LATERALLY BRACED COLD-FORMED STEEL FLEXURAL MEMBERS

WITH EDGE STIFFENED FLANGES

By B. W. Schafer1 and T. Peköz,2 Members, ASCE

ABSTRACT: The moment capacity of a laterally braced cold-formed steel flexural member with edge stiffened
flanges (e.g., a channel or zee section) may be affected adversely by local or distortional buckling. New pro-
cedures for hand prediction of the buckling stress in the local and distortional mode are presented and verified.
Numerical investigations are employed to highlight postbuckling behavior unique to the distortional mode.
Compared with the local mode, the distortional mode is shown to have (1) heightened imperfection sensitivity,
(2) lower postbuckling capacity, and (3) the ability to control the failure mechanism even in cases when the
elastic buckling stress in the local mode is lower than in the distortional mode. Traditional design methods do
not explicitly recognize distortional buckling, nor do they account for the observed phenomena in this mode. A
new design method that integrates distortional buckling into the unified effective width approach, currently used
in most cold-formed steel design specifications, is presented. For each element a local buckling stress and a
reduced distortional buckling stress are compared to determine the effective width. Comparison with experi-
mental tests shows that the new approach is more consistent and reliable than existing design methods.

INTRODUCTION the objective here. The procedure begins with closed-form pre-
diction of the local and distortional buckling stresses. Inter-
Finite strip analysis of a flexural member with an edge stiff- action of the flange, web, and lip, in both local and distortional
ened flange (Fig. 1) reveals three fundamental buckling modes: buckling, is considered. A need for the integration of the dis-
local, distortional, and lateral-torsional. For a laterally braced tortional mode into the design procedure is highlighted by two
flexural member the lateral-torsional buckling mode is re- behavioral phenomena. First, the distortional mode has less
stricted. Therefore, the two primary modes of concern are local postbuckling capacity than the local mode. Second, the dis-
and distortional buckling. tortional mode has the ability to control failure even when it
The American Iron and Steel Institute (AISI) Specification occurs at a higher critical stress than the local mode. A design
for the design of cold-formed steel structural members (AISI method incorporating these phenomena is needed to provide
1996), hereon referred to as the AISI specification, attempts to an integrated approach to strength prediction involving local
account for distortional buckling through an empirical reduc- and distortional buckling.
tion of the local plate buckling coefficient k. The empirical k
values do not agree with the actual distortional buckling stress. ELASTIC BUCKLING
The experimental work (Desmond et al. 1981) that was
conducted to determine the empirical k expressions actually Elastic buckling of cold-formed steel members can be pre-
concentrated efforts on local buckling of the flange. This dicted readily by numerical methods. However, for design pur-
was accomplished by testing back-to-back sections. This ex- poses, closed-form solutions still are required. Therefore, new
perimental setup strongly restricts buckling in the web and hand methods are developed for prediction of the buckling
hence distortional buckling as well. More recent experiments stress in the local and distortional modes.
on laterally braced flexural members with edge stiffened
flanges by Willis and Wallace (1990), Schuster (1992), Mor- LOCAL BUCKLING PREDICTION
eyra (1993), and Ellifritt et al. (1997) demonstrate unconser- An element model and a semiempirical interaction model
vative strength predictions using the AISI specification. are presented for closed-form approximation of the buckling
A hand method for the prediction of the distortional buck- stress in the local mode [see Fig. 1(a)]. The element model
ling stress in compression members was derived by Lau and ignores interaction of the flange, web, and/or lip and treats the
Hancock (1987). Hancock extended this approach to flexural buckling of each element independently as is done in the cur-
members in Hancock (1995, 1997). In Hancock et al. (1996) rent AISI specification. For instance, for a compression flange,
a method for evaluating the strength in distortional buckling
is proposed. Hancock et al.’s method provides an independent
strength calculation for distortional buckling. The suggested
design strength is the minimum of the AISI specification
method and a distortional buckling method. Comparison of
this approach with test data is favorable, though the method
proves overly conservative in many cases.
A unified treatment of local and distortional buckling in lat-
erally braced flexural members with edge stiffened flanges is
1
Sr. Engr., Simpson, Gumpertz & Heger, Inc., 297 Broadway Ave.,
Arlington, MA 02474; formerly, Instructor, Cornell Univ., Ithaca, NY
14853.
2
Prof., Cornell Univ., 220 Hollister Hall, Ithaca, NY.
Note. Associate Editor: C. Dale Buckner. Discussion open until July
1, 1999. To extend the closing date one month, a written request must
be filed with the ASCE Manager of Journals. The manuscript for this
paper was submitted for review and possible publication on February 2,
1998. This paper is part of the Journal of Structural Engineering, Vol. FIG. 1. Finite Strip Analysis of Flexural Member with Edge
125, No. 2, February, 1999. 䉷ASCE, ISSN 0733-9445/99/0002-0118 – Stiffened Flange: (a) Local Buckling; (b) Distortional Buckling;
0127/$8.00 ⫹ $.50 per page. Paper No. 17532. (c) Lateral-Torsional Buckling

118 / JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999

 FIG. 2. Local Buckling of Isolated Flange and Lip FIG. 3. Finite Strip Analysis of Isolated Flange and Lip

it is assumed that the element is simply supported on all four DISTORTIONAL BUCKLING PREDICTION
sides and thus a plate buckling coefficient of k = 4 is em-
ployed. For the semiempirical interaction model, local buck- Prediction of distortional buckling, as shown in Fig. 1(b),
ling of the flange is influenced by its attachment to a lip and is complicated because of the sensitivity to the solution to the
a web. rotational restraint at the web/compression flange juncture.
Expressions for the plate buckling coefficients for the ele- Consider an isolated flange and lip (similar to the inset of Fig.
ment model and the semiempirical interaction model follow. 2) in which the web/flange juncture is idealized as either a
All of the k values are written in terms of the critical buckling simple support or a fixed support. Finite strip analysis (Fig. 3)
stress of the flange, where shows that for local buckling the change in the plate buckling
coefficient is small regardless of the boundary condition. How-
␲2D ever, for distortional buckling the potential differences are sig-
fcr = k (1) nificant.
b2t
Closed-form prediction of the distortional buckling stress is
Several of the elements are subjected to a stress gradient, based on an examination of the rotational restraint at the web/
which is defined in terms of ␰ flange juncture. The rotational stiffness may be expressed as
f1 ⫺ f2 the summation of the elastic and stress-dependent geometric
␰= (2) stiffness terms with contributions from both the flange and the
f1 web
where f1 and f2 = stresses at the opposite edges of the element. k␾ = (k␾ f ⫹ k␾w)e ⫺ (k␾ f ⫹ k␾w)g (10)
For the web, f1 is at the web/compression flange juncture. For
the lip, f1 is at the lip/compression flange juncture. Compres- Buckling ensues when the elastic stiffness at the web/flange
sion stresses are positive (tension stresses negative). juncture is eroded by the geometric stiffness, i.e.
Element model k␾ = 0 (11)
flange: ( fcr)f k=4 (3) Using (11) and writing the stress-dependent portion of the ge-
ometric stiffness explicitly
web: ( fcr)w k = (0.5␰ 3web ⫹ 4␰ web
2
⫹ 4)(b/h)2 (4)
k␾ = k␾ fe ⫹ k␾we ⫺ f (k˜ ␾ fg ⫹ k˜ ␾wg) = 0 (12)
lip: ( fcr)l k = klip(b/d )2 (5)
Therefore, the critical buckling stress ( f ) is
for 0 < ␰lip ⱕ 1.1 2
klip = 1.4␰ lip ⫺ 0.25␰lip ⫹ 0.425 (6)
k␾ fe ⫹ k␾we
f=˜ (13)
for 1.1 < ␰lip ⱕ 2 3
klip = 13␰ lip ⫺ 65.5␰ lip
2
⫹ 131␰lip ⫺ 80 (7) k␾ fg ⫹ k˜ ␾wg
Semiempirical interaction model Analytical models are needed for determining the rotational
stiffness contributions from the flange and the web. For the
flange/lip: ( fcr)fl k = (8.55␰lip ⫺ 11.07)(d/b)
2
flange, cross-sectional distortion is not important [Fig. 1(b)];
⫹ (⫺1.59␰lip ⫹ 3.95)(d/b) ⫹ 4 hence the flange is modeled as a column undergoing torsional-
flexural buckling. This is similar to the approach of Sharp
for ␰lip ⱕ 1 and d/b ⱕ 0.6 (8) (1966), Lau (1988), Seah and Rhodes (1993), Davies and Jiang
(1996), and Hancock (1997). For the web, cross-sectional dis-
flange/web: ( fcr)fw k = 1.125 min{4, (0.5␰ 3web ⫹ 4␰ 2web ⫹ 4)(b/h)2} (9) tortion must be considered, so the web is modeled as a single
With the exception of the k = 4 solution, all of the foregoing finite strip. Therefore, the transverse shape function is a cubic
expressions are new. The equations are determined by fitting polynomial. The longitudinal shape functions of the flange and
expressions to finite strip analysis results. Fig. 2 shows the web are matched by using a single half-sine wave for each.
comparison for the local buckling expressions of an isolated
Distortional Buckling — Model for the Flange
flange and lip. The element model provides a lower bound,
whereas the semiempirical interaction model closely approxi- Consider the torsional-flexural buckling of a column with
mates the finite strip analysis, within the prescribed param- springs along one edge as shown in Fig. 4. The governing
eters. differential equations are
JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999 / 119
EIyf
d 4u
dz4
⫹ EIxyf
d 4v
dz4
⫹ P
d 2u
dz2
⫹ y0
d 2␾
dz2 冉 冊
⫹ kxf (u ⫹ ( y0 ⫺ hy)␾) = 0 (14)

EIxf
d v
dz4
4
d u
⫹ EIxyf 4 ⫹ P
dz
d v
dz2
d ␾ 4

冉 2

⫺ x0 2 ⫹ kyf (v ⫺ (x0 ⫺ hx)␾) = 0

dz
2

冊
(15)

ECwf
d 4␾
dz4
⫺ GJf ⫺
Iof
Af
P
dz2
冉
d 2␾
⫺ P x0
d 2v
dz2
⫺ y 0 冊
d 2u
dz2
冉 冊 FIG. 4. Flange Model

b2 ⫺ d 2 cos(␪)
x0 = (26)
⫹ kxf (u ⫹ ( y0 ⫺ hy)␾)( y0 ⫺ hy) ⫺ kyf (v ⫺ (x0 ⫺ hx)␾)(x0 ⫺ hx) 2(b ⫹ d )

⫹ k␾ f ␾ = 0 (16) ⫺d 2 sin(␪)
hy = y0 = (27)
2(b ⫹ d )
where Ixf , Iyf , Ixyf , Iof , Cwf , Jf , and A f = section properties of the
flange; kxf , kyf , and k␾f = springs; x0 and y0 = distances from ⫺(b2 ⫹ 2db ⫹ d 2 cos(␪))
the centroid to the shear center; and hx and hy = distances from hx = (28)
2(b ⫹ d )
the centroid to the springs. The following shape functions,
consistent with a simply supported column, are used: x0 ⫺ hx = b (29)

␾ = A1 sin 冉冊 ␲z
L
, u = A2 sin 冉冊 ␲z
L
(17a,b)
Distortional Buckling — Model for the Web
Cwf = 0 (30)

v = (x0 ⫺ hx)A1 sin

␲z
L 冉冊 (17c) Derivation of the rotational restraint provided by the web to
the web/flange juncture is based on using a single finite strip
For this application the kx f spring stiffness is assumed zero and as shown in Fig. 5. The finite strip solution for the plate bend-

再 冎 冋冋 册
the kyf spring stiffness is assumed infinite. The typical approach ing terms may be represented symbolically as
is to find the buckling load Pcr. However, the goal here is to F1 k11 k12 k13 k14
write the solution in terms of the rotational restraint the flange M1 k21 k22 k23 k24
provides at the web/flange juncture. The shape functions in =
F2 k31 k32 k33 k34
(17) are substituted into (14) – (16) and the load P is written

冋冋 册册册再 冎
M2 k41 k42 k43 k44
in terms of the uniform stress f1. If terms of order f 2 are ne- E

glected then the flange rotational restraint may be written in k11 k12 k13 k14 w1
the linear form given in (12). The resulting rotational stiffness k21 k22 k23 k24 ␪1
terms are ⫺
k31 k32 k33 k34 w2

冉冊冉 冊 ␪2
4
␲ 2
I xyf k41 k42 k43 k44 G
(31)
k␾ fe = EIxf (x0 ⫺ hx)2 ⫹ ECwf ⫺ E (x0 ⫺ hx)2
L Iyf where F and M = consistent nodal loads or moments; and kij

冉冊
2 = stiffness coefficients for the plate bending finite strip matrix
␲ [e.g., Cheung (1976)]. For simply supported edges, the terms
⫹ GJf
L (18) of interest are

k̃␾ fg =冉冊冋 冉 冉冊␲

L
2

Af (x0 ⫺ hx)2
Ixyf
Iyf
2

⫺ 2y0(x0 ⫺ hx) 冉冊 Ixyf

Iyf
再 冎 冋冋
M1
M2
=
k22
k42
k24
k44 册 冋E
⫺
k22
k42 册 册再 冎
k24
k44 G
␪1
␪2
(32)

To find k␾w, consider the strip to be unloaded along edge

⫹h ⫹y 冊 册
2
x
2
0 ⫹ Ixf ⫹ Iyf
(19)
two and loaded along edge one (the web/compression flange
juncture) with a sinusoidal edge moment of M sin(␲y/L). The
consistent nodal moments M2 = 0 and M1 = 1/2ML are sub-
For a simple lip stiffened flange (Fig. 4) the section prop- stituted. The solution then is written in terms of ␪1. If ␪1 = 1,
erties in (18) and (19) are only a function of b, d, ␪, and t then M = k␾w; therefore

Af = (b ⫹ d )t
Jf = 1/3bt 3 ⫹ 1/3dt 3
(20)
(21)
k␾w =
2
L 冉 (k22e ⫺ k22g) ⫺
(k24e ⫺ k24g)(k42e ⫺ k42g)
(k44e ⫺ k44g) 冊 (33)

The web rotational stiffness k␾w is decomposed into elastic and

t(t 2b2 ⫹ 4bd 3 ⫺ 4bd 3 cos2(␪) ⫹ t 2bd ⫹ d 4 ⫺ d 4 cos2(␪)) geometric parts
Ixf =
12(b ⫹ d )
k␾w = k␾we ⫺ k␾wg (34)
(22)

Iyf =
t(b4 ⫹ 4db3 ⫹ 6d 2b2 cos(␪) ⫹ 4d 3b cos2(␪) ⫹ d 4 cos2(␪))
12(b ⫹ d )
k␾we =
2
L
冉 k22e ⫺
2
k24e
k22e
冊 (35)

(23)
tbd sin(␪)(b ⫹ d cos(␪))
2
⫺k␾wg =
2
L
冉冋 (k22e ⫺ k22g) ⫺
(k24e ⫺ k24g)(k42e ⫺ k42g)
(k44e ⫺ k44g)
册
冉 冊冊
Ixyf = (24)
4(b ⫹ d ) k224e
⫺ k22e ⫺
tb3 bt 3 td 3 k22e (36)
Iof = ⫹ ⫹ (25)
3 12 3 The kij terms may be substituted directly to yield the com-
120 / JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999
 FIG. 5. Finite Strip Idealization of Web

plete analytical expressions for k␾. Although exact, the ex- L. This minimization is complicated by the K4 term — the web
pressions have an inordinate number of terms. Simplifications geometric stiffness. If the f (L) in the K4 term is approximated
are made to provide a more compact solution. The elastic ro- as 1/L2 then the C3 and K4 terms drop out. This assumption is
tational stiffness in (35) is truncated by converting to partial made; therefore, the general solution for Lcr is

冉 冊
fractions on the length L and keeping the constant term, the 1/4

1/L2 term, and the 1/L4 term. The resulting expression asymp- df C1 ⫹ K 3
= 0 → Lcr = (47)
totes to the full expression and provides a reasonable approx- dL K1
imation of the elastic rotational stiffness The appropriate terms for C1, K 3, and K1 are substituted, re-

冉 冉冊 冉冊 冊
2 4
3 ␲ 19h ␲ 3
h sulting in

冉 冉 冊
k␾we = D ⫹ ⫹ (37)
h L 60 L 240 4␲4h(1 ⫺ ␯2) 2
I xyf
Lcr = Ixf (x0 ⫺ hx)2 ⫹ Cwf ⫺ (x0 ⫺ hx)2
For the geometric rotational stiffness in (36) the first ap- t3 Iyf
proximation made is to linearize the stress f1. This is adequate
冊
1/4
␲4h4
for stress gradients near pure bending (␰web ⬃ 2), but breaks ⫹
down as the stress approaches pure compression (␰web = 0). 720 (48)
With this simplification, the geometric rotational stiffness takes If the flange is assumed to be pinned [as is done in the
the form critical length derivation of Lau (1988)] then the (Ixyf)2/Iyf term

⫺k␾wg =
2 f1
L
冉
2k22e k24e k24g ⫺ k22g k224e ⫺ k22g k222e
k222e
冊 (38)
is assumed negligible.

Elastic Distortional Buckling — Summarized

Further simplification is provided after substituting in the kij To find the critical buckling stress in the distortional mode
terms by converting the solution to partial fractions on the ( fcr)dist, use (13). The rotational stiffness terms in (13) are
length L. The general expression is then expressed in three found in (18), (19), (37), and (41). The rotational stiffness
terms, in which the dominators are terms should be evaluated at Lcr via (48) unless Lb < Lcr.
— — —
⫹ 4 ⫹ 4 (39) COMPARISON OF ELASTIC BUCKLING METHODS
L2 (h ⫹ L2h2 ⫹ L4) (h ⫹ L2h2 ⫹ L4)2
Thirty-two members are examined via finite strip analysis
Parametric analysis shows the final term to be insignificant;
to compare with the proposed hand methods. The critical local
thus it is neglected. The first two terms are combined to form
buckling moment (Mlocal) and critical distortional buckling mo-
the approximation of the rotational geometric stiffness, where
ment (Mdist) are recovered from the finite strip analysis. The
k␾w = k␾we ⫺ f1k˜ ␾wg (40) geometry of the studied members is summarized in Table 1
and the comparison of the predictions is shown in Table 2.
2
ht␲
k̃␾wg = {[(45360(1 ⫺ ␰web) ⫹ 62160)(L/h)2 ⫹ 448␲2 TABLE 1. Geometry of Members
13440
⫹ (h/L)2(53 ⫹ 3(1 ⫺ ␰web))␲4]/[␲4 ⫹ 28␲2(L/h)2 h b d ␪
(1) (2) (3) (4)
⫹ 420(L/h)4]} (41) 50 25 6.25, 12.5 45, 90
100 25 6.25, 12.5 45, 90
50 6.25, 12.5, 25 45, 90
Distortional Buckling — Critical Length 150 25 6.25, 12.5 45, 90
50 6.25, 12.5, 25 45, 90
The buckling stress is a function of length L; hence to ap- 75 6.25, 12.5, 25, 37.5 45, 90
proximate the L at which f is a minimum, the rotational stiff- 200 25 6.25, 12.5 45, 90
ness terms are written explicitly in terms of L 50 6.25, 12.5, 25 45, 90
75 6.25, 12.5, 25, 37.5 45, 90
k␾fe = (1/L)4C1 ⫹ (1/L)2C2 (42) 100 6.25, 12.5, 25, 37.5, 50 45, 90

k̃␾fg = (1/L)2C3 (43)

TABLE 2. Performance of Elastic Buckling Methods
k␾we = K1 ⫹ (1/L) K 2 ⫹ (1/L) K 3
2 4
(44)
DISTORTIONAL
k̃␾wg = K4( f (L)) (45) LOCAL BUCKLING BUCKLING
Element Interaction Proposed
This gives the solution for the distortional buckling stress f
Model Model method
as
Statistics Mpredicted /Mlocal Mpredicted /Mlocal Mpredicted /Mdist
(1/L)4C1 ⫹ (1/L)2C2 ⫹ K1 ⫹ (1/L)2K 2 ⫹ (1/L)4K 3 (1) (2) (3) (4)
f= (46)
(1/L)2C3 ⫹ K4( f (L)) Average 0.74 0.90 0.95
Standard deviation 0.12 0.05 0.08
The critical length is found by minimizing f with respect to
JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999 / 121
 TABLE 3. Geometry of Edge Stiffened Flanges A longitudinal through thickness flexural residual stress of
Pcr,local
30% fy also is modeled.
The geometry of the members investigated is summarized
b/t d/t ␪ Pcr,dist
in Table 3. The thickness is 1 mm and fy = 345 MPa. Based
(1) (2) (3) (4)
on the displaced shape and location of plasticity two basic
25 4.00 – 19.0 90 1.82 – 0.25 failure mechanisms from the finite element analysis are iden-
6.25 – 12.5 45 19.4 – 0.96 tified. It is observed that the final failure mechanism is con-
50 5.00 – 25.0 90 1.58 – 0.27
6.25 – 25.0 45 1.76 – 0.51 sistent with the distortional mode even in cases when the dis-
75 6.25 – 37.5 90 1.34 – 0.18 tortional buckling stress is higher than the local buckling
6.25 – 37.5 45 1.73 – 0.35 stress. Consider Fig. 6, which shows the final failure mecha-
100 6.25 – 50.0 90 1.40 – 0.14 nism for all the members studied. Based solely on elastic buck-
6.25 – 50.0 45 1.75 – 0.23 ling, one would expect the local buckling mode to control in
all cases in which ( fcr)local /( fcr)dist < 1; as the figure shows, this
is not the case.
The models proposed for the local buckling stress in (3) – Finite element analysis also reveals that the postbuckling
(9) do not directly provide a direct prediction of the critical capacity in the distortional mode is less than that in the local
buckling moment. For the element model the governing local mode. Consider Fig. 7, for the same slenderness values the
buckling stress is assumed to be the minimum of the flange, distortional failures exhibit a lower ultimate strength. Similar
web, or lip. For the interaction model the governing local loss in strength is observed experimentally and summarized in
buckling stress is assumed to be the minimum of the flange/ Hancock et al. (1994).
lip and flange/web calculation. The governing local buckling The geometric imperfections are modeled as a superposition
stress then is used to determine the local buckling moment. of the local and distortional buckling modes. The magnitude
For local buckling prediction the interaction model performs of the imperfection is selected based on the statistical summary
markedly better than the element model. The overly conser- provided in Schafer and Peköz (1998). The error bars in Fig.
vative nature of the element model is driven largely by poor 7 demonstrate the range of strengths predicted for imperfec-
predictions when the lip controls the local buckling stress. In tions varying over the central 50% portion of expected im-
cases when the lip controls, the average Mpredicted /Mlocal ratio is perfection magnitudes. The greater the error bars, the greater
0.6. The proposed distortional buckling method compares fa- the imperfection sensitivity.
vorably with the finite strip analysis. Predictions for the ␪ = The percent difference in the strength over the central 50%
45⬚ members are slightly less conservative than for the ␪ = portion of expected imperfection magnitudes is used as a mea-
90⬚ members. Fortunately, the ratio for the ␪ = 45⬚ degree sure of imperfection sensitivity
members is still 0.98 and the standard deviation is lower than ( fu)25%imp ⫺ ( fu)75%imp
for the ␪ = 90⬚ members. ⫻ 100% (49)
1
(( fu)25%imp ⫹ ( fu)75%imp)
2
POSTBUCKLING BEHAVIOR OF EDGE
A contour plot of this imperfection sensitivity statistic (49)
STIFFENED ELEMENTS
is shown in Fig. 8. Stocky members prone to failure in the
To investigate the postbuckling behavior in the local and distortional mode have the greatest sensitivity. In general, dis-
distortional modes, nonlinear finite-element analysis of iso- tortional failures are more sensitive to initial imperfections
lated flanges is completed using ABAQUS (Hibbitt, Karlsson than local failures. Areas of imperfection sensitivity risk are
& Sorensen, Inc. 1995). The flange is modeled as fixed at the assigned.
web/flange juncture and nine node shell elements (S9R5) are
employed. The material model is elastic-plastic with strain DESIGN OF FLEXURAL MEMBERS
hardening. Initial imperfections in the local and distortional The current AISI specification approach for the capacity of
mode are superposed to form the initial imperfect geometry. a laterally braced flexural member involves determining an

FIG. 6. Failure Mode of Edge Stiffened Flanges

122 / JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999

 FIG. 7. Postbuckling Capacity of Edge Stiffened Flanges

FIG. 8. Imperfection Sensitivity of Edge Stiffened Flanges

proach to all members of the cross section is based on the

unified effective width approach of Peköz (1987). Once the
effective width is calculated, determination of the flexural
strength becomes a relatively straightforward manner, as
shown in Table 4.

Design — Effective Width of Elements

The effective width of the flange (or lip, replace b with d )
is
be = ␳b (50)
where ␳ is defined as
FIG. 9. Typical (a) Gross and (b) Effective Sections ␳ = (1 ⫺ 0.22/␭)/␭ for ␭ > 0.673 otherwise ␳=1 (51)

effective section modulus to account for local buckling. As The slenderness parameter ␭ is
shown in Fig. 9, each element is reduced from its gross width ␭ = 兹 fy /( fcr)flange (52)
(e.g., b) to an effective width (e.g., be). The reduction is based
on an empirical correction to the work of von Kármán et al. Portioning of the effective width for the flange is simple
(1932) completed by Winter (1947). The extension of this ap- (Fig. 9). However, in the case of a stiffened element under a
JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999 / 123
 TABLE 4. Example for Effective Section Calculation
Element A y Ay Ay 2 Iown
(1) (2) (3) (4) (5) (6)
Compression
flange bet — — — —
Web 1 h1t h1 /2 (h1)2t/2 (h1)3t/4 t(h1)3t/12
Web 2 h2t h ⫺ (ht ⫹ h2 /2) — — —
Web 3 htt h ⫺ (ht /2) — — —
Tensor flange bt h — — —
Compression lip det de /2 — — —
Tensor lip dt h ⫺ d/2 — — —

冘 A 冘 冘 冘 Ay Ay 2 Iown

冘 Ay
冘 冘 冉冘 冊
冘
Ieff
Mn = Seff f1 Seff = yeff = Ieff = Ay 2 ⫹ Iown ⫺ A y 2eff
yeff A

FIG. 10. Simply Supported Plate in Pure Bending

stress gradient (i.e., the web), the portioning of h to h1, h 2, and
ht is not straightforward. The expressions currently used in the
AISI specification for a stiffened element under a stress gra- 0 ⱕ ␳ < 0.77 ␻ = 0.30␳ (57a)
dient are discontinuous (Cohen 1987) and unconservative (Fig. 0.77 ⱕ ␳ < 0.95 ␻ = 0.23 (57b)
10). Other specifications, such as the Canadian standard for
cold-formed steel structural members (Canadian Standards As- 0.95 ⱕ ␳ ⱕ 1.00 ␻ = ⫺4.6␳ ⫹ 4.6 (57c)
sociation 1994), yield results more consistent with numerical
The resulting expressions agree with numerical analysis
analysis.
(Fig. 10). Further, the effective width of the web is a function
A new approach is proposed for the effective width of stiff-
of ␳. Thus, for the first time, the effective width of an element
ened elements under a stress gradient (i.e., webs). Consider
under a stress gradient is a function of the degree of non-
the effective width of an element in pure compression as
linearity in the postbuckling stress distribution, as reflected
shown in Fig. 11. Determination of the effective width is based
through ␳.
on (1) an approximation of the nonlinear postbuckling stress
via ␳ and (2) a force balance between the approximated non-
linear stress and the effective section. For an element under a Design — Integrating Distortional Buckling
stress gradient (Fig. 11) the natural extension to this method- into Procedure
ology is to determine the effective width by insuring that both If distortional buckling is considered then the critical buck-
a force and a moment balance are maintained between the ling stress of an element (flange, web, or lip) is no longer
approximated nonlinear stress and the effective section. The solely dependent on local buckling. To properly integrate dis-
solution of this force and moment balance result in the follow- tortional buckling, reduce postbuckling capacity in the distor-
ing expressions: tional mode and the ability of the distortional mode to control
h1 = h␻/␰web for ␰web > 0 (53) the failure mechanism even when at a higher buckling stress
than the local mode must be incorporated. Consider defining
h 2 = (h/␰web)兹␻ ⫺ 2␻ ⫹ ␳ 2
for ␰web > 0 (54) the critical buckling stress of the element used in (52) or (55)
␭ = 兹 fy /( fcr)web (55) as

␳ = (1 ⫺ 0.22/␭)/␭, for ␭ > 0.673, else ␳=1 (56) ( fcr) = min[( fcr)local, Rd ( fcr)dist] (58)

FIG. 11. Effective Width Determination

124 / JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999

 TABLE 5.
Model 1 ⬃ M1
Elastic Buckling Stress Determination

Model 2 ⬃ M2
Rd = min 冉 1,
1.17
␭d ⫹ 1 冊
⫹ 0.3 where ␭d = 兹 fy /( fcr)dist (60)
(no local buckling interaction) (local interaction included)
Two models are advanced for predicting the critical buck-
( fcr)web = min[( fcr)w, Rd ( fcr)d] ( fcr)web = min[( fcr)f w, Rd ( fcr)d] ling stress of the elements. The models are summarized in
( fcr)flange = min[( fcr)f , Rd ( fcr)d] ( fcr)flange = min[( fcr)f w , ( fcr)f l, Rd ( fcr)d] Table 5. With fcr of the element known, the effective width of
( fcr)lip = min[( fcr)l, Rd ( fcr)d] ( fcr)lip = min[( fcr)f l, Rd ( fcr)d]
each element may be determined readily. The procedure out-
lined in Table 4 is completed to calculate the section capacity.
For strength, if the reduced distortional mode governs, then Comparison with Experimental Data
(51) or (56) become
Experimental tests on laterally braced flexural members
␳ = 兹Rd(1 ⫺ 0.22兹Rd /␭)/␭ (59) with edge stiffened flanges from Winter (1947), Desmond et
al. (1981), LaBoube and Yu (1978), Schardt and Schrade
For Rd < 1, this method provides an additional reduction on (1982), Elhouar and Murray (1985), Cohen (1987), Willis and
the postbuckling capacity. Further, the method also allows the Wallace (1990), Ellifritt et al. (1992, 1997), Schuster (1992),
distortional mode to control in situations when the distortional Moreyra (1993), Shan et al. (1994), and Rogers and Schuster
buckling stress is greater than the local buckling stress. Thus, (1995) are gathered and examined. Based on the information
Rd provides a framework for solving the problem of predicting available from the tests, the type of sections tested, and the
the failure mode and reducing the postbuckling capacity in the loading arrangement, the applicability for use in this compar-
distortional mode. The selected form for Rd based on Figs. 6 ison is assessed. The experimental data of Winter (1947), Des-
and 7 and the experimental results of Hancock et al. (1994) is mond (1978), Elhouar and Murray (1985), and Ellifritt et al.

TABLE 6. Summary of Test to Predicted Ratios

JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999 / 125

 CONCLUSIONS
Laterally braced cold-formed steel flexural members with
edge stiffened flanges have two important buckling phenom-
ena: local and distortional. Current AISI specification methods
do not explicitly treat the distortional mode. Distortional buck-
ling deserves special attention because it has the ability to con-
trol the final failure mechanism in many cases and is observed
to have a lower postbuckling capacity than local buckling. New
hand methods are developed to predict the critical buckling
stress in both the local and the distortional mode. A design
method for strength prediction, based on the unified effective
FIG. 12. Performance of AISI Specification versus h/b width approach, is developed. The design method uses the new
expressions for prediction of the local and distortional buckling
(1992) are deemed to have poor applicability. Desmond’s and stress and also introduces a new approach for determining the
Winter’s tests use back-to-back webs, which provide an un- effective width of the web. The resulting design method is
realistic rotational restraint. Ellifritt et al.’s (1992) tests pri- compared with a large body of experimental results and is
marily fail in the lateral-torsional mode. Elhouar and Murray’s shown to provide more consistent and conservative prediction
(1985) summary of proprietary tests does not provide enough than the existing AISI specification. Proper incorporation of the
detailed information on loading and bracing. distortional buckling phenomena is imperative for accurate
The majority of the remaining tests are on face-to-face chan- strength prediction of cold-formed steel members.
nels in two-point bending. The channels typically have signif-
icant bracing at the load application point as well as a regularly ACKNOWLEDGMENTS
spaced angle or bar attached across the two channels in both The sponsorship of the AISI in conducting this research is gratefully
the compression flange and the tension flange. The bracing is acknowledged.
to insure that lateral-torsional buckling does not occur and to
approximate the effect of sheeting. The small spacing of the APPENDIX I. REFERENCES
attached angles or bars (often 300 mm ⬃12 in. or 150 mm
⬃6 in.) partially restricts the distortional mode. American Iron and Steel Institute (AISI). (1996). AISI specification for
the design of cold-formed steel structural members. American Iron and
If the bracing length (Lb) is less than the predicted Lcr from Steel Institute, Washington, D.C.
(41), then Lb is used in determining the distortional buckling Canadian Standards Association. (1994). S136-M94 cold-formed steel
stress. In many tests Lb < Lcr. Therefore capacity lower than structural members, Rexdale (Toronto), ON, Canada.
the experimentally observed strength is possible even for a Cheung, Y. K. (1976). Finite strip method in structural analysis. Perga-
laterally braced member, because of the fact that the distor- mon, Pergamon, New York.
tional mode is partially restricted. Cohen, J. M. (1987). ‘‘Local buckling behavior of plate elements.’’ Dept.
of Struct. Engrg. Rep., Cornell University, Ithaca, N.Y.
The flexural capacity of the remaining test data is assessed Davies, J. M., and Jiang, C. (1996). ‘‘Design of thin-walled beams for
via the AISI specification (MAISI) and the two proposed meth- distortional buckling.’’ Proc., 13th Int. Spec. Conf. on Cold-Formed
ods M1 and M 2 (Table 5). The statistical results are summarized Steel Struct., W. W. Yu and R. A. Laboube eds., University of Missouri,
in Table 6. One striking feature that Table 6 does not bring Rolla, Mo., 141 – 154.
out is the systematic error that exists in the current AISI spec- Desmond, T. P., Peköz, T., and Winter, G. (1981). ‘‘Edge stiffeners for
ification method for large h/b values (Fig. 12). thin-walled members.’’ J. Struct. Div., ASCE, 107, 329 – 353.
Elhouar, S., and Murray, T. M. (1985). ‘‘Adequacy of proposed AISI
From Table 6, the overall performance of the AISI specifi- effective width specification provisions for Z- and C-purlin design.’’
cation method appears adequate. A more detailed analysis re- Rep. No. FSEL/MBMA 85-04, Fears Structural Engineering Laboratory,
veals several inadequacies. For one, several of the individual University of Oklahoma, Norman, Okla.
tests groups yield consistently unconservative predictions (n < Ellifritt, D., Glover, G., and Hren, J. (1997). ‘‘Distortional buckling of
1). Second, the systematic error for large h/b is problematic. channels and zees not attached to sheathing.’’ Rep. No. 4504530-12,
Third, the AISI method is not a function of bracing length. American Iron and Steel Institute, Washington, D.C.
Ellifritt, D., Sputo, T., and Haynes, J. (1992). ‘‘Flexural capacity of dis-
Therefore, the same members at longer unbraced lengths (but cretely braced C’s and Z’s.’’ Proc., 11th Int. Spec. Conf. on Cold-
members still not failing in lateral-tortional buckling) have the Formed Steel Struct., W. W. Yu and R. A. Laboube eds., University of
same strength prediction via the AISI specification. This is Missouri, Rolla, Mo., 109 – 129.
inadequate — until Lb exceeds Lcr, the distortional buckling Hancock, G. J. (1995). ‘‘Design for distortional buckling of flexural mem-
stress and the strength will decrease. bers.’’ Proc., 3rd Int. Conf. on Steel and Aluminum Struct., G. A. Aşkar,
An integrated design method that employs local and distor- ed., Boǧaziçi University, Istanbul, Turkey, 275 – 284.
Hancock, G. J. (1997). ‘‘Design for distortional buckling of flexural mem-
tional buckling calculations is possible and reliable. The sys- bers.’’ Thin-Walled Struct., 27(1), 3 – 12.
tematic error for large h/b values observed in the AISI speci- Hancock, G. J., Kwon, Y. B., and Bernard, E. S. (1994). ‘‘Strength design
fication is alleviated in either of the proposed methods (M1 or curves for thin-walled sections undergoing distortional buckling.’’ J.
M 2). The test to predicted ratio is slightly on the conservative Constr. Steel Res., 31(2-3), 169 – 186.
side (>1) for the overall results of the proposed models. The Hancock, G. J., Rogers, C. A., and Schuster, R. M. (1996). ‘‘Comparison
standard deviation and number of unconservative predictions of the distortional buckling method for flexural members with tests.’’
Proc., 13th Int. Spec. Conf. on Cold-Formed Steel Struct., W. W. Yu
are both lower than the AISI specification for the overall results. and R. A. Laboube eds., University of Missouri, Rolla, Mo., 125 – 140.
Often, including the local buckling interaction (M 2) actually Hibbitt, Karlsson & Sorensen, Inc. (HKS). (1995). ABAQUS version 5.5.
yields a more conservative prediction than that determined by Hibbitt, Karlsson & Sorensen, Pawtucket, R.I.
ignoring it and using an element approach (M1). However, in- LaBoube, R. A., and Yu, W. (1978). ‘‘Structural behavior of beam webs
dividual cases are observed where including the local buckling subjected to bending stress.’’ Civil engineering study structural series,
interaction yields markedly better results. Local buckling initi- Rep. 78-1, Department of Civil Engineering, University of Missouri-
Rolla, Rolla, Mo.
ated by long edge stiffeners and local buckling with highly Lau, S. C. W. (1988). ‘‘Distortional buckling of thin-walled columns,’’
slender webs and compact flanges are examples where includ- PhD thesis, University of Sydney, Sydney, Australia.
ing the interaction is observed to improve the strength predic- Lau, S. C. W., and Hancock, G. J. (1987). ‘‘Distortional buckling formulas
tion markedly. for channel columns.’’ J. Struct. Engrg., ASCE, 113(5), 1063 – 1078.

126 / JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999

Moreyra, M. E. (1993). ‘‘The behavior of cold-formed lipped channels D = plate rigidity;
under bending,’’ MS thesis, Cornell University, Ithaca, N.Y. f = stress;
Peköz, T. (1987). ‘‘Development of a unified approach to the design of f1 = edge stress on an element;
cold-formed steel members.’’ AISI Res. Rep. No. CF 87-1, American f2 = edge stress on an element;
Iron and Steel Institute, Washington, D.C.
Rogers, C. A., and Schuster, R. M. (1995). ‘‘Interaction buckling of
fcr = minimum buckling stress;
flange, edge stiffener and web of C-sections in bending.’’ Res. Into fu = ultimate failure stress for a member;
Cold Formed Steel, Final Rep. of CSSBI/IRAP Proj., Department of fy = material yield stress;
Civil Engineering, University of Waterloo, Waterloo, Canada. h = web height;
Schafer, B. W., and Peköz, T. P. (1998). ‘‘Computational modeling of ht = portion of web in tension;
cold-formed steel: Characterizing geometric imperfections and residual h1 = portion of effective width of a web;
stresses.’’ J. Const. Steel Res., 47(3), 193 – 210. h2 = portion of effective width of a web;
Schardt, R., and Schrade, W. (1982). Kaltprofil-Pfetten. Institut Für Statik, k = plate buckling coefficient;
Technische Hochschule Darmstadt, Bericht Nr. 1, Darmstadt, Germany k␾ = rotational stiffness at web flange juncture;
(in German).
Schuster, R. M. (1992). ‘‘Testing of perforated C-stud sections in bend-
k␾fe = elastic rotational stiffness of flange;
ing.’’ Report for the Canadian Sheet Steel Building Institute, University k␾fg = geometric rotational stiffness of flange;
of Waterloo, Waterloo, ON, Canada. k␾we = elastic rotational stiffness of web;
Seah, L. K., and Rhodes, J. (1993). ‘‘Simplified buckling analysis of plate k␾wg = geometric rotational stiffness of web;
with compound edge stiffeners.’’ J. Engrg. Mech., ASCE, 119(1), k xf = flange model spring stiffness in x-direction;
19 – 38. k yf = flange model spring stiffness in y-direction;
Shan, M., LaBoube, R. A., and Yu, W. (1994). ‘‘Behavior of web elements k␾g = k␾g /f;
with openings subjected to bending, shear and the combination of bending L = length;
and shear.’’ Civil Engineering Study Structural Series, Rep. 94-2, De- Lb = unbraced length;
partment of Civil Engineering, University of Missouri-Rolla, Rolla, Mo.
Sharp, M. L. (1966). ‘‘Longitudinal stiffeners for compression members.’’ Lcr = length at which f is a minimum;
J. Struct. Div., ASCE, 102, 187 – 211. M = moment;
von Kármán, T., Sechler, E. E., and Donnell, L. H. (1932). ‘‘The strength M = consistent nodal moment;
of thin plates in compression.’’ Trans. ASME, 54, 53 – 57. M1 = moment capacity by proposed method 1;
Willis, C. T., and Wallace, B. (1990). ‘‘Behavior of cold-formed steel M2 = moment capacity by proposed method 2;
purlins under gravity loading.’’ J. Struct. Engrg., ASCE, 116(8), MAISI = moment capacity by AISI specification;
2061 – 2069. Rd = reduction factor for distortional buckling stress;
Winter, G. (1947). ‘‘Strength of thin steel compression flanges.’’ Trans. t = thickness;
ASCE, Paper No. 2305, 112, 1 – 50.
u = flange model displacement in x;
v = flange model displacement in y;
APPENDIX II. NOTATION
␪ = orientation angle of the edge stiffener (lip);
The following symbols are used in this paper: ␭ = slenderness;
␰ = stress gradient;
b = flange width; ␳ = postbuckling reduction factor; and
be = effective flange width; ␾ = flange model twist.

JOURNAL OF STRUCTURAL ENGINEERING / FEBRUARY 1999 / 127