “Background and illustrative examples of pro-
Dept. of Civil and Environmental Engineering, Bucknell Univ.,
posed direct analysis method for stability design of moment frames.”
Downloaded from [Link] by University of Texas at Austin on 03/19/18. Copyright ASCE. For personal use only; all rights reserved.

Lewisburg, PA.
Rep. to Task Committee 10, AISC, Chicago.
McKenna, F., Fenves, G. L., and Scott, M. H. (2000). “Open system for
Denavit, M. D., and Hajjar, J. F. (2014). “Characterization of behavior of
earthquake engineering simulation.” Univ. of California, Berkeley, CA,
steel-concrete composite members and frames with applications for
design.” Newmark Structural Laboratory Rep. Series, Newmark 〈[Link]
Structural Laboratory Rep. NSEL-034, Univ. of Illinois at Urbana- Perea, T., Leon, R. T., Hajjar, J. F., and Denavit, M. D. (2014). “Full-scale
Champaign, Urbana, IL. tests of slender concrete-filled tubes: Interaction behavior.” J. Struct.
Denavit, M. D., Hajjar, J. F., Leon, R. T., and Perea, T. (2014). “Analysis Eng., 10.1061/(ASCE)ST.1943-541X.0000949, 04014054.
and design of steel-concrete composite frame systems.” Structures Surovek-Maleck, A. E., and White, D. W. (2004a). “Alternative approaches
Congress 2014, ASCE, Reston, VA, 2605–2616. for elastic analysis and design of steel frames. I: Overview.” J. Struct.
Galambos, T. V., and Ketter, R. L. (1959). “Columns under combined bend- Eng., 10.1061/(ASCE)0733-9445(2004)130:8(1186), 1186–1196.
ing and thrust.” J. Eng. Mech. Div., 85(2), 1–30. Surovek-Maleck, A. E., and White, D. W. (2004b). “Alternative approaches
Hage, S. E., and MacGregor, J. G. (1974). “The second-order analysis of for elastic analysis and design of steel frames. II: Verification studies.”
reinforced concrete frames.” Structural Engineering Rep. No. 49, Dept. J. Struct. Eng., 10.1061/(ASCE)0733-9445(2004)130:8(1197),
of Civil Engineering, Univ. of Alberta, Edmonton, AB, Canada. 1197–1205.

© ASCE 04015157-12 J. Struct. Eng.

J. Struct. Eng., 2016, 142(3): 04015157