Engineering Structures 31 (2009) 2812–2824

Contents lists available at ScienceDirect

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

Elastic buckling of cold-formed steel columns and beams with holes

Cristopher D. Moen a,∗ , B.W. Schafer b
a
Department of Civil & Environmental Engineering, Virginia Tech, 102-D Patton Hall, Blacksburg, VA 24061, USA
b
Department of Civil Engineering, Johns Hopkins University, 203 Latrobe Hall, Baltimore, MD 21218, USA

article info abstract

Article history: Simplified methods for approximating the global, distortional, and local critical elastic buckling loads
Received 27 May 2009 of cold-formed steel columns and beams with holes are developed and summarized. These methods
Received in revised form provide engineering approximations appropriate for design, but are intended to be general enough to
7 July 2009
accommodate the range of hole shapes, locations, and spacings common in industry. The simplified
Accepted 7 July 2009
Available online 19 July 2009
methods are developed as a convenient alternative to shell finite element eigen-buckling analysis,
which require laborious and subjective visual identification methods as well as commercial software
Keywords:
not generally accessible to the engineering community. Global buckling of cold-formed steel beams and
Perforations columns, including the influence of holes, is predicted with approximate ‘‘weighted average’’ cross-
Holes section properties formulated from classical energy-based stability solutions. Distortional and local
Prediction method buckling of a cold-formed steel member with holes are determined with the semi-analytical finite
Local buckling strip method, considering appropriate modifications to the element thickness and choice of buckling
Distortional buckling half-wavelength. The proposed methods are verified with shell finite element eigen-buckling studies.
Global buckling Unambiguous, simple methods for elastic buckling prediction of members with holes is central to
Direct Strength Method the extension of the Direct Strength Method (DSM) for cold-formed steel member ultimate strength
Cold-formed steel
determination.
© 2009 Elsevier Ltd. All rights reserved.

1. Introduction However, discontinuities caused by holes cannot be explicitly

modeled with the finite strip method, and therefore, formal
Holes can be found in most cold-formed steel structural determination of the elastic buckling loads requires shell finite
components. For example, in low and midrise construction, evenly- element (FE) models with the holes explicitly included. An example
spaced holes are placed in the webs of cold-formed steel columns of the results for such models is provided in Fig. 1 (see the ‘‘•’’ in
and beams, allowing electrical, plumbing, and heating services Fig. 1 which are determined from FE and may fall on or below the
to pass through walls and ceilings. Cold-formed steel members ‘‘ ’’ FSM results). Compared to the finite strip method, FE models
without holes can be designed with the American Iron and Steel are computationally inefficient, but worse than this the results are
Institute’s Direct Strength Method (DSM) [1, Appendix 1], [2], subjective and laborious to obtain. Each of as many as 100’s of FE
which utilizes the local, distortional, and global (Euler) elastic buckling modes must be visually inspected to determine whether
buckling properties of a cross-section to predict ultimate strength. the buckling mode is local, distortional, or global. For example, in
Fig. 1 local is FE mode 1, distortional is FE mode 28 and global is FE
(The use of thin sheet steel makes cold-formed steel members
mode 117; these modes were identified through mode-by-mode
susceptible to local buckling. Distortional buckling, a form of
visual inspection and judgment is required in the separation.
buckling related to intermediate and/or edge stiffeners, is also
This manuscript describes a procedure for approximating the
commonly observed in open cross-sections, see Fig. 1.) The elastic
change in the elastic buckling load (or moment) due to the pres-
buckling properties for members without holes can be determined
ence of holes. The procedure enlists simplified methods developed
from an elastic buckling curve generated with the semi-analytical as a convenient alternative to full finite element eigen-buckling
finite strip method (Fig. 1). The finite strip method (FSM) reveals analysis. The end result is a verified series of engineering approx-
and readily identifies, with remarkable computational efficiency, imations that predict the elastic buckling of cold-formed steel
the buckling modes which must be considered in design: the two beams and columns including the influence of holes without ambi-
minima, local and distortional, along with the final global buckling guity and with sufficient robustness for engineering design.
mode occurring at the physical length of the member.
2. Global buckling of cold-formed steel columns and beams
with holes
∗ Corresponding author. The nominal global axial capacity, Pn , of a cold-formed steel
E-mail address: [email protected] (C.D. Moen). column is dictated by the relationship between slenderness and
0141-0296/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2009.07.007
 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824 2813

Fig. 1. Elastic buckling curve for a cold-formed steel lipped C-section with the finite strip method, and the influence of holes with shell finite element eigen-buckling
analysis.

Fig. 2. Cold-formed steel column design curve.

strength as depicted in Fig. 2 [1]. As the global elastic buckling load,

Fig. 3. Rayleigh–Ritz column coordinate system and notation.
Pcre , decreases relative to the squash load, Py , the global slenderness
of the column, λc = (Py /Pcre )0.5 , increases — and the more slender global flexural (Euler) buckling load, Pcre , of a uniaxially loaded
column has less capacity. A similar design approach is employed simply-supported column with holes. This approximate method is
for laterally unbraced cold-formed steel beams except the nominal motivated by existing analytical solutions for the elastic buckling
flexural strength, Mn , is predicted with the global slenderness of of a column with a variable cross-section [4–6]. For the solution,
the beam, λb = (My /Mcre )0.5 , and My is the yield moment for the holes are placed along the length of the column, forming regions
beam. of alternating gross and net cross-sections. The coordinate system
Quantifying the reduction in Pcre or Mcre caused by the presence and dimensional notation for the column with holes is defined in
of holes can be achieved with FE eigen-buckling analysis, although Fig. 3.
this option is subjective, laborious, and not typically accessible The simply-supported column is assumed to deform in one half-
to practicing engineers. Nonetheless, FE analysis has been used sine wave during elastic flexural buckling:
by researchers to demonstrate that the global buckling load
πx
of cold-formed steel rack columns decrease with the presence v(x) = α sin , (1)
of closely-spaced hole patterns provided for shelving height L
adjustment [3]. As an alternative, the following conservative, en- where α is an arbitrary constant defining the magnitude of the out-
gineering approximations for calculating Pcre and Mcre of structural of-plane column displacement. Note, the use of Eq. (1) eliminates
members with holes are proposed. the potential for the buckling mode to localize near the hole and
focuses the analysis on global buckling alone, alternative methods
2.1. Global flexural buckling are discussed later in this manuscript for evaluating localized
buckling modes.
2.1.1. Rayleigh–Ritz energy solution Equilibrium in the deformed (bent) state is governed by [7,
A Rayleigh–Ritz energy solution is employed in this section Appendix]:
to develop an approximate equation for the critical elastic δ(U + W ) = 0, (2)
2814 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824

where Eq. (2) represents the variation in total potential energy, U

is the strain energy in the column due to bending, and W is the
external work performed by the axial load P on the column, as the
column shortens axially due to bending. The bending strain energy
in the column is:
2
L
d2 v(x)
Z 
1
U = EI (x) dx, (3)
2 0 dx2
where E is the modulus of elasticity. If the moment of inertia, I (x),
of the column varies discretely as shown in Fig. 3 between the gross
cross-section, Ig , and the net cross-section, Inet , then substituting
Eq. (1) into Eq. (3):
"
E α2 π 4 πx
2n
X Z x̂i+1
U = Ig sin2 dx
2L4 i=0 x̂i L Fig. 4. ABAQUS column boundary and loading conditions.
#
2n
X Z x̂i+1 πx Pcre depending upon the location of the holes in the column. For
+ Inet sin2 dx , (4)
L example, when considering a single hole, the minimum Pcre occurs
i=1 x̂i
when the hole is located at the longitudinal midlength of the
where column (i.e., when c = L/2). This result is consistent with physical
x̂0 = 0 intuition.
If holes are spaced symmetrically about the longitudinal
x̂1 = c1 − 0.5Lhole,1 midline of the column (x = L/2), then T = 0 and Eq. (9) reduces to
x̂2 = c1 + 0.5Lhole,1
π 2E
 
Ig Lg + Inet Lnet
.. (5)
Pcre = . (12)
. L2 L
x̂2n−1 = cn − 0.5Lhole,n Pcre is conveniently proportional to the weighted average of Ig and
Inet . The validity of this ‘‘weighted average’’ approximate method
x̂2n = cn + 0.5Lhole,n
for Pcre is explored next for flexural buckling of a column with
x̂2n+1 = L. evenly spaced holes. Obviously, the most conservative approxima-
The distance from the end of the member to the centerline of tion to Eq. (12) would be to replace Ig with Inet , resulting in a ‘‘net
hole j is cj , and Lhole,j is the length of hole j for j = 1 . . . n holes. section’’ approximation.
The external work of the applied load P as the column is axially
compressed is approximated as: 2.1.2. Verification for flexural buckling of a column with holes
Thin shell finite element eigen-buckling analysis in ABAQUS [8]
dv(x)
Z L
 2
1
W =− P dx. (6) is employed to evaluate the ‘‘weighted average’’ approximation of
2 0 dx Eq. (12) for the global flexural critical elastic buckling load of a
Substituting Eq. (1) into Eq. (6): column with evenly spaced holes. This example considers weak-
axis flexural buckling of an industry standard cold-formed steel
P α2 π 2 L πx P α2π 2
Z
Structural Stud Manufacturers Association (SSMA) 1200S162-68
W =− cos2 dx = − . (7)
2L2 0 L 4L long column [9] with circular web holes. The length of the column,
L, is 2500 mm, the hole spacing, S , is 500 mm, and the diameter of
For this single parameter problem solution, Eq. (2) takes the form
the circular hole is varied from hhole /H = 0.10 to hhole /H = 0.90
dU dW where hhole is the hole depth and H is the out-to-out depth of the
+ = 0. (8)
dα dα cross-section. The ABAQUS finite element models are loaded in
Eq. (8) can be solved directly for the column critical elastic buckling compression at the member ends and have warping free boundary
load, Pcre , including the influence of holes, after substituting Eqs. (4) conditions, as shown in Fig. 4. The modulus of elasticity, E, is
and (7). assumed as 203.4 GPa and Poisson’s ratio, ν , as 0.3. The weak-axis
The critical elastic buckling load, Pcre , for n arbitrarily spaced flexural buckling load of the column without holes is Pcre,nohole =
net-section regions is: 35.2 kN as predicted by the classical Euler buckling equation.
The ‘‘weighted average’’ approximate method of Eq. (12) is
π 2E Ig Lg + Inet Lnet + T (Ig − Inet )
 
shown to be an accurate predictor of the weak-axis flexural
Pcre = , (9)
L2 L buckling load in Table 1 and Fig. 5 over the range of hhole /H
considered in this study. (Note that the ABAQUS-to-predicted
where
mean and standard deviation are designated as µ and σ in this
2π cj π Lhole,j
n    
L X manuscript.) The ‘‘weighted average’’ approximation for I is more
T = cos sin . (10)
2π j=1 L L accurate than using the net-section moment of inertia (Inet ) in the
classical Euler prediction for Pcre (Table 1 and Fig. 5), which results
The total length of the net-section regions (i.e., regions containing in increasingly conservative predictions as hhole /H increases.
holes) in the column is
n
X 2.2. Global flexural–torsional buckling
Lnet = Lhole,j , (11)
j=1 2.2.1. Torsional properties for members with holes
and the length of the column without holes is Lg = L − Lnet . The The critical elastic buckling loads for flexural and flexu-
trigonometric term T in Eq. (10) is interesting in that it decreases ral–torsional buckling modes of a column are described by the
 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824 2815

Table 1
Influence of hole diameter on global buckling.
Buckling mode Comparison hhole /H ABAQUS-to-predicted statistics
0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 µ σ
Column weak-axis Pcre,ABAQUS /Pcre,no hole a
0.93 0.92 0.91 0.89 0.88 0.84 0.79 0.72 0.61 – –
flexural buckling
Pcre,ABAQUS /Pcre,weighted av erage a 0.93 0.93 0.92 0.92 0.92 0.90 0.90 0.88 0.84 0.90b 0.03
Pcre,ABAQUS /Pcre,net section a 0.95 0.96 0.98 1.00 1.03 1.05 1.09 1.15 1.25 1.05b 0.10
Column flexural–torsional Pcre,ABAQUS /Pcre,no hole 0.99 0.96 0.92 0.86 0.79 0.71 0.64 0.55 0.44 – –
buckling
Pcre,ABAQUS /Pcre,weighted av erage 1.01 1.02 1.02 1.02 1.02 1.01 1.03 1.05 1.03 1.02 0.01
Pcre,ABAQUS /Pcre,net section 1.08 1.15 1.20 1.25 1.28 1.29 1.34 1.37 1.45 1.27 0.12
Beam lateral–torsional Mcre,ABAQUS /Mcre,no hole 0.99 0.98 0.97 0.96 0.94 0.91 0.86 0.79 0.68 – –
buckling
Mcre,ABAQUS /Mcre,weighted av erage 1.00 1.00 1.01 1.02 1.03 1.05 1.06 1.08 1.10 1.04 0.04
Mcre,ABAQUS /Mcre,net section 1.00 1.01 1.02 1.04 1.05 1.07 1.10 1.15 1.25 1.08 0.08
a
ABAQUS results are systematically 7% lower than the Euler buckling solution. The difference is caused by the assumption of a rigid cross-section in the classical stability
equations, see [18] for details.
b
Compare µ (mean) to a baseline of 0.93, not 1.0, because of 7% systematic difference caused by the rigid cross-section assumption.

The x- and y-distances from the shear center to the centroid of the
cross-section are xo and yo respectively, and ro is the polar moment
of inertia about the shear center, defined as ro = (x2o + y2o )0.5 .
Eq. (13) utilizes the moment of inertia of a column, Ix and Iy , the St.
Venant torsional constant, J, and the warping torsion constant, Cw ,
all of which are reduced by the presence of holes. For a column with
holes in flexural buckling, Ix and Iy were shown to be accurately
predicted by the ‘‘weighted average’’ formulation in Eq. (9). Finite
element analysis is employed in this section to determine if the
torsional properties, J and Cw , of a column, including the influence
of holes, can also be predicted with the same ‘‘weighted average’’
approach, allowing for the extension of Eq. (13) to columns with
holes.
Focusing directly on torsion, the differential equation for non-
uniform torsion is defined as [6]:

dβ d3 β
T = GJ − ECw , (15)
Fig. 5. Comparison of ‘‘weighted average’’ and ‘‘net section’’ approximations to dx dx3
ABAQUS global buckling load, Pcre (weak-axis flexure).
where β is the angle of twist of the cross-section and G is the shear
modulus (G = 78.2 GPa is assumed for cold-formed steel). Eq. (15)
three roots of the classical cubic column buckling equation [6,10]:
is used in conjunction with a static ABAQUS analysis to solve for
(Pcrey − Pcre )(Pcrex − Pcre )(Pcreφ − Pcre ) the average J and Cw of a column as hhole /H varies from 0.10 to
2 2 2 2 0.90. Specifically, the average J is calculated by applying a unit twist
Pcre x0 Pcre y0
− (Pcrey − Pcre ) − (Pcrex − Pcre ) = 0, (13) at the end of the column, βo , about the gross cross-section shear
r02 r02 center while keeping the opposite column end fixed against twist
where as shown in Fig. 6. If both ends of the column are free to warp (again
π 2 EIx π 2 EIy see Fig. 6), the variation in twist along the column is constant as
Pcrex = , Pcrey = , shown in Fig. 7 and warping resistance does not contribute to the
L2 L2
resulting torsion (i.e. d3 β/dx3 = 0). The variation in twist is not
π 2 ECw
 
1 sensitive to increasing hole diameter in this case, and therefore the
Pcreφ = 2 GJ + . (14)
r0 2 L solid line of Fig. 7 is the same regardless of the hhole /H considered.

Fig. 6. ABAQUS boundary conditions for the study of J and Cw including holes.
2816 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824

The ABAQUS boundary conditions are now modified such that

warping is restrained at the fixed column end (Fig. 6). A unit twist,
βo , is again applied at the free end, and the resulting angle of twist,
β , along the length of the column is measured in ABAQUS. Because
of the warping-fixed end condition, β is nonlinear along the length
(i.e. d3 β/dx3 6= 0) of the column (Fig. 7) and warping resistance
contributes to the torsional stiffness. Since the distribution of β
along the column is not influenced by hhole /H, an indirect solution
of Cw,FE as a function of Cw,g can be derived:
dβ
Cw,FE To − GJFE (x = 0)
= dx
dβ
, (18)
Cw,g To,g − GJg dx (x = 0)
where for each ABAQUS model (hhole /H = 0.10 to 0.90), the
torque To associated with the unit twist βo is reported directly from
ABAQUS, and dβ/dx (x = 0) is calculated from Fig. 7 (dashed line).
Fig. 8b shows that the ‘‘weighted average’’ approximation for
Cw overestimates the average Cw,FE predicted with ABAQUS. The
Fig. 7. Angle of twist decreases linearly in the SSMA 1200S162-68 column with reason that Cw does not follow the ‘‘weighted average’’ prediction
warping free end conditions, and is nonlinear for a warping fixed end condition at can be explained by exploring the mechanics of warping torsion.
x = L.
Restrained warping creates longitudinal stresses (σx ) and related
shear stresses (τxy ) around a cross-section, as shown in Fig. 9. The
This is not to say that the hole has no effect, indeed the stiffness magnitude of the torsional moment created by τxy about the shear
decreases, but the angle of twist remains linear. (The twist β along center of the cross-section is:
the column is measured in ABAQUS as the relative rotation of the d3 β(x)
Z m
flange-web corners). ECw = τxy (s, x)tr (s)ds. (19)
The average St. Venant torsional constant (derived from the dx3 0
finite element model) for the warping free column, JFE , is calculated Therefore, the influence of a hole on Cw can be evaluated by
by rearranging Eq. (15): observing changes in τxy at the net cross-section of a member.
The warping shear stress (τxy ) is plotted in Fig. 10 at a section
To L
JFE = . (16) through a hole (Section B–B, Fig. 6) for the SSMA 1200S162-68
G βo column. Shear stress cannot be carried at the hole, and there is a
The shear modulus, G, and column length, L, are known and To ‘‘disturbed’’ region adjacent to the hole where τxy is not carried
is determined directly from ABAQUS as the end torque resulting effectively by the cross-section. The disturbed region extends along
from the unit rotation βo . The resulting JFE from ABAQUS is the full length of the column, because discontinuities created by
compared against Jav g calculated with the ‘‘weighted average’’ the discrete holes do not allow the warping resistance at the end of
approximation, i.e. a column to develop along the length. Thus, the ‘‘weighted average’’
approximation is not valid when calculating Cw for structural
Jg Lg + Jnet Lnet
Jav g = . (17) members with holes.
L The dimension of the hole and the disturbed region together,
(Note that Jnet can readily be calculated with the built-in section hhole∗ , is related to the size of the hole relative to the web depth,
property calculations in the freely available open source program hhole /H, and can be described for the web holes in this study with
CUFSM [11] by assuming the thickness is zero at the hole.) Jav g the relationship:
calculated with Eq. (17) is consistent with JFE and is a viable  0.2
predictor of the St. Venant torsional constant for this member 1 hhole
(Fig. 8a). hhole∗ = hhole + (H − hhole ) . (20)
2 H

a b

Fig. 8. (a) The ‘‘weighted properties’’ approximation for Jav g matches closely with the ABAQUS prediction for the SSMA 12S00162-68 column with holes but (b) overpredicts
warping torsion resistance when compared to ABAQUS results.
 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824 2817

Fig. 9. Warping stress coordinate system.

Fig. 11. Comparison of ‘‘weighted average’’ and ‘‘net section’’ approximations to

ABAQUS global buckling load, Pcre (flexural–torsional buckling).

section properties to calculate Pcre produces overly conservative

predictions as hole size increases relative to web depth.

2.2.3. Calculating flexural–torsional buckling of a column with holes

The global buckling load, Pcre , for a column with holes is
calculated by first obtaining the ‘‘weighted average’’ of Ix , Iy , J, ro ,
xo , and yo with the pertinent form of Eq. (9). Note, the net- and
gross-section properties may be readily obtained from the built-
in section property calculations of CUFSM [11] or CUTWP [12].
(The net-section properties are calculated by setting the thickness
t = 0 at the location of the holes in the cross-section). The warping
torsion constant, Cw , is then obtained by setting the cross-section
thickness to zero in the disturbed region of a hole, i.e. hhole∗ . Once
the section properties have been calculated, they are substituted
into Eq. (13). The three roots of Eq. (13) are then solved, the lowest
root being Pcre . Custom Matlab software [13] is available to perform
Fig. 10. Influence of a hole on warping shear stress, τxy , hhole /H = 0.30. the calculations, or alternatively, a spreadsheet program such as
Microsoft Excel [14] can be used.

Eq. (20) is an empirical relationship derived to produce Cw,av g

2.3. Lateral–torsional buckling
consistent with Cw,FE over the range of hole sizes considered
(Fig. 8b). (Cw,av g may be determined with the built-in section
2.3.1. Extension of column flexural–torsional buckling to beams
property calculation of CUFSM [11] by setting the cross-section
sheet thickness equal to zero, i.e. t = 0, over hhole∗ .) The ‘‘weighted average’’ method developed for columns with
holes is extended to beams with holes in this section. The classical
lateral–torsional stability equation for a simply-supported beam
2.2.2. Verification for flexural–torsional buckling of a column with loaded with a constant moment along its length is [6]:
holes s
The second elastic global buckling mode of an SSMA 1200S162- π π2
 
68 column is flexural–torsional, with Pcre,nohole = 59.6 kN for Mcre = EIy GJ + ECw . (21)
L L2
the column without holes. Table 1 demonstrates that Pcre decreases
with increasing hhole /H, and that the flexural–torsional mode The approximate method is implemented by replacing Iy and J with
is more sensitive to hole diameter than the weak-axis flexural Iyav g and Jav g and by calculating Cwav g assuming the cross-section
mode. When hhole /H = 0.50, the reduction in Pcre for the thickness is zero over the disturbed region hhole∗ .
weak-axis flexure mode is 5% (discounting the 7% systematic
difference caused by the rigid cross-section assumption) but for 2.3.2. Verification for lateral–torsional bucking of a beam with holes
flexural–torsional buckling the reduction caused by the presence The cold-formed steel SSMA 1200S162-68 member evaluated in
of holes is 21%. The reason for the large reduction in Pcre for the the previous sections as a column is now evaluated as a beam with
flexural–torsional mode is the degradation of Cw caused by the a uniform moment. The beam ends are modeled as warping-free
discrete holes along the length of the column (see discussion in and the cross-section at the longitudinal midline is warping-fixed
previous section). as shown in Fig. 4. Mcre decreases with increasing hole diameter as
Utilizing ‘‘weighted average’’ properties for I, J, etc. as well shown in Table 1, which is consistent with the decreasing trends
as the additional reduction in Cw,net embodied in Eq. (20), it in Pcre for columns with holes. The ‘‘weighted average’’ prediction
is shown that the provided engineering approximations are an method is demonstrated to be a viable, conservative predictor
accurate predictor of Pcre as shown in Fig. 11 over the range of Mcre when compared to the ABAQUS eigen-buckling results
of hhole /Hconsidered in this study. Further, using just the net (Fig. 12).
2818 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824

C-section cold-formed steel column. The web of the cross-section

is isolated as a simply-supported plate and loaded with imposed
rotations with magnitudes varying as a half-sine wave to simulate
distortional deformation over one half-wavelength (Fig. 13). The
plate dimensions in ABAQUS are chosen to correspond to the web
of an SSMA 250S162-68 cross-section [9]. The plate width h =
61 mm, the plate length L = 305 mm (consistent with the
distortional half-wavelength for the gross cross-section, Lcrd =
305 mm), and t = 1.81 mm. One slotted hole with Lhole =
101.6 mm is centered in the plate. The width of the hole is varied,
i.e. hhole /h = 0.21, 0.40, 0.50, 0.62, and 0.73 (and subsequently rhole
varies). The modulus of elasticity, E, is assumed as 203.4 GPa and
Poisson’s ratio, ν , as 0.30 in all finite element models.
At each node where an imposed rotation is applied, the
associated moment is obtained from ABAQUS and plotted in Fig. 14
as a transverse bending stiffness per unit length. The presence of
a hole causes a sharp reduction in bending stiffness locally at the
hole, but has a minimal influence on bending stiffness away from
Fig. 12. Comparison of ‘‘weighted average’’ and ‘‘net section’’ approximations to the hole. The stiffness reduction is shown to be insensitive to the
ABAQUS global buckling moment, Mcre (lateral–torsional buckling).
ratio of hole width to plate width except for peaks in stiffness that
increase with hhole /h at the rounded edges of the slotted hole.
3. Distortional buckling of cold-formed steel columns and If Kθ represents the cumulative transverse rotational stiffness
beams with holes for the plate without a hole (the area under the no hole curve in
Fig. 14), then the reduced Kθ including the presence of the hole can
Distortional buckling is recognized as a design limit state for be approximated as:
cold-formed steel columns and beams with open cross-sections,  
separate from that of global or local–global buckling interaction Lhole
Kθ hole = 1− Kθ . (22)
[15,16]. The capacity of a cold-formed steel member undergoing Lcrd
distortional buckling is calculated with a design curve similar
An explicit expression for Kθ hole which relates the dimensions and
to that for global buckling (Fig. 2), but instead the distortional
thickness of the plate to transverse rotational stiffness can be
slenderness, λd = (Py /Pcrd )0.5 for columns or λd = (My /Mcrd )0.5
determined by realizing that the rotational stiffness of the plate is
for beams is employed. The bending stiffness provided by the
linearly related to the plate rigidity, D [10]:
web of an open cross-section provides rotational restraint to the
flange, and therefore, at least in part, dictates the magnitude of E (tweb,hole )3
Pcrd and Mcrd . If a hole is introduced into the web of a lipped C- Kθ hole Dhole 12(1−ν 2 ) (tweb,hole )3
∝ = = . (23)
section, the rotational restraint provided by the web is decreased, Kθ D E (tweb)3 (tweb )3
resulting in a lower critical distortional buckling load. For example, 12(1−ν 2 )

see Fig. 1 or the extensive work in [17] where the decrease Substituting Eq. (22) into Eq. (23) and rearranging in terms of
in Pcrd of cold-formed steel structural studs was observed in a tweb,hole , the web thickness corresponding to the reduced trans-
finite element study evaluating the influence of slotted web holes verse bending stiffness from the hole is approximated as:
provided for improved thermal performance in cold climates [17].  1/3
The magnitude of the rotational stiffness reduction is explored and Lhole
tweb,hole = 1− tweb . (24)
quantified in the following section. Lcrd
An expression similar to Eq. (24) has been proposed for the evalua-
3.1. Distortional buckling of columns with web holes
tion of distortional buckling behavior of web-slotted thermal struc-
tural cold-formed steel studs [17].
3.1.1. Loss of transverse stiffness due to hole To calculate Pcrd including the influence of holes, first a finite
A shell finite element model is developed in ABAQUS to study strip analysis is performed with the gross cross-section to identify
the influence of a web hole on the web bending stiffness of a lipped the distortional half-wavelength, Lcrd . Then, the web thickness is

Fig. 13. ABAQUS boundary conditions and imposed rotations for the web plate undergoing ‘‘distortional-type’’ deformations.
 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824 2819

Table 2
Column and beam dimensions considered in simplified method verification studies.
Specimen Boundary conditions Specimen Extrema Dlip /t H /t B /t H /B Dlip /B L /H hhole /H H /Lhole S /Lhole
type quantity

Columnsa fixed–fixed 63 min 6.3 46.3 19.3 1.9 0.23 1.7 0.14 0.9 2.2a
max 20.0 172.7 65.0 4.9 0.32 13.3 0.57 7.0 24.0a
weak-axis pinned 15 min 6.6 46.2 20.4 2.3 0.33 7.7 0.14 2.3 18.0a
max 10.3 71.6 31.7 2.3 0.30 17.9 0.43 7.0 78.2a
Beams Simply supported, four point 144 min 5.5 40.5 16.3 1.5 0.18 15.9 0.12 0.6 5.2
bending
max 22.1 257.1 58.3 7.7 0.42 62.0 0.62 3.0 12.0
a
Columns have a single hole located at the longitudinal midline, S is set equal to L, Dlip is the out-to-out flange stiffening lip length and B is the out-to-out flange width.

Fig. 14. Transverse rotational stiffness of a plate is reduced in the vicinity of a hole. Fig. 15. The finite strip approximate method for distortional buckling is consistent
with ABAQUS results for the SSMA 250S162-68 column with holes.

modified with Eq. (24) and Pcrd (including the influence of holes)
is obtained with another finite strip analysis performed just at Lcrd
of the gross cross-section. It is important to note that the second
analysis only be conducted at Lcrd as this is the only length for which
the reduced thickness tweb,hole has any relevance. Alternatively, the
modified web thickness in Eq. (24) can be used directly with the
hand method for predicting Pcrd derived in [2] and provided in the
AISI specification [1, Eq. C3.1.4-14].

3.1.2. Verification for distortional bucking of a column with web holes

To demonstrate the finite strip prediction method for distor-
tional buckling, Pcrd is approximated for a long column (L =
2540 mm) with an SSMA 250S162-68 cross-section and five evenly
spaced slotted web holes, where S = 508 mm and Lhole =
102 mm. The width of the hole is varied relative to the web width,
and ABAQUS eigen-buckling results are used to evaluate the viabil-
ity of the method. The squash load of the column, Py , is calculated
with the gross cross-sectional area assuming Fy = 345 MPa. Fig. 16. The finite strip approximate method is a viable predictor of Pcrd when
The ABAQUS distortional buckling mode shape is provided in compared to ABAQUS predictions.
Fig. 15 for hhole /h = 0.63. Nine distortional half-waves form along
the member, with every other half-wave containing one slotted length increases relative to the fundamental distortional half-
hole. The finite strip prediction method is compared over a range of wavelength, Lcrd . The prediction accuracy is highly variable when
hhole /h to ABAQUS eigen-buckling results in Fig. 15, demonstrating L/Lcrd < 1 because of the boundary condition influence for stocky
that the approximation is a viable predictor of Pcrd . columns.
The approximate prediction method for distortional buckling
of columns with holes was also verified across the parameters in 3.2. Distortional buckling of beams with web holes
Table 2, which includes 78 lipped C-section column specimens
from an experimental database. (Full details of the verification 3.2.1. Extension of reduced thickness concept to beams with web holes
study, including the determination of Pcrd with thin shell finite The finite strip distortional buckling prediction method for
element eigen-buckling analysis for each of the specimens cold-formed steel columns with holes presented in the previous
including the influence of holes and test boundary conditions, section is now extended to cold-formed steel beams with holes.
are available in [18].) The results of the verification study are The expression for tweb,hole of Eq. (24), derived for columns, is
summarized in Fig. 16. Prediction accuracy improves as the column proposed as an approximate method for beams with web holes.
2820 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824

Fig. 17. Finite element model boundary conditions for beam eigen-buckling analyses.

holes based on thin shell finite element eigen-buckling stud-

ies [20]. These studies demonstrated that local buckling can occur
either at the location of the hole as ‘‘unstiffened strip’’ buckling,
or away from a hole as plate buckling, depending upon the plate
and hole geometry. This idea is extended in the following sections
from plates to entire members by utilizing the finite strip method
to evaluate the buckling of the ‘‘unstiffened strip’’ adjacent to a hole
including the influence of connected elements in the cross-section.

4.1. Local buckling load of cold-formed steel columns with holes

4.1.1. Prediction equations and implementation

The local critical elastic buckling load, Pcr ` , is calculated for a
cold-formed steel column with holes as

Pcr ` = min(Pcr `nh , Pcr `h ), (25)

where Pcr `nh is the local buckling load of the gross section, ignoring
Fig. 18. Finite strip approximation of Mcrd is consistent with ABAQUS predictions.
the hole, which may be calculated by finite strip or other methods
(see commentary to Appendix 1 of the AISI-S100 [1]). The buckling
Although the longitudinal stress distribution on a web plate is load including the hole, Pcr `h , may be calculated by a finite strip
different for a column and a beam, the rotational stiffness is still analysis of the net cross-section (e.g. in CUFSM), but restraining
linearly related to the plate rigidity D (see Section 3.1.1). the deformations to local buckling (see Fig. 19) and examining
only those buckling half-wavelengths shorter than the length of
3.2.2. Verification for distortional buckling of a beam with web holes the hole.
A verification study was conducted across the parameter ranges When performing a finite strip analysis of the net section, it
summarized in Table 2 with the critical elastic distortional buckling is important to avoid fully restraining a cross-section element
moments of 144 previously-tested cold-formed beam specimens (i.e., flange or web), since this prevents Poisson-type deformations
containing evenly spaced web holes [19]. The buckling moments and artificially stiffens the cross-section. For example, Fig. 19a
including the influence of holes were obtained with eigen-buckling restrains the corners in the z-direction only to prevent distortional
analyses in ABAQUS (Fig. 17). Special care was taken to simulate buckling, while still accommodating transverse deformation of the
the experimental boundary conditions when modeling in ABAQUS, flanges. The only condition when a corner should be fixed in both
including simple supports with vertical roller restraints, aluminum the x and z directions is when two isolated elements intersect
angle straps connecting the top and bottom channel flanges, lateral (i.e., C-section with a flange hole, see Fig. 19a). In addition, when
bracing of the top flange in the constant moment region, and the a hole isolates an unsupported strip in the net cross-section (as
application of load through the webs in four point bending, see full shown in Fig. 19b, e.g., a hat section with flange holes), the isolated
details in [18]. Fig. 18 shows that the simplified prediction method portion of the cross-section should be deleted since it is assumed
is accurate over a wide range of hole depths (i.e. hhole /h where h is to no longer contribute to the stiffness.
the flat depth of the web) when compared to the ABAQUS results. Once the net cross-section is restrained, an eigen-buckling
(Note that the influence of the boundary conditions is isolated from analysis is performed, and an elastic buckling curve similar to
the influence of holes on the ABAQUS Mcrd predictions to allow for Fig. 20 is generated. Lcr `h is identified on the curve as the half-
a consistent comparison to the approximate method, see [18] for wavelength corresponding to the minimum buckling load. When
details.) Lhole < Lcr `h (Fig. 20a), Pcr `h is equal to the buckling load at the
length of the hole. (This case is common for circular and square
4. Local buckling of cold-formed steel columns and beams with holes, where Lhole is less than the width of the cross-sectional
holes element containing the hole.) If Lhole ≥ Lcr `h (Fig. 20b), Pcr `h is
the minimum on the buckling curve. To ensure force equilibrium
Recent research motivated by the extension of DSM to cold- between the net section and the end of the column [20] and a
formed steel members with holes has resulted in simplified consistent comparison of Pcr `h to Pcr `nh , the reference stress used
element-based approximations of elastic buckling of plates with in the gross section and net-section finite strip analyses should be
 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824 2821

Fig. 19. Guidelines for modeling a column net cross-section in CUFSM, examples include (a) lipped C-section and (b) a hat section.

a b

Fig. 20. Local elastic buckling curve of net cross-section when (a) hole length is less than Lcr `h and (b) when hole length is greater than Lcr `h .

calculated with the same reference load (e.g., 1 kN applied to the eigen-buckling analyses are modeled with FSM-style boundary and
gross section, 1 kN applied to the net section). loading conditions identical to those shown in Fig. 4. The modulus
Use of the net cross-section for buckling half-wavelengths of elasticity, E, is assumed as 203.4 GPa and Poisson’s ratio, ν as
greater than Lhole is unduly conservative and fails to reflect the 0.3 in all finite strip and finite element models. Pcr ` is normalized
stiffness contributions of the gross section. Finite element analysis when plotted by Py , the squash load of the column calculated with
detailed in [20] demonstrates that local buckling waves triggered the gross cross-sectional area and a yield stress, Fy , of 345 MPa.
in the unstiffened strips adjacent to a hole are of primary concern. The first example is an SSMA 362S162-33 cross-section [9] with
Knowledge of the specific buckling half-wavelength of interest a slotted web hole. Fig. 21 compares the finite strip and ABAQUS
allows the finite strip method to be extended by utilizing the net mode shapes for hhole /hC = 0.14 where hC is the C-section web
section, but only for half-waves less than Lhole . depth measured from the centerline flange to centerline flange.
The finite strip approximate method predictions are plotted for
4.1.2. Verification for local buckling of a column with holes a range of hhole /hC (Fig. 21), and compared with ABAQUS eigen-
Two examples are presented here that approximate the local buckling predictions to demonstrate the viability of the prediction
critical elastic buckling load, Pcr ` , for cold-formed steel columns method. For this example, smaller hole widths lead to reductions in
with holes using the finite strip approximation detailed in the Pcr ` when compared to a member without a hole or members with
previous section. For both examples, the length of the column larger holes. This counter-intuitive result occurs because for small
L = 2540 mm and five slotted holes are spaced at S = 508 mm. holes the unstiffened strip controls the local buckling behavior
The typical length of the hole Lhole = 102 mm. All ABAQUS and for large holes, local plate buckling occurs between the holes.
2822 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824

Fig. 21. Comparison of CUFSM and ABAQUS predictions of unstiffened strip Fig. 23. The approximate finite strip approach for Pcr ` is a viable approach,
buckling. becoming more conservative as hhole /h increases.

account wavelength stiffening effects [20] which can boost Pcr `

as the hole(s) becomes large relative to the column cross-section
(Fig. 23). For those column specimens with small holes relative to
web width in Fig. 25, the holes are often circular or square. The
buckling load of the unstiffened strip, Pcr `h , is predicted higher than
Pcr `nh in these cases (i.e., unstiffened strip buckling is predicted not
to occur), which is consistent with the actual buckled behavior of
stiffened elements with circular and square circular holes [18,20].

4.2. Local buckling moment of beams with holes

4.2.1. Prediction equations and implementation

The local critical elastic buckling moment, Mcr ` , is calculated for
a cold-formed steel beam with holes as
Mcr ` = min(Mcr `nh , Mcr `h ). (26)
Local buckling is assumed to occur as the minimum of Mcr `nh of the
gross cross-section (again see commentary of the Direct Strength
Fig. 22. CUFSM and ABAQUS local buckling mode shapes are consistent when Method [1] for FE, FSM and analytical methods) and local buckling
considering a slotted flange hole. of the compressed unstiffened strip adjacent to the hole, Mcr `h . The
unstiffened strip buckling mode is evaluated with the procedure
(One must keep in mind that for strength the net section in described in Fig. 20, i.e. when Lhole > Lcr `h , Mcr `h occurs as the
yielding, as well as the elastic buckling load, ultimately determine minimum on the elastic buckling curve, otherwise Mcr `h is equal
the capacity, not just Pcr ` .) to the buckling moment at Lhole . To ensure moment equilibrium
The next example evaluates the influence of a slotted flange between the net section and the end of the beam, and a consistent
hole on Pcr ` for an SSMA 362S162-33 cross-section. The finite comparison of Mcr `h to Mcr `nh , the reference stress used in the gross
strip prediction method results are plotted for varying flange hole section and net-section finite strip analyses should produce the
width, bhole , relative to centerline flange width, bC , and compared same reference moment (e.g., 1 kN m applied to the gross section,
to ABAQUS eigen-buckling predictions in Fig. 22. Pcr `h decreases 1 kN m applied to the net section).
with increasing flange hole width for both CUFSM and ABAQUS To predict Mcr `h from the net cross-section in CUFSM, the cross-
results. The decreasing trend in the critical elastic buckling load section is restrained to isolate local buckling from distortional
demonstrates the importance of the flange in web local buckling buckling as shown in Fig. 24. Compressed corners should be
dominated cross-sections. Local buckling occurs primarily in the restrained in the direction normal to the neutral axis about which
web and flange strips in both CUFSM and ABAQUS mode shapes bending occurs as shown for the C-section in Fig. 24a and the
(Fig. 22). hat section in Fig. 24b. (Corners experiencing tension need not
A more extensive validation study was performed across the be restrained.) The only time both the x and z directions of a
parameter ranges in Table 2 by comparing the local buckling of corner should be restrained is if a hole isolates two compressed
78 column specimens determined with ABAQUS eigen-buckling intersecting elements (as in the case of a flange hole in a C-section,
analysis (the same database was used in Section 3.1.2 for the see Fig. 24a). Finally, when holes isolate a compressed element in
validation of the distortional buckling approximate method) to a cross-section (similar to the hat section with the flange hole in
the finite strip approximate method for predicting Pcr ` . Fig. 25 Fig. 19b), the isolated element should be removed from the cross-
compares the ABAQUS Pcr ` to the predicted Pcr ` for the 78 section.
specimens and demonstrates the approximate method is accurate
for smaller holes relative to column size and becomes conservative 4.2.2. Verification for local buckling of a beam with holes
for increasing hhole /h. One contribution to the conservative nature The finite strip prediction method is used to predict Mcr ` for the
of the predictions is that the approximation does not take into 144 cold-formed steel beam specimens described in Section 3.2.2
 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824 2823

Fig. 24. Guidelines for restraining beam net cross-sections in the finite strip local buckling approximate method for (a) a C-section and (b) a hat section in bending.

with explicit holes introduced) it is shown with properly reduced

cross-section properties flexural, and flexural–torsional buckling
in columns with holes as well as lateral–torsional buckling of
beams with holes may be readily predicted using classical closed-
form analytic expressions.
Distortional and local buckling of members with holes are
approximated using careful modifications to finite strip models
of the cross-section. Since the influence of the hole is different
on each buckling mode separate models are needed for each
case. In distortional buckling the primary impact of the hole is
to reduce the available rotational stiffness to the member under
distortional buckling. An expression for a reduced thickness is
provided, but this thickness is only used for determining the loss
in distortional buckling capacity at the original (no hole) buckling
half-wavelength. The approximation is verified against FE models
with explicit holes introduced. In local buckling a net-section
model of the cross-section in employed, which is similar to the
unstiffened strip idealization commonly used at the plate-level in
Fig. 25. Comparison of Mcr ` finite strip predictions to ABAQUS results. design of members with holes. The important difference is that
the net-section model of the cross-section is only investigated
and detailed in [18]. The predictions are compared to the ABAQUS for local buckling half-wavelengths less than the length of the
eigen-buckling results and demonstrates that the finite strip hole. In addition a gross section model must also be checked for
approximate method is viable over a wide range of hole widths and local buckling as local buckling may occur at, or away from, the
beam depths. The prediction method is shown to be conservative hole — depending on the hole size and member geometry. This
when web hole depth is small relative to the beam depth because approximation is again verified using FE models with explicit holes.
unstiffened strip buckling is predicted although local buckling of Taken together the approximate methods provide an important
the full cross-section is actually occurring. building block for the extension of the Direct Strength Method to
cold-formed steel structural members with holes.
5. Conclusions
Acknowledgement
Global, distortional, and local buckling of cold-formed steel
members with holes can be approximated using simple, unam- The American Iron and Steel Institute’s sponsorship of this
biguous procedures that may readily be employed in engineering research is gratefully acknowledged.
design. Global buckling of members with holes may be approx-
imated through careful application of a ‘‘weighted average’’ ap-
References
proach in determination of member cross-section properties (I, J,
Cw , etc.). Even hole location along the member length may be prop- [1] AISI-S100. North American specification for the design of cold-formed steel
erly treated in such an approach. The primary difficulty is the cor- structural members. Washington (DC): American Iron and Steel Institute;
rect handling of the ‘‘weighted average’’ in warping torsion. It is 2007.
[2] Schafer BW. Local, distortional, and Euler buckling of thin-walled columns.
shown that a hole significantly disturbs the shear stress in a re- J Struct Eng 2002;128(3):289–99.
gion around the holes, and this must be accounted for, as pro- [3] Sarawit A. Cold-formed steel frame and beam–column design. Ph.D. thesis,
vided herein. Using examples and validation studies (FE models Cornell University, Ithaca, 2003.
2824 C.D. Moen, B.W. Schafer / Engineering Structures 31 (2009) 2812–2824

[4] Falk S. Die Knickformeln fuer den Stab mit n Teilstuecken konstanter [13] Moen CD. Perforated column Euler buckling calculator, V 1.0 www.moen.cee.
Biegesteifigkeit (Buckling formulas for members with n segments of constant vt.edu, accessed June 2009.
bending resistance). Ingenieur–Archiv 1956;24(2):85–91. [14] Microsoft. Microsoft Office Excel 2007, V 12.0.6341.5001, www.microsoft.
[5] Timoshenko SP. Buckling of bars with variable cross-section. Kiev: Bull. com. 2006.
Polytech. Inst.; 1908. [15] Lau SCW, Hancock GJ. Distortional buckling formulas for channel columns.
[6] Timoshenko SP, Gere JM. Theory of elastic stability. New York: McGraw-Hill; ASCE J Struct Eng 1987;113(5):1063–78.
1961. [16] Schafer BW. Design aids and examples for distortional buckling, Technical
[7] Simitses G, Hodges D. Fundamentals of structural stability. Boston (MA): Note G100-08. Washington, D.C.: Cold-Formed Steel Engineers Institute,
Butterworth-Heinemann; 2006. www.cfsei.org, 2008.
[8] ABAQUS. ABAQUS/Standard Version 6.7-3. Providence (RI): Dassault Systèmes, [17] Kesti J. Local and distortional buckling of perforated steel wall studs.
http://www.simulia.com/, 2007. Dissertation/Thesis, Helsinki University of Technology, Espoo, Finland,
[9] SSMA. Product technical information, ICBO ER-4943P. Steel stud manufactur- 2000.
ers association, www.ssma.com, 2001. [18] Moen CD. Direct strength design for cold-formed steel members with
[10] Chajes A. Principles of structural stability. Englewood Cliffs (NJ): Prentice Hall perforations. Ph.D. thesis, Johns Hopkins University, Baltimore, 2008.
College Div; 1974. [19] Shan MY, LaBoube RA. 1994, Behavior of web elements with openings
[11] Schafer BW, Ádàny S. Buckling analysis of cold-formed steel members using subjected to bending, shear, and the combination of bending and shear. Civil
CUFSM: conventional and constrained finite strip methods. In: Eighteenth engineering study 94-2, cold-formed steel series, University of Missouri-Rolla,
international specialty conference on cold-formed steel structures. 2006. Rolla.
[12] Sarawit A. CUTWP Thin-walled section properties, December 2006 update [20] Moen CD, Schafer BW. Elastic buckling of thin plates with holes in compression
www.ce.jhu.edu/bschafer/cutwp, accessed January 2008. or bending. Thin-Walled Struct 2009.