Collapse Resistance of Steel Columns under

Combined Axial and Lateral Loading

J. Fogarty, S.M.ASCE 1; and S. El-Tawil, Ph.D., P.E., F.ASCE 2

Abstract: The collapse resistance of steel columns subjected to combined axial and reversed lateral loading is investigated computation-
ally. Detailed finite-element models that are validated using available experimental data are presented. Simulation results using these
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

models suggest that local buckling as well as lateral torsional buckling in deep, slender sections under combined axial and lateral loading
can lead to a substantial reduction in column ductility. The results of an extensive parametric study suggest that the current design guide-
lines are potentially unconservative and are used to develop a simple-to-use design aid. DOI: 10.1061/(ASCE)ST.1943-541X.0001350.
© 2015 American Society of Civil Engineers.
Author keywords: Steel column; Lateral drift; Collapse; Seismic; Metal and composite structures.

Introduction fully address column behavior. In fact, the columns in the subas-
semblages tested were not subjected to axial loading.
The 1994 Northridge earthquake shook the faith of the structural When used in tall moment-resisting frames, lower floor column
engineering community in then-prevalent seismic design provi- members can be subjected to high axial loads due to seismically
sions for steel buildings. The unexpected nonductile failure of induced overturning moments. When deep columns are used, their
seismically designed steel moment connections motivated the lower torsional characteristics compared to, for example, W14 sec-
SAC Steel Project, a joint research venture between the Struc- tions, make them especially vulnerable to lateral torsional instabil-
tural Engineering Association of California (SEAOC), Applied ity under combined axial and lateral loading. This is important
Technology Council (ATC), and Consortium of Universities for because column integrity during seismic events is critical in avoid-
Research in Earthquake Engineering (CUREE). Among the many ing a vertical progressive collapse scenario. Past research on ver-
issues that arose during the research effort was that then-existing tical progressive collapse has focused primarily on the vulnerability
steel column provisions were deficient. Major issues identified in- of reinforced concrete columns (Moehle et al. 2002; Setzler and
cluded: (1) the need to account for material overstrength to ensure Sezen 2008; Wu et al. 2010) due to the lack of ductile detailing in
strong-column–weak-beam behavior; (2) concerns about weak panel structures built prior to modern seismic design guidelines. In con-
zone response; and (3) the increased fracture potential for beam- trast, much less research has been devoted to steel columns, and
flange-to-column-flange welds for columns with thick flanges. little is known about the degradation in axial capacity that they may
Before Northridge, almost all steel columns used in seismic suffer as a result of lateral seismic loading. Only a few researchers
zones were W14s. Such columns were popular because they have have recognized this problem and, as a result, there is limited data
a small footprint, thereby maximizing architectural flexibility. available about it.
However, the new provisions that emerged from the SAC Project
made it challenging to continue to design columns using W14 sec-
tions because their small depth made it difficult to satisfy the new Previous Research
and more stringent design requirements. To alleviate this problem, Newell and Uang (2006) experimentally investigated nine full-
engineers turned to deeper steel columns, but that trend generated scale W14 steel sections meeting the American National Standards
new concerns regarding their ability to resist the torsional demands Institute (ANSI)/AISC 341-05 (ANSI/AISC 2010) seismic com-
that arise when plastic hinges form in connected beams. Experi- pactness requirements under a combined axial compression and
mental research by Ricles et al. (2003) and Chi and Uang (2002) biaxial flexure loading sequence. In their tests, an initial gravity
showed that the concerns were unfounded and led the way to wide- load was applied, then followed by in-phase, increasing amplitude
spread use of deep steel columns in special moment-resisting steel cyclic axial load and story drift. Flange local buckling was ob-
frames. While the studies discussed alleviated concerns about served in all but one specimen, but the level of buckling was con-
the torsional vulnerability of deep columns in a moment-resisting sidered insignificant at drift levels far exceeding the maximum
connection, these studies and others (e.g., Shen et al. 2002) did not expected drift. The authors concluded that this was due to the
stocky nature of the chosen columns. Macrae et al. (1990) exper-
1
Ph.D. Candidate, Dept. of Civil and Environmental Engineering, Univ. imentally investigated eight small-scale W10 × 49 columns sub-
of Michigan, Ann Arbor, MI 48109 (corresponding author). E-mail: jected to axial load ratios (P=Py ) ranging from 0.0 to 0.8 as well
[email protected] as cyclic strong axis bending. The specimens all experienced sig-
2
Professor, Dept. of Civil and Environmental Engineering, Univ. of
nificant local flange buckling but displayed ductility greater than
Michigan, Ann Arbor, MI 48109. E-mail: [email protected]
Note. This manuscript was submitted on February 26, 2015; approved 0.02–0.03 rad.
on May 7, 2015; published online on June 25, 2015. Discussion period Cheng et al. (2013) subjected six deep wide-flange cantilever
open until November 25, 2015; separate discussions must be submitted sections to a constant axial load combined with a cyclic lateral dis-
for individual papers. This paper is part of the Journal of Structural placement up to 0.02 rad and then monotonic lateral displacement
Engineering, © ASCE, ISSN 0733-9445/04015091(12)/$25.00. until failure. The failure mechanism for each section was attributed

© ASCE 04015091-1 J. Struct. Eng.

J. Struct. Eng.
 to local instabilities. The authors also noted that individual plate Table 1. Current AISC Seismic Design Guidelines (Data from ANSI/
limits common in most current design specifications, and AISC AISC 2010)
(2010) in particular, may not adequately predict the behavior of Limiting
the sections due to interaction between the flange and web ele- Width-to- width-to-thickness
ments. Zargar et al. (2014) investigated a 1:8 scale W36 × 652 sub- thickness ratio, λhd (highly
jected to large drifts under high axial demands. The dominant Description of element ratio ductile members)
pffiffiffiffiffiffiffiffiffiffiffi
modes of failure for this particular section under various cyclic Flanges of rolled I-shaped b=t 0.30 E=Fy
and monotonic loading histories are lateral-torsional buckling sections
and web local bucking. The Cheng et al. (2013) and Zargar et al. Webs of rolled I-shaped h=tw For Ca ≤ 0.125
sections used as columns pffiffiffiffiffiffiffiffiffiffiffi
(2014) studies are the only existing experimental studies the au- 2.45 E=Fy ð1 − 0.93Ca Þ
thors could find that specifically focused on deep steel sections sub- For Ca > 0.125
jected to combined axial and lateral loading, highlighting the lack pffiffiffiffiffiffiffiffiffiffiffi
0.77 E=Fy ð2.93 − Ca Þ
of data in this area. More importantly, the fact that the deep steel pffiffiffiffiffiffiffiffiffiffiffi
≥1.49 E=Fy
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

members in these studies experienced significant local buckling

and failure under drifts less than 0.04 rad shows that there is a need where Ca ¼ Pu =ϕc Py ðLRFDÞ
for further investigation. Note: LRFD = load and resistance factor design.
Newell and Uang (2006) computationally investigated the
behavior of three W27 sections and noticed severe local flange
and web buckling. Significant strength degradation was noticed Ductility Limits
as a result of the local buckling and several of their models were 10
unable to complete the applied loading sequence. Elkady and
9
Lignos (2015) also computationally investigated deeper sections
under a combined axial and cyclic loading scheme, focusing on 8
the axial shortening and precapping rotations of typical first-story
interior columns. They recommend a reduction in the current AISC 7

high ductility compactness limits for sections used in lower stories 6

b/2tf Ratio
of special moment systems, but note that their results are dependent
on the applied loading and boundary conditions and do not con- 5
sider different member lengths.
4
Prior work using computational and analytical methods indi-
cates that any interruption of the load path in a column, e.g., due 3
to local flange damage, can lead to a significant decrease in the
axial resistance of wide-flange steel columns (Fogarty et al. 2
2013; Fogarty and El-Tawil 2013). Local buckling, which has been 1
observed to occur under cyclic flexural loading (Krishnan and
Muto 2012; Lamarche and Tremblay 2011; Suita et al. 2007, 0
0 10 20 30 40 50 60
2008; Newell and Uang 2006), can create such interruptions, po-
h/tw Ratio
tentially leading to a severe reduction in axial load-carrying capac-
ity. This observation, coupled with the limited information about Fig. 1. Selected columns’ slenderness ratios with respect to seismic
the behavior of deep steel members subjected to combined axial design guideline limits
loading and lateral drift, motivated the research presented herein.
The specific research objectives are (1) to investigate the relation-
ship between column axial capacity and lateral drift, and (2) assess
the adequacy of the current AISC seismic design guidelines (AISC
2010) pertaining specifically to highly ductile members, as de- is the ratio of the required strength, Pu , to available strength,
fined next. Py ¼ Fy Ag , multiplied by the strength reduction factor for com-
pression, ϕc ¼ 0.90. Limitations for the width-to-thickness ratios,
λhd for highly ductile members, are shown in Fig. 1 for the
specific material properties of the members used in this study (dis-
Current AISC Seismic Design Guidelines
cussed subsequently). The current member requirements for the
To choose appropriate sections to investigate for the first objective width-to-thickness ratios are based on small-scale, compact
and achieve the second objective, a summary of the current design sections subjected to axial compression and pure bending under
guidelines, which are based on individual plate slenderness limita- monotonic—but not cyclic—loading (Lay 1965; Kemp 1986;
tions, is useful. General requirements are outlined in Chapter D of Dawe and Kulak 1986).
“Seismic Provisions for Structural Steel Buildings” (ANSI/AISC Based on the current design guidelines as well as the behavior
2010) for both moderately and highly ductile steel members and seen in the limited experimental studies conducted to date, the
are summarized for highly ductile column members in Table 1. parameters of interest include the flange and web width-to-
As defined in the specification (ANSI/AISC 2010), a highly thickness ratios, the length of the column, the level of applied axial
ductile member is one that is “expected to undergo significant plas- load, and the boundary conditions. The slenderness ratios of the
tic rotation (more than 0.04 rad) from either flexure or flexural column as well as variation in the applied axial load are of concern
buckling under the design earthquake.” Table 1 indicates that because they are the basis for the current member requirements.
the flange width-to-thickness (b=t) ratio is based solely on material The vulnerability of deeper sections to lateral torsional buckling
properties, E and Fy , while the web width-to-thickness (h=tw ) ratio indicates a need to study the length and boundary conditions of
is determined by these properties as well as Ca ¼ Pu =ϕc Py , which these members under a combined axial and cyclic loading scheme.

© ASCE 04015091-2 J. Struct. Eng.

J. Struct. Eng.
 Finite-Element Modeling plasticity model (LS-DYNA MAT_003) is used and calibrated with
available steel coupon data (Kaufman et al. 2001) as shown in
To investigate the collapse behavior of steel columns and the cur- Fig. 2. It is clear from the figure that the stiffness and hardening
rent AISC seismic design guidelines, a number of carefully selected response of the model matches reasonably well with experimental
wide-flange column members are subjected to a combined axial data. The model employed does not capture fracture, but it is
and cyclic lateral load protocol using finite-element simulation. deemed adequate for this particular study because fracture is not
expected to dominate response at the deformation levels of interest
Prototype Column Selection (around 4% drift).
Typical column cross sections found in the steel prototype build-
ings commissioned by the National Institute of Science and Length, Boundary Conditions, and Loading Scheme
Technology (NIST) and the SAC Steel Project are the basis for
Exterior first-story columns in moment-resisting bays are poten-
the selected sections in this study (Khandelwal et al. 2008). These
tially the most critical columns because they are likely to see
columns consist of sections that make up the lateral load-resisting
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

some of the largest axial loads due to overturning moments during

frame on the first or second story of a structure where axial loads
a seismic event. For this reason, the columns are either 3.05, 4.57,
are expected to be largest during a seismic event. Several other col-
or 6.10 m long to cover the range of typical first-story column
umns are chosen due to the availability of experimental data or in an
member heights used in the West Coast. In addition, the boundary
effort to provide a wider range of local slenderness ratios.
conditions of the finite-element models are chosen to idealize those
As seen in Fig. 1, the selected columns meet the highly ductile
of a first-story column. The bottom of the column is fixed while
seismic requirements (in the region designated TYPE I in Fig. 1),
the top has some released degrees of freedom as shown in Fig. 3.
are highly ductile for specific axial load ranges (TYPE II), or are on
The top of the column is free to translate in the x- and z-directions.
the cusp of being highly ductile (TYPE III). TYPE III members
It is also free to rotate about the x-axis. This is meant to simulate
exceed the highly ductile flange width-to-thickness ratio, but were
the member having a moment connection perpendicular to the
chosen to explore the current guidelines in the event that they are
strong axis and a shear tab connection parallel to the strong
found to be overly conservative.
axis. Rigid plate caps are attached at either end to distribute the
Based on the AISC (2010) guidelines, the W36 × 194 is slender
applied loads over the entire cross section without causing local
in compression due to the h=tw , but meets the seismic member
deformation.
requirements for a highly ductile member as long as the ratio of
The loading protocol entails application of a force-controlled
required strength to nominal axial strength, Ca , does not exceed
axial load in phase with a displacement-controlled lateral load.
0.64. The other columns listed under Type II in Table 2 are similar
The axial compressive load increases as the drift increases until
to the W36 × 194 with different ratios of required strength to avail-
the target load and drift are reached simultaneously. The displace-
able strength needed to meet the highly ductile requirements. This
ment then reverses direction and the axial load is reversed (poten-
is also shown in Fig. 1 by the fact that these columns are below
tially going into tension) to represent the loading of an actual
the horizontal solid line (flange width-to-thickness limit) and to
column experiences during a seismic event. Members are investi-
the right of the leftmost solid vertical line (h=tw limit for Ca ¼ 1).
gated under maximum compression loads ranging from 0.2Py to
The maximum expected value of P=Py occurs when Ca ¼ 1,
0.9Py ðPy ¼ Fy AÞ. The combined axial and cyclic loading scheme
implying that the ratio of axial demand to the squash load of
is shown in Fig. 4 where the maximum compression load is 0.5Py
the section is equal to ϕc , which is the resistance factor for com-
and the load fluctuates about an average of 0.15Py in compression
pression, or 0.90. This is assuming the compression strength of the
to represent the gravity load on the column. Table 3 summarizes
member is capable of reaching 0.90Py prior to buckling. To account
the story drift ratio and number of cycles for the loading scheme.
for buckling, the expected value for Type I and Type II members
This loading protocol is based on the basic loading history devel-
listed in Table 2 is the smaller of the values calculated using (1) the
oped by Krawinkler et al. (2000) for moment frames and modified
limits in Table 1, and (2) the nominal compressive strength equa-
by Newell and Uang (2006) for braced frames. The loading rate of
tions of Chapter E (AISC 2011) divided by Py. The interaction of
0.46 m=s is selected through a trial and error process as a compro-
the axial load and moment is neglected in determining the expected
mise to reduce computational run time while minimizing the dy-
axial load per ANSI/AISC (2010) Chapter D Section 4a(2) and a
namic effects associated with rapid loading.
conservative value of K ¼ 1 is used to determine the capacity of the
members as suggested in the specification. For Type III members,
an expected value of P=Py ¼ not highly ductile ðNHDÞ indicates Failure Criterion
that the member cannot be considered highly ductile under the cur-
Three failure criteria are used to define failure of the member under
rent AISC seismic design guidelines and is not expected to reach
this loading scheme: (1) 10% reduction in the lateral load; (2) 10%
4% drift under the combined loading scheme.
reduction in the moment at the base of the column, as discussed
next; and (3) complete loss in the ability to carry the applied axial
Model Setup load (collapse). The moment at the base of the column, M, is cal-
Detailed finite-element models of the cross sections shown in Fig. 1 culated as follows:
are created using the commercial software Hypermesh and ana- 1
lyzed using the commercial code LS-DYNA. The columns are M ¼ ðHL − PΔÞ ð1Þ
2
modeled using fully integrated shell elements (ELFORM 16) based
on the formulation described by Engelmann et al. (1989) for a where the axial load, P, is positive in tension, H is the lateral force
large-displacement small-strain context. The geometry is modified on the column, L is the length of the column, and Δ is the lateral
to include an initial imperfection that assumes a sinusoidal shape displacement at the top of the column.
with a maximum deflection of L=1,000 at midheight. Sensitivity The moment criterion has been used previously to define failure
studies are performed to ensure an appropriate mesh is selected (Newell and Uang 2006). The lateral load criterion was used by
for further investigations. A mixed isotropic-kinematic hardening Cheng et al. (2013) to investigate the structural behavior of the

© ASCE 04015091-3 J. Struct. Eng.

J. Struct. Eng.
 Table 2. Column Data Used for Regression Analysis and Comparison with Expected Axial Load
Critical axial Maximum
pffiffiffiffiffiffiffi Fez load ratio expected P=Py
Section h=tw b=t L=ry JSx =ho JI y =G (MPa) (finite element) (AISC guidelines)
Type I
W14 × 132 17.7 7.15 47.90 0.0043 0.0025 1,205 0.80 0.85
W14 × 176 13.7 5.97 44.75 0.0068 0.0046 1,515 0.70 0.86
W14 × 257 9.71 4.23 43.58 0.0131 0.0099 2,183 0.70 0.87
W14 × 500 5.21 2.43 40.67 0.0383 0.0377 4,627 0.80 0.89
W18 × 86 33.4 7.2 68.44 0.0014 0.0008 1,279 0.30 0.71
W18 × 86 33.4 7.2 45.63 0.0014 0.0019 672 0.50 0.86
W18 × 119 24.5 5.31 89.39 0.0026 0.0009 1,474 0.30 0.56
W18 × 119 24.5 5.31 67.04 0.0026 0.0016 841 0.40 0.72
W18 × 119 24.5 5.31 44.70 0.0026 0.0036 620 0.60 0.86
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

W21 × 93a 32.3 4.53 97.58 0.0015 0.0007 1,022 NMb 0.50
W21 × 111 34.1 7.05 62.18 0.0013 0.0013 1,538 0.50 0.75
W21 × 111 34.1 7.05 41.46 0.0013 0.0030 783 0.60 0.88
W24 × 176 28.7 4.81 59.14 0.0022 0.0033 1,822 0.50 0.63
W24 × 176 28.7 4.81 78.85 0.0022 0.0019 976 0.50 0.77
W24 × 176 28.7 4.81 39.42 0.0022 0.0074 680 0.60 0.89
W27 × 178 32.9 5.92 73.81 0.0015 0.0018 1,985 0.40 0.67
W27 × 178 32.9 5.92 36.91 0.0015 0.0073 996 0.50 0.80
W27 × 178 32.9 5.92 55.36 0.0015 0.0033 650 0.60 0.90
W30 × 326 23.4 3.75 66.71 0.0033 0.0062 2,621 0.30 0.72
W30 × 326 23.4 3.75 50.03 0.0033 0.0110 1,410 0.50 0.83
W30 × 326 23.4 3.75 33.35 0.0033 0.0248 986 0.80 0.90
W33 × 241 35.9 5.66 66.21 0.0013 0.0057 2,458 0.20 0.71
W33 × 241 35.9 5.66 49.65 0.0013 0.0057 1,195 0.30 0.82
W33 × 241 35.9 5.66 33.10 0.0013 0.0128 753 0.60 0.90
W36 × 487 21.4 3.19 60.38 0.0023 0.0133 2,944 0.40 0.76
W36 × 487 21.4 3.19 45.29 0.0023 0.0236 1,573 0.50 0.86
W36 × 487 21.4 3.19 30.19 0.0023 0.0530 1,093 0.80 0.90
W40 × 278a 33.3 3.31 71.41 0.0017 0.0057 795 NMb 0.69
Type II
W16 × 31a 51.6 6.28 153.8 0.0006 0.0001 208 NMb 0.13
W16 × 77 31.2 6.77 72.84 0.0017 0.0021 640 0.30 0.30
W24 × 55a 54.6 6.94 134.3 0.0004 0.0002 237 NMb 0.07
W24 × 76a 49.0 6.61 93.79 0.0007 0.0005 405 NMb 0.26
W27 × 129a 39.7 4.55 81.58 0.0012 0.0014 355 NMb 0.61
W27 × 146 39.4 7.16 56.14 0.0010 0.0022 1,876 0.30 0.72
W27 × 146 39.4 7.16 37.43 0.0010 0.0049 913 0.50 0.72
W30 × 124a 46.2 5.65 80.83 0.0008 0.0012 536 NMb 0.39
W33 × 130a 51.7 6.73 75.45 0.0006 0.0012 580 NMb 0.13
W33 × 201 41.7 6.85 50.60 0.0009 0.0039 2,327 0.40 0.61
W33 × 201 41.7 6.85 33.74 0.0009 0.0087 1,107 0.50 0.61
W36 × 194a 42.4 4.81 70.18 0.0009 0.0028 704 NMb 0.58
W44 × 335 38.0 4.50 51.57 0.0013 0.0092 1,187 0.30 0.81
Type III
W16 × 26a 56.8 7.97 161.1 0.0004 0.0000 176 NMb NHDc
W21 × 48a 53.6 9.47 108.7 0.0004 0.0002 303 NMb NHDc
W24 × 117a 39.2 7.53 61.26 0.0010 0.0014 783 NMb NHDc
W30 × 99a 51.9 7.80 85.83 0.0005 0.0007 467 NMb NHDc
W36 × 135a 54.1 7.56 75.61 0.0005 0.0012 584 NMb NHDc
a
Sections not used in regression analysis.
b
Critical axial load ratio is not measured (NM) because section did not reach 4% drift under 0.20Py.
c
Not highly ductile (NHD) member based on current design guidelines.

member due to second-order moments (M 2 ¼ PΔ) by comparing displacement) to right (negative displacement), the moment and
the total moment, M, with the moment due to only first-order ef- lateral force begin to degrade as seen in Fig. 5. Once the moment
fects (M 1 ¼ HL). The axial failure criterion, or inability to carry and lateral force drop 10% below their peak values as shown by the
the applied axial load, i.e., explicit collapse, is unique to this study solid line in Figs. 5(a and b), the furthest drift where this line in-
and used due to its obvious physical meaning. tersects the moment or force-displacement diagram corresponds to
To illustrate use of the three failure criteria, consider a 4.57- the failure of the column. The axial failure criterion is obtained by
m-long column with a W36 × 487 section. As the load fluctuates determining the time step at which the axial force is unable to reach
between tension (positive force) and compression (negative the target load as shown in Fig. 6(b) and choosing the largest drift
force) in phase with the lateral drift oscillating from left (positive reached prior to that time step [Fig. 6(a)]. This criterion is obvious

© ASCE 04015091-4 J. Struct. Eng.

J. Struct. Eng.
 during the simulation because the column collapses suddenly as the AISC seismic specifications (AISC 2010; ANSI 2010). With
indicated in Figs. 6(b and c). The results for this particular section emphasis on the axial load failure criterion in this paper, separating
for various levels of applied axial load are summarized in Table 4 elastic deformations from total deformation is not clear cut. The
and Fig. 7. problem is geometrically nonlinear, and so elastic deformations
are strongly dependent on the applied axial load history, which
varies during the loading protocol. Also, in some cases the failure
Measures of Performance point is low on the softening portion of the lateral load versus drift
There is no clear consensus in the AISC seismic specifications curve, making it unclear how to compute the elastic component of
(AISC 2010; ANSI 2010) regarding the use of total versus plastic deformations. To avoid confusion, lateral drift, which is equivalent
deformations as a measure of structural performance. For example, to total rotation for the column configurations used, is utilized in-
while member performance is described as a function of plastic ro- stead of plastic rotation to gauge structural performance in the
tation, connection response is specified as a function of total drift in AISC sense. As such, highly ductile columns are those that reach
drifts of at least 0.04 rad. Other researchers, such as Newell and
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

Uang (2006) and Fadden and McCormick (2014), have also used
600 total rotations in their work.
The critical axial load ratio (CALR) is defined as the maximum
400 axial load ratio that a member can sustain under the given loading
scheme and reach 4% lateral drift. It is a key measure of perfor-
Stress (MPa)

200
mance that specifies axial capacity at the limiting drift level for
0 ductile members and subassemblage according to the AISC seismic
specifications (AISC 2010). Determination of the appropriate P=Py
-200
value for the CALR of W36 × 487 is shown in Fig. 7 and is also
-400 bolded in Table 4. This table indicates that the top of the 4.57-m
W36 × 487 column is able to move laterally 229 mm, or 5% of the
-600 column height, while carrying 15,902 kN, or 50% of the section’s
-0.05 -0.03 -0.01 0.01 0.03 0.05
Strain (mm/mm) squash load (0.5Py ), before failing based on the axial load failure
criterion. If the load is increased to 0.6Py , the W36 × 487 is only
Fig. 2. Material model calibration with coupon data (data from
Kaufman et al. 2001)
Table 3. Loading Sequence after Initial Axial Gravity Load is Applied
Story drift ratio Number of cycles
Top Boundary Condition:
Free translation in x and z direction. 0.001 6
Free rotation about the x axis. 0.0015 6
All other conditions fixed. Force controlled axial load 0.002 6
0.003 4
0.004 4
Displacement 0.005 4
controlled lateral 0.0075 2
force 0.01 2
0.015 2
0.02a 1
0.03 1
0.04 1
0.05 1
Bottom Boundary Condition: 0.06 1
Fixed translation in all three directions. 0.07 1
Fixed rotation about all axes. 0.08 1
0.09 1
Fig. 3. Finite-element validation model showing mesh and boundary 0.10 1
conditions

0.15 0.30
Lateral Displacement (%)

0.10 0.15
Axial Load (P/Py)

0.05 0.00
0.00 -0.15
-0.05 -0.30
-0.10 -0.45
-0.15 -0.60
(a) Steps (b) Steps

Fig. 4. Loading protocol for (a) lateral displacement; (b) axial load

© ASCE 04015091-5 J. Struct. Eng.

J. Struct. Eng.
 W36x487 - 4.57 m - 0.2Py W36x487 - 4.57 m - 0.2Py
20000 8000

End Moment (kN-mm)

Lateral Force (kN)

10000 a 4000 a

0 0
-400 -200 b 0 200 400 -400 -200 b 0 200 400
-10000 -4000
10% Reduction from
10% Reduction from
Peak Moment
-20000 -8000 Peak Lateral Force
(a) Lateral Displacement (mm) (b) Lateral Displacement (mm)

Fig. 5. Failure criterion for 4.57 m W36 × 487-0.2Py column based on a 10% reduction in (a) moment; (b) lateral force
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

able to reach 3% drift and therefore the 4.57 m W36 × 487 should the ANSI/AISC 341-10 (ANSI/AISC 2010) seismic compactness
not be considered a highly ductile member under that loading. As requirements subjected to a combined axial compression and flex-
such, the CALR of this member is 0.5Py . ure loading sequence.

Finite-Element Model Validation Model Setup

Model validation is conducted using the Newell and Uang (2006) Two detailed finite-element models of the test setup are created
experimental data, which focused on full-scale specimens meeting and analyzed using the commercial code LS-DYNA. The models
consist of 4.57 m W14 × 132 and W14 × 176 columns subjected
to various axial and lateral drift loads. To address the fact that the
experimental setup is physically unable to achieve the ideal
case of full fixity that is feasible in a finite-element program, ver-
tical springs are added at the ends of the columns (as shown in
Fig. 8) to match the stiffness seen in the experimental plots of
end moment and lateral force versus lateral drift. As deployed,
the springs, each with a stiffness of 8,756 kN=mm, provide rota-
tional flexibility to both column ends. The upper column end is
allowed to translate vertically (z) and horizontally (x) as shown
in Fig. 8.
The material properties from a cyclic coupon test performed at
Lehigh University (Kaufman et al. 2001) are deemed an adequate
representation per Newell and Uang of the A992 steel member
specimens used in their experimental work. Therefore, steel with
a yield strength of 372 MPa, an ultimate strength of 796 MPa,
and an elongation of 36% are utlized for this validation study.
Residual stresses are not included in the computational study be-
cause Newell and Uang (2006) showed through their work that the
residual stresses have an insignificant impact on the behavior of the
column under the combined loading case.

Loading Scheme
The loading scheme consists of applying a displacement-controlled
lateral load in phase with a force-controlled axial load similar to
that seen in Fig. 4, which is consistent with the experimental tech-
niques used by Newell and Uang (2006). While the experimental
loading protocol is similar to that used in the finite-element para-
metric study, the two loads are not actually applied simulataneously
in phase due to limitations of the experimental setup. The physical
loading scheme consisted of laterally displacing the column to the
desired drift, then applying the axial displacement until the target
axial load was reached. This can be seen in Fig. 9 where there are
Fig. 6. Progression of axial load failure for 4.57-m W36 × 487-0.7Py large vertical jumps in the lateral force and end moment at constant
column as shown by (a) last half-cycle able to reach target displacement drift levels. This discrepancy in loading is replicated in the valida-
under target axial load; (b) inability to reach target axial load at target tion study by using a displacement-controlled lateral load followed
displacement; (c) sudden drop in axial load and increase in axial by a force-controlled axial load applied after reaching the target
displacement drift ratio.

© ASCE 04015091-6 J. Struct. Eng.

J. Struct. Eng.
 Table 4. Summary of Drift Reached Based on Different Failure Criteria under Various Axial Compressive Loads for a 4.57-m W36 × 487 Column
10% lateral force criterion 10% moment criterion Axial force criterion
P=Py Millimeters Percentage Millimeters Percentage Millimeters Percentage
0.2 174 3.80 183 4.00 320 7.00
0.3 137 3.00 165 3.60 274 6.00
0.4 110 2.40 128 2.80 274 6.00
0.5 82.3 1.80 86.9 1.90 229 5.00
0.6 54.7 1.20 54.4 1.30 137 3.00
0.7 22.8 0.40 32.0 0.70 91.4 2.00
0.8 22.8 0.40 22.8 0.40 68.6 1.50
0.9 9.09 0.20 9.09 0.20 22.9 0.50
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

1.00
4% Limit Simulation Results
0.90 A W14 × 132 column with an applied axial load that is 75% of the
column squash load (W14 × 132-75) and W14 × 176 column with
0.80
an applied axial load that is 55% of the column squash load
0.70 (W14 × 176-55) are investigated and comparison of the column re-
Axial Load (P/Py)

sponses are shown in Fig. 9. A comparison of the deformed shapes

0.60 CALR
Moment at 10% lateral drift is shown in Fig. 10. The different directions of
0.50 Failure Criterion the buckled flanges between the experimental setup and the finite-
Axial
element model are attributed to differences in initial imperfections.
0.40
Lateral Force Failure Criterion However, the location and length of local buckling is consistent
0.30 Failure Criterion with that seen in the experimental setup. As seen in the plots of
force-displacement, the interaction diagram, and general deformed
0.20
behavior, the computational study matches well with the available
0.10 experimental data and provides confidence that the current
study using similar modeling techniques is providing reasonable
0.00
0.00 2.00 4.00 6.00 8.00 10.00 behavior.
Lateral Drift Ratio (%)

Fig. 7. Summary of failure criterion for 4.57-m W36 × 487 column

with critical axial load ratio (CALR) shown W14x132-75 W14x176-55
4.5

End Moment (kN-m)

4.5
End Moment (kN-m)

2.3 2.3
0 0
-2.3 -2.3

• Lateral motion in direction of arrow and vertical -4.5 -4.5

-500 -250 0 250 500 -500 -250 0 250 500
motion permitted.
(a) Lateral Displacement (mm) (d) Lateral Displacement (mm)
• Arrow indicates direction of applied lateral load.
• Rotation about y-axis is allowed.
1800
Lateral Force (kN)

Lateral Force (kN)

1800
900 900
0
0
-900
-900
-1800 -1800
-500 -250 0 250 500 -500 -250 0 250 500
(b) Lateral Displacement (mm) (e) Lateral Displacement (mm)

13350 13350
8900
Axial Force (kN)

8900
Axial Force (kN)

4450 4450
0 0
-4450 -4450
-8900 -8900
-13350 -13350
-2.3 -1.1 0 1.1 2.3 -3.4 -2.3 -1.1 0 1.1 2.3 3.4
(c) End Moment (kN-m) (f) End Moment (kN-m)
• No translation is allowed.
• Rotation about y-axis is allowed. Fig. 9. Comparison of (a) moment versus lateral displacement;
springs (b) lateral force versus lateral displacement; (c) P-M interaction re-
sponse between computational model (dashed lines) and experimental
Fig. 8. Finite-element validation model showing mesh and boundary results (solid lines) for W14 × 132-75; similar comparisons are shown
conditions in (d through f) for W14 × 176-55

© ASCE 04015091-7 J. Struct. Eng.

J. Struct. Eng.
 Nearly half of the columns in this study failed to achieve the
plastic rotation of 0.04 rad that is expected of a highly ductile
member under the lowest considered axial load (P=Py ¼ 0.20).
None of the sections in this study, including the compact W14 sec-
tions, are able to carry their maximum expected Ca value and reach
the 4% lateral drift expected of a highly ductile member.
While some of these failures can be attributed to lack of com-
pliance with current AISC guidelines, all of the members that do
satisfy the requirements fail to reach the axial load levels specified
as shown in Table 2 where the actual P=Py value is the CALR re-
corded from the finite-element simulations. A value of P=Py ¼
not measured ðNMÞ merely indicates that the member was unable
to reach 4% drift under the lowest axial load (P=Py ¼ 0.20) inves-
tigated. To facilitate discussion of the behavior of the different sec-
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

tion types hereon, a roman numeral following a section indicates

the type of section as described previously (e.g., W18 × 211 − I
where W18 × 211 is Type I).

Flange Width-to-Thickness Ratio

There is a slight downward trend with a high amount of scatter
between the CALR and flange width-to-thickness ratio for the sec-
tions studied as shown in Fig. 11. In terms of the range of behavior
seen at the extremes as circled in Fig. 11(a), the W18 × 86 − I
(b=t ¼ 7.2, h=tw ¼ 33.3, L=ry ¼ 68.4) experiences significant
local web and flange buckling leading to flexural-torsional
Fig. 10. Comparison of deformed member at 10% drift for buckling while the W30 × 326-I (b=t ¼ 3.75, h=tw ¼ 23.4,
W14 × 132 − 75 (reprinted from Newell and Uang 2006, with L=ry ¼ 66.7) is dominated by flexural-torsional buckling leading
permission) to local flange and web buckling. On the high CALR end, the
W14 × 132-I (b=t ¼ 7.15, h=tw ¼ 17.7, L=ry ¼ 47.9) is domi-
nated by local flange buckling, then local web buckling, and finally
flexural-torsional buckling while the W14 × 500-I (b=t ¼ 2.43,
8 AISC Highly Ductile Limit h=tw ¼ 5.21, L=ry ¼ 40.7) is dominated by flexural buckling
7 W18x86-15ft W14x132-15ft behavior and minor local flange buckling. Columns with lower
6
CALR tend to have higher values of L=ry while the columns with
b/t Ratio

5
4
higher CALR tend to have lower values of L=ry .
3 W30x326-20ft The W40 × 278-I (b=t ¼ 3.31, h=tw ¼ 33.3, L=ry ¼ 71.4)
2 W14x500-15ft initially experiences flange local buckling and subsequently web
1 local buckling, then eventually significant flexural torsional behav-
0 ior. Compared with the W18 × 86-I (b=t ¼ 7.2, h=tw ¼ 33.3,
0 0.2 0.4 0.6 0.8 1
L=ry ¼ 68.4), which is similar in web width to thickness but differ-
(a) CALR (P/Py)
ent in flange width to thickness, the sequence of failure and global
7070 failure of the two columns is similar in terms of the general location
AISCHighly
HighlyDuctile
DuctileLimit
Limit
6060 AISC of the largest out-of-plane displacement as seen in Fig. 12(a).
5050 None of the members with b=t ratios greater than the highly
Ratio
h/twwRatio

W27x146-15ft
4040 ductile limit (Type III) are able to achieve a CALR of 0.2Py or
3030 W30x326-10ft greater. Member W24 × 117-III (largest h=tw for Type III)
h/t

2020 W18x119-20ft experiences local flange buckling followed by flexural-torsional

1010 W14x500-15ft buckling and web local buckling. The W21 × 48 − III (highest
00 b=t) experiences failure due to local flange buckling as well as
0 0.2 0.4 0.6 0.8 1
(b) 0 0.2 0.4 0.6
CALR (P/P y)
0.8 1 flexural-torsional behavior. The W16 × 26-III (lowest h=tw for
CALR (P/P ) Type III) experiences flange and web buckling followed by flexural
Fig. 11. Capacity of columns with trends regarding (a) flange width-to- buckling.
thickness ratio; (b) web width-to-thickness ratio In addition to Type III members failing to maintain this low axial
load up to 4% drift, there were also several members below this
limit that failed to maintain the lowest axial load level (designated
Results of Parametric Study as triangles in Fig. 1). While the conclusion can be drawn that
members above this limit cannot be highly ductile, it does not hold
In total, 70 different members are used in this parametric study. true that members below this limit are necessarily highly ductile.
Thirty different sections are selected and 20 of those sections
are investigated at three different column lengths. Although there
Web Width-to-Thickness Ratio
are exceptions, the general mode of failure for these deeper sections
consists of local web and/or flange buckling leading to global fail- There is a clear negative correlation between the CALR and h=tw as
ure through a combination of torsional and flexural behavior. shown in Fig. 11(b). This corresponds to the AISC seismic design

© ASCE 04015091-8 J. Struct. Eng.

J. Struct. Eng.
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 12. Displaced shape just prior to axial failure criterion for columns under 0.20Py comparing (a) different h=tw (same b=t); (b) different b=t
(similar h=tw )

guidelines approach, which limits h=tw based on the expected Unbraced Length
axial load on the column. In terms of the range of behavior seen
Unbraced lengths of 3.04, 4.57, and 6.10 m for several members are
at the extremes as circled in Fig. 11(b), the W27 × 146-II considered to investigate the effect of unbraced length on capacity.
(b=t ¼ 7.16, h=tw ¼ 39.4, L=ry ¼ 56.1) experiences significant As shown in Fig. 13, the same general trends previously observed
local web and flange buckling while the W18 × 119-I are seen, i.e., there is a severe reduction in ductility with increasing
(b=t ¼ 5.31, h=tw ¼ 24.5, L=ry ¼ 89.4) is dominated by flexural- axial load level. The effect appears to be accentuated as the column
torsional buckling followed by local web and flange buckling. global slenderness increases, which is attributed to a shift in the
The W30 × 326-I (b=t ¼ 3.75, h=tw ¼ 23.4, L=ry ¼ 33.4) experi- failure mode. Local buckling is the dominant failure mode in
ences local flange and web buckling and finally flexural-torsional the shortest column, while lateral torsional buckling becomes more
buckling, while, as mentioned previously, the W14 × 500-I dominant as the length of the column increases.
(b=t ¼ 2.43, h=tw ¼ 5.21, L=ry ¼ 40.7) is dominated by flexural
buckling behavior and minor local flange buckling. Columns with
lower CALR tend to have higher values of L=ry and b=t while the Regression Analysis
columns with higher CALR tend to have lower values of L=ry Using the CALR for each section based on the axial load failure
and b=t. criterion a multivariable regression analysis is undertaken to deter-
The web width-to-thickness ratio appears to influence the mine which factors influence the amount of load a column section
failure mode of the section as shown in Fig. 12(b). The W24 × is able to carry up to 4% lateral drift. The data input into the re-
176-I (b=t ¼ 4.81, h=tw ¼ 28.7) is dominated by flange local gression model are shown in Table 2.
buckling that leads to the formation of plastic hinges near the col- Because it is clear that the width-to-thickness ratios alone do not
umn ends while the W36 × 194-II (b=t ¼ 4.81, h=tw ¼ 42.4) is adequately predict the capacity of sections under this combined
dominated by web local buckling that leads to significant twisting loading scheme, a linear regression model is used to evaluate
and ultimately failure due to flexural-torsional buckling. Despite the potentially important parameters based on the behavior seen
the fact that these sections have the same b=t ratio, the initiation in the parametric study. The first two parameters, h=tw and b=t,
and ultimate failure modes of these two sections are significantly are already accounted for in the current guidelines, but the addi-
different. For the sections studied, h=tw appears to play a more tional terms are included to assess their influence on the CALR.
dominant role in the strength of the columns than the b=t Because columns under this loading scheme seem to exhibit sig-
ratio. nificant twisting behavior after local buckling has begun, several
Nine of the members, as shown by the triangles in Fig. 1, that terms that are involved with global buckling, particularly torsional
satisfy the AISC guidelines for highly ductile members are inca- behavior, are investigated. The third term, L=ry , is used for calcu-
pable of reaching a 4% lateral drift when P=Py ¼ 0.20. The plate lating the nominal compressive strength due to flexural buckling
slenderness ratios of the sections alone are not a clear indication of and is very similar to the ratio of unbraced length to effective
the capacity of these sections under a combined loading scheme as radius of gyration, Lb =rts , which is used to calculate the nominal
seen by the fact that the W27 × 129-II (b=t ¼ 4.55, h=tw ¼ 39.7) flexural strength due
pffiffiffiffiffiffiffi to lateral-torsional buckling. Variations of the
and W21 × 93-I (b=t ¼ 4.53, h=tw ¼ 32.3) cannot carry P=Py ¼ fourth term, JI y =L2 , can be found in solutions for the critical
0.20 up to 4% lateral drift, while a W44 × 335 − II (b=t ¼ 4.50, load at which a simply supported beam will buckle laterally
h=tw ¼ 38.0), which is very similar in terms of both ratios, is able (Gambhir 2004), where J is the torsional constant, I y is the moment
to carry P=Py ¼ 0.30 up to 4% lateral drift. This suggests that other of inertia about the y-axis, and L is the length. The fifth term,
parameters are influential. JSx =h0 , is also used to calculate the nominal flexural strength

© ASCE 04015091-9 J. Struct. Eng.

J. Struct. Eng.
 1
4% Limit
0.9
0.8
0.7 4.57 m

Axial Load (P/Py)

0.6
0.5 3.05 m

0.4
0.3
0.2
6.10 m
0.1
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

0
0.00 1.00 2.00 3.00 4.00 5.00 6.00
Lateral Drift Ratio (%)
(a) (b) (c) (d)
Fig. 13. Axial load of W24 × 176 column versus (a) lateral drift ratio at moment failure criterion as well as column displaced shape (2× magnified) at
moment failure for W24 × 176 with P=Py ¼ 0.40 for unbraced lengths; (b) 6.10 m; (c) 4.57 m; (d) 3.04 m

due to lateral-torsional buckling, where Sx is the elastic section are not highly ductile. However, members that meet the highly duc-
modulus about the x-axis and ho is the distance between the flange tile limit can be evaluated using Eq. (4) and the design aid described
centroids. The final term is the torsional elastic stress, Fez , and is subsequently, which is based on it.
used in the calculation of nominal compressive strength based on
a torsional buckling limit state. These terms account for different
variations of geometric properties that contribute to a section’s abil- Evaluation of Current Seismic Design Guidelines
ity to resist lateral or torsional behavior, and a combination of
one or more of these parameters along with the two parameters that The parametric study described previously suggests that the current
are related to local buckling behavior are expected to enable the AISC guidelines are potentially unconservative and can be im-
prediction of the CALR. proved. The simulations indicate that CALR can be reasonably pre-
To predict the response variable (RV), the general linear model dicted by h=tw and L=ry provided b=t meets current design
used is guidelines for highly ductile members. Instead of having separate
limits as shown in Table 1, limits on the two ratios are interrelated
RV ¼ b1 þ b2 logðX 1 Þ þ b3 logðX 2 Þþ · · · þbn logðX n−1 Þ ð2Þ and can be plotted as a family of curves, each based on a given

where bi = regression coefficients; and X i = predictor variables.

Using the predictor variables in conjunction with the response var-
iable, P=Py , the regression model can be written as
     
P h b L
¼ b1 þ b2 log þ b3 log þ b4 log
Py tw t ry
pffiffiffiffiffiffiffi  
JI y JSx
þ b5 log þ b6 log þ b7 logðFez Þ ð3Þ
G ho

Using backward elimination (Chaterjee et al. 2000) with stan-

dard t-tests until all the parameters are statistically significant at the
5% level results in h=tw and L=ry being the main parameters influ-
encing the critical axial load. Eq. (3) is reduced to the proposed
regression model
   
P h L
¼ 2.6224 − 0.2037 log − 0.3734 log ð4Þ
Py tw ry

The coefficient of multiple determination for the linear regres-

sion of the CALR on h=tw and L=ry is estimated to be 0.722, which
means that these two geometric parameters jointly explain 72.2% of
Fig. 14. Solid curves for logarithmic expectation function for CALR
the variation in the critical axial load ratio. While b=t has been
plotted along with members used in the regression analysis (filled
deemed critical in predicting the behavior in the past, it is unsur-
circles) as well as members that could not sustain 0.2Py up to 4% lateral
prising that it is statistically insignificant based on the high amount
drift (unfilled circles); test members (squares) are also shown with
of scatter seen in Fig. 11. As mentioned previously, the current
arrows indicating the predicted CALR based on the regression equation
flange width-to-thickness ratio can be used to identify sections that

© ASCE 04015091-10 J. Struct. Eng.

J. Struct. Eng.
 Table 5. Evaluation of Plotted Regression Equation Using Test Sections significantly at lateral drift levels below 4% despite the fact that
Expected many of the columns studied can be classified as highly ductile
CALR Actual CALR members based on current AISC seismic design requirements.
Section h=tw L=ry (Fig. 14) (finite element) This is concerning because the overturning effects can easily mag-
Type I
nify a gravity load of 0.15Py . Of the 57 members studied that can
W18 × 211 15.1 63.8 0.5 0.5 be considered highly ductile, only 32 of them were able to support
W24 × 162 30.6 59.0 0.4 0.5 a CALR of 0.20 or higher. This value drops to 24 members when
W36 × 652 16.3 43.9 0.6 0.5 the CALR is 0.40, indicating that approximately half of the
W36 × 652 16.3 52.7 0.5 0.5 members that can be considered highly ductile based on current
W40 × 324 34.2 50.3 0.4 0.3 specifications cannot reach 4% lateral drift under axial loads
Type II (0.20 − 0.40Py ) that are feasible during a seismic event. The stud-
W30 × 148 41.6 79.0 NHDa NMb ied columns indicate that h=tw has a more significant effect on the
a
NHD member based on proposed design guide. strength of the column under the combined loading scheme than
b
CALR is NM because section did not reach 4% drift under 0.20Py. b=t when the current AISC flange width-to-thickness ratio highly
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

ductile limit is satisfied. L=ry also plays a significant role in pre-

dicting the CALR and can be seen by changes in the failure mode
CALR. Fig. 14 shows plots for CALRs ranging from 30 to 80% of different-length members. The dominant buckling mode shifts
along with the scatter of sections that achieved the associated from local buckling to lateral torsional buckling with increasing
CALR based on the finite-element analysis. The range for h=tw unbraced length. An evaluation of the current AISC seismic design
is based on the maximum value of 57.5 (W30 × 90) for sections guidelines suggests that the provisions are potentially unconserva-
deeper than W16s and the range for L=ry is based on columns rang- tive. It is suggested that improvements can be made to the guide-
ing from 3.05 to 6.10 m in length with radius of gyrations ranging lines based on the results of this study and a design aid is provided.
from 28.4 mm (W16 × 26) to 104 mm (W36 × 652). Additional investigations concerning boundary conditions and ad-
Sections that are unable to carry the lowest target axial load in- ditional cyclic loading protocols are necessary to ensure the design
vestigated up to a drift of 4% are also plotted in Fig. 14. While these aid is adequate for predicting the behavior of all column members
sections are not useful in performing the regression analysis, they used in seismic force resisting systems.
are useful in evaluating the limit on h=tw as a function of L=ry for
sections that can serve as highly ductile members.
Acknowledgments

This work was supported by the University of Michigan and

Design Aid National Science Foundation (NSF) grants numbers CMMI-
Fig. 14 can also serve as a powerful design aid. A designer can 0928193 and CMMI-1344372. Any opinions, findings, conclu-
enter the figure with their known h=tw and L=ry and determine sions, and recommendations expressed in this paper are those of
up to what critical axial load ratio their column can be considered the authors and do not necessarily reflect the views of the sponsor.
a highly ductile member. Based on the location of the section re-
garding the two geometric parameters, the closest line to the right
should provide a reasonable approximation for the section’s CALR. References
To evaluate the usefulness of this guideline, six sections that are
AISC. (2010). Seismic design manual, 2nd Ed., Chicago.
not used in the calibration of the model are analyzed using finite- AISC. (2011). Steel construction manual, 14th Ed., Chicago.
element models and compared with the expected values from the ANSI (American National Standards Institute)/AISC. (2010). “Seismic
regression model. These findings are summarized in Table 5 and provisions for structural steel buildings.” ANSI/AISC 341-10, Chicago.
shown in Fig. 14, where the selected sections are represented by Chatterjee, S., Hadi, A., and Price, B. (2000). Regression analysis by
solid squares with arrows pointing to the associated curve. In four example, 3rd Ed., Wiley, New York.
of the six test sections, the equation provided the appropriate criti- Cheng, X., Chen, Y., and Pan, L. (2013). “Experimental study on steel
cal axial load ratio or a more conservative value compared with the beam-columns composed of slender H-sections under cyclic bending.”
finite-element analysis. In two of these sections, the CALR was J. Constr. Steel Res., 88, 279–288.
overestimated by 0.1Py. This is unsurprising because the regression Chi, B., and Uang, C. (2002). “Cyclic response and design recommenda-
tions of reduced beam section moment connections with deep col-
equation only accounts for approximately 70% of the variation in
umns.” J. Struct. Eng., 10.1061/(ASCE)0733-9445(2002)128:4(464),
the CALR and there is clearly scatter in the data, which could 464–473.
possibly be considered within a probabilistic design framework. Dawe, J. L., and Kulak, G. L. (1986). “Local buckling behavior of beam-
Regardless, the proposed process does provide a more reasonable columns.” J. Struct. Eng., 112(11), 2447–2461.
estimate for the actual CALR and behavior of deeper sections than Elkady, A., and Lignos, D. (2015). “Analytical investigation of the cyclic
the current seismic member design guidelines as seen by the much behavior and plastic hinge formation in deep wide-flange steel beam-
smaller difference between the expected and actual critical axial columns.” Bull. Earthquake Eng., 13, 1097–1118.
loads when comparing Table 5 with Table 2. Engelmann, B. E., Whirley, R. G., and Goudreau, G. L. (1989). “A simple
shell element formulation for large-scale elastoplastic analysis.” Ana-
lytical and computational models of shells, A. K. Noor, T. Belytschko,
and J. C. Simo, eds., ASME, New York.
Conclusions Fadden, M., and McCormick, J. (2014). “Finite element model of the cyclic
bending behavior of hollow structural sections.” J. Constr. Steel Res.,
The behavior of steel wide-flanged beam-columns is investigated 94, 64–75.
under a combined axial and cyclic lateral loading scheme. Fogarty, J., and El-Tawil, S. (2013). “Collapse behavior of steel columns
Results from the parametric study indicate that local buckling under lateral loading.” Proc., ASCE Structures Congress, ASCE,
and lateral torsional buckling are reducing the column strength Reston, VA.

© ASCE 04015091-11 J. Struct. Eng.

J. Struct. Eng.
 Fogarty, J., Yossef, N., and El-Tawil, S. (2013). “Collapse resistance of Moehle, J. P., Elwood, K. J., and Sezen, H. (2002). “Gravity load collapse
locally damaged steel columns.” J. Constr. Steel Res., 82, 195–202. of building frames during earthquakes.” S. M. Uzumeri Symp.: Behavior
Gambhir, M. L. (2004). Stability analysis and design of structures, and Design of Concrete Structures for Seismic Performance, American
Springer, Berlin. Concrete Institute, 215–238.
Hypermesh Version 11.0 [Computer software]. Troy, MI, Altair Engineering. Newell, J., and Uang, C. (2006). “Cyclic behavior of steel columns with
Kaufman, E., Metrovich, B., and Pense, A. (2001). “Characterization of combined high axial load and drift demand.” Rep. No. SSRP-06/22,
cyclic inelastic strain behavior on properties of A572 Gr. 50 and Dept. of Structural Engineering, Univ. of California, San Diego,
A913 rolled sections.” ATLSS Rep. No. 01-13, National Center for La Jolla, CA.
Engineering Research on Advanced Technology for Large Structural Ricles, J., Mao, C., Lu, L.-W., and Fisher, J. (2003). “Ductile details for
Systems, Lehigh Univ., Bethlehem, PA. welded unreinforced moment connections subject to inelastic cyclic
Kemp, A. R. (1986). “Factors affecting the rotation capacity of plastically loading.” J. Eng. Struct., 25(5), 667–680.
designed members.” Struct. Eng., 64B(2), 28–35. Setzler, E., and Sezen, H. (2008). “Model for the lateral behavior of rein-
Khandelwal, K., El-Tawil, S., Kunnath, S., and Lew, H. (2008). “Macro- forced concrete columns including shear deformations.” Earthquake
Spectra, 24(2), 493–511.
model-based simulation of progressive collapse: Steel frame struc-
Shen, J.-H. J., Astaneh-Asl, A., and McCallen, D. B. (2002). “Use of deep
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 09/25/15. Copyright ASCE. For personal use only; all rights reserved.

tures.” J. Struct. Eng., 10.1061/(ASCE)0733-9445(2008)134:7(1070),

columns in special steel moment frames.” Steel Tip Rep. No. 24, Struc-
1070–1078.
tural Steel Educational Council, Moraga, CA.
Krawinkler, H., Gupta, A., Medina, R., and Luco, N. (2000). “Development
Suita, K., Yamada, S., Tada, M., Kasai, K., Matsuoka, Y., and Sato, E.
of loading histories for testing of steel beam-to-column assemblies.” Dept.
(2007). “E-defense tests on full-scale steel buildings. Part 2: Collapse
of Civil and Environmental Engineering, Stanford Univ., Stanford, CA.
experiments on moment frames.” Proc., ASCE Structures Congress,
Krishnan, S., and Muto, M. (2012). “Mechanism of collapse of tall steel ASCE, Reston, VA.
moment frame buildings under earthquake excitation.” J. Struct. Eng., Suita, K., Yamada, S., Tada, M., Kasai, K., Matsuoka, Y., and Sato, E.
10.1061/(ASCE)ST.1943-541X.0000573, 1361–1387. (2008). “Results of recent E-defense tests on full-scale steel buildings.
Lamarche, C., and Tremblay, R. (2011). “Seismically induced cyclic buck- Part 1: Collapse experiment on 4-story moment frame.” Proc., ASCE
ling of steel columns including residual—Stress and strain-rate effects.” Structures Congress, ASCE, Reston, VA.
J. Constr. Steel Res., 67(9), 1401–1410. Wu, C.-L., Yang, Y.-S., Hwang, S.-J., and Loh, C.-H. (2010). “Dynamic
Lay, M. G. (1965). “Flange local buckling in wide-flange shapes.” J. Struct. collapse of reinforced concrete columns.” Proc., 9th U.S. National
Div., 91(6), 94–116. and 10th Canadian Conf. on Earthquake Engineering, EERI, Oakland,
LS-DYNA [Computer software]. Livermore, CA, Livermore Software Tech- CA.
nology Corp. Zargar, S., Medina, R. A., and Miranda, E. (2014). “Cyclic behavior of
MacRae, G. A., Carr, A. J., and Walpole, W. R. (1990). “The seismic re- deep steel columns subjected to large drifts, rotations, and axial loads.”
sponse of steel frames.” Rep. No. 90-6, Dept. of Civil Engineering, Proc., 10th U.S. National Conf. on Earthquake Engineering, EERI,
Univ. of Canterbury, Christchurch, U.K. Oakland, CA.

© ASCE 04015091-12 J. Struct. Eng.

J. Struct. Eng.