Thin-Walled Structures 98 (2016) 565–577

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Full length article

Torsion of cold-formed steel lipped channels dominated by warping

response
Guanbo Bian a,n, Kara D. Peterman a,b, Shahabeddin Torabian a,c, Benjamin W. Schafer a
a
Department of Civil Engineering, Johns Hopkins University, Baltimore, MD, USA
b
Department of Civil & Environmental Engineering, Northeastern University, Boston, MA, USA
c
School of Civil Engineering, College of Engineering, University of Tehran, Iran

art ic l e i nf o a b s t r a c t

Article history: The objective of this paper is to provide benchmark test results, explanatory shell finite element models,
Received 3 August 2015 and preliminary Direct Strength Method prediction for cold-formed steel lipped channels undergoing
Received in revised form torsion dominated by warping response. Although the elastic theory for the torsional response of a thin-
16 October 2015
walled cold-formed steel lipped channel member is well-developed, the extent to which warping torsion
Accepted 26 October 2015
dominates the response of cold-formed steel members is not widely appreciated. Further, for cold-
formed steel members in torsion little exists in terms of experimental benchmarks and even less on
Keywords: situations beyond the classic elastic theory, including geometric nonlinearity and post-buckling, and/or
Cold-formed steel material nonlinearity from partial to full plastification of the section. Here, a typical cold-formed steel
Torsion
lipped channel member loaded experimentally in torsion exhibits significant strength beyond first yield.
Warping
Shell finite element models of the testing correlate well with the experiments and indicate the extent of
Direct Strength Method
Torsional buckling plastification as the thin-walled member undergoes torsion dominated by warping response. Idealized
end boundary conditions are developed for the shell finite element model that is conservative with
respect to the response, and in agreement with classical expressions in the elastic regime. The shell finite
element model with idealized end boundary conditions is used to develop a parametric study on ulti-
mate torsional capacity for members dominated by warping torsion. The results indicate that torsional
slenderness may be used to predict torsional capacity and indicate that Direct Strength Method pre-
dictions for torsion for members dominated by warping torsion are possible. Preliminary design ex-
pressions for warping torsion strength prediction are provided.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction In the classical theory [3,4] torsion (T) is resisted by shear (Tsv)
and by shear related to restrained warping (Tw):
Torsion is fundamental to the response of cold-formed steel
T = Tsv + Tw = GJθ′ − ECwθ‴ (1)
members since common sections are open, with relatively weak
torsional resistance, and often singly- or un-symmetric. For thin- where G is the shear modulus, J is the St. Venant torsional
walled cold-formed steel members torsion manifests itself in di- constant, E is the elastic modulus, Cw is the warping torsion con-
rect form or through instability. For example, cold-formed steel stant, and θ is the angle of twist (and ’ denotes derivatives). For
lipped channel beams (joists, purlins, girts, etc.) loaded away from thin-walled cold-formed steel sections, as discussed further in the
their shear center develop torsion. Torsion also plays a key role in next section, the shear contribution (Tsv) and related shear stresses
buckling instabilities at the member level with lateral-torsional (τt) are relatively small and the dominant resistance develops from
buckling of beams, and flexural-torsional buckling of columns; and warping restraint (e.g., see [6]). Warping restraint creates long-
at the cross-section level with (flange/web) distortional buckling. itudinal (sw) and shear stresses (τw) in the cross-section. The
The theory for the elastic torsional response of thin-walled open longitudinal stresses, which are the primary contributor to in-
sections was developed by Vlasov [1] and utilized by Timoshenko stability and degraded strength in thin-walled members, may be
[2] and others, and remains the primary tool for design prediction determined from:
methods (see [3]–,[5]).
σw = Eωθ″ = Bω/Cw (2)

n
Corresponding author. where ω is the sectorial coordinates, and B is the bimoment.
E-mail address: [email protected] (G. Bian). Warping stress sw may be found directly through differentiation of

http://dx.doi.org/10.1016/j.tws.2015.10.023
0263-8231/& 2015 Elsevier Ltd. All rights reserved.
566 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577

A A

T=1
T=1
A
A A
A

Fig. 1. Beam with mid-span torsion load: (a) torsionally supported – warping-free and (b) torsionally supported – warping-fixed.

the twist ([3], [4], and [5]), or through B, which develops in the torsion interaction, where R is the ratio of the maximum bending
section as it responds to torsion, T, and is available from numerical stress to the combined bending (sb) plus warping stress (sw), i.e.:
beam finite element solutions (e.g. MASTAN [7], [8]). The shear (σw )max
stresses due to warping vary around the cross-section according R= ≤ 1.0
σb + σw (5)
to:
ES wθ‴ The resulting reduction, which is applied to the bending ca-
τw = − pacity calculated without consideration of torsion, provides a re-
t (3)
duction similar to a longitudinal stress-based linear interaction
where Sw is the warping static moment, and t is the thickness of equation. The dominance of a stress-based approach to account for
the member. The shear stresses due to St. Venant torsion vary a limit state (torsion) is unusual in modern design where strength-
through the thickness, per: based limit states are used for all other actions.
τt = Gtθ′ (4) The approach taken in this paper is to explore basic aspects of
torsion in cold-formed steel members first from a review of clas-
Research on cold-formed steel beams in torsion demonstrated sical elastic response, then from a pilot set of experiments at a
the detrimental role of the torsional response on bending strength single length, and finally from companion shell finite element
and the importance of including this response in design [9]. models that extend into the nonlinear geometric response and
Analysis demonstrates that the torsional warping stresses change yielding. Torsional response in buckling, initial yielding, and full
significantly as the beam twists and are highly sensitive to the end plastification are all explored. Ultimately, the goal is to provide
boundary conditions [10]. Exploration of the stability of the sec- strength-based expressions for the prediction of torsional limit
tion further indicates that cross-section buckling is also sensitive states, when warping torsion dominates the response, that can be
to the longitudinal warping stresses that develop in the twisted integrated into design through appropriate interaction equations.
section [10]. For the common case of a cold-formed steel beam
with restraint on one flange the torsional stresses that develop are
even more complex, but their correct inclusion can aid design 2. Classical elastic torsional response in cold-formed steel
([11], [12], and [13]). Research provides significant insight on tor- members
sional response of cold-formed steel members, but less has been
done to examine torsion in isolation for cold-formed steel mem- While it is generally understood that thin-walled open mem-
bers – the approach that has long been used to understand axial, bers, such as those used in cold-formed steel applications, rely on
bending, and shear actions. warping to restrain torsion the extent to which this is true and the
In cold-formed steel design the basic philosophy is to try to conditions under which this is true are less well understood.
eliminate torsion to the greatest extent possible. For example, Distribution of torsion between Tw and Tsv is dependent on cross-
AISI-S100 [5] provides prescriptive bracing criteria to limit torsion section properties, boundary conditions, and the member length.
in beams. When torsion must be considered, design directly or To illustrate, consider the torsional response of a 400S162-54
indirectly applies stress-based interaction expressions to limit the [345 MPa (50 ksi)] cold-formed steel member (nomenclature per
impact of torsion. Eurocode [14] limits the total longitudinal stress AISI-S200 [15]). Two cases with midspan torsion applied, as shown
from all actions, including torsion, to be less than the yield stress, in Fig. 1 are considered: torsionally supported – warping-free (i.e.
Fy (divided by a partial safety factor). AISI-S100 [5] employs a re- the ideal fork type boundary conditions), and torsionally sup-
duction factor, R, on bending strength to account for bending- ported – warping-fixed. The members are modeled in MASTAN [8]

Fig. 2. Diagram of restrained torsion along member length for 400S162-54 for different member lengths (a) torsionally supported – warping-free and (b) torsionally
supported – warping-fixed.
 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577 567

Fig. 3. Diagram of restrained torsion along member length for different αː (a) torsionally supported – warping-free and (b) torsionally supported – warping-fixed.

and the results post-processed in terms of the warping torsion (Tw) lipped channel members in North America [15] are also provided
diagram and provided in Fig. 2 for a series of different L/H ratios, in Fig. 4. (For L/H of 2 and 20 the 400S162-54 of Fig. 2a and b has
where L is the member length and H is the cross section depth (i.e. an α of 48 and 0.48, respectively.) Fig. 4 indicates that for the vast
100mm (4 in.)). The L/H ratio is varied between 2, consistent with majority of cold-formed steel members and boundary conditions
a well-braced framing member, to 20 a large torsionally-unbraced Tw dominates the elastic response. Only for warping-free members
span. The response, provided in Fig. 2b, indicates that at the with long unbraced lengths (high L/H) does Tsv play a significant
midspan and at the fixed end of a warping-fixed beam Tw dom- role.
inates response. In a warping-fixed beam Tw is a minimum at the ¼ The stresses developed in a cold-formed steel cross-section due
points, but even for L/H as high as 20, Tw still dominates response to torsion include sw, τw, and τt, as introduced in Eqs. (2)–(4). The
(Fig. 2b). For a beam with warping continuity, but warping-free at distribution of each of these stresses for a typical cold-formed
its ends, Tw is a minimum at the ends, as shown in Fig. 2a. In this steel section is depicted in Fig. 5. The extent to which the stresses
case, Tw is still as much as 80% of the total torsion, but Tsv plays an of Fig. 5 are realized is dependent on how the torsion is distributed
important role. between Tw and Tsv, and the derivatives of the twist as related to
The selected member cross-section dimensions influence the bimoment in Eq. (2).
results and this may be captured by considering the non-dimen- The impact of the developed stresses may be understood in
sional variable α ¼ECw/(GJL2) as a means to classify the section. For terms of the von Mises stress, which is commonly associated with
our typical 400S162-54 cold-formed steel section with L/H ¼6, prediction of steel yielding in multi-axial stress states, and in this
α ¼5.34. Fig. 3 shows the warping torsion diagram response for α case simplifies to:
from 1 to as small as 0.01. Here one can see that in sections that
σeff = σw2 + 3(τw2 + τt 2) (6)
have relatively large J in comparison with Cw the dominance of
warping torsion response is decreased. The maximum warping stress (sw) in the cross section is
Systematic study of the ratio of warping torsion (Tw) to the total compared to the von Mises stress (seff) along the member for the
torsion (Tw þ Tsv) across a broad range of cross-sections (α‘s) at 400S162-54 at L/H ¼6 for the studied torsion cases in Fig. 6. For
three key locations: midspan in the warping-fixed case, ¼ span in warping-fixed end conditions, sw is a reasonable estimate of seff
the warping-fixed case, and at the member ends in the warping- except near the ¼ point, where stress is low. Even for warping-free
free case are provided in Fig. 4. Histograms of the α at L/H of 2 and end conditions, only at the member ends, where seff is attributable
20 for all commercially available structural cold-formed steel to τw and τt alone is their significant error – generally seff is

Fig. 4. Ratio of restrained torsion to total torsion for warping-free and warping-fixed boundary conditions across different member properties.
568 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577

e e e
a a
c d a
d
b b d c b
c
B = 3.44N·m2 Tw = 5.56N·m Tsv = 0.13N·m
σw τw τt
(MPa) (MPa) (MPa)
-1.00 0.00 1.00
-0.52 -0.67 1.00
-0.00 -1.00 1.00
0.47 -0.77 1.00
0.00 0.22 1.00
Fig. 5. Stress analysis of 400S162-54 under torsion (a) warping longitudinal stress (b) warping shear stress (c) pure shear stress (note: B¼ 3.44N m2 is equivalent to a mid-
span T of 1.15 N m in the 400S162-54 with L of 0.61 m).

dominated by sw and specifically this holds true at maximum sections with welded end plates and a circular shaft inserted into
stress locations. the grips as shown in Fig. 7. Specimens were held constant in the
For cold-formed steel sections warping torsion is more likely to axial direction and twisted to approximately 45 degrees with ac-
dominate response than St. Venant torsion. However, the elastic tuator torsional moment and angle of twist recorded. The bottom
stresses developed in the classical theory are based on the un- grip, per Fig. 7, applied the twist to the specimen while the top
deformed cross-section, and provide no insight on what happens remained fixed.
after yielding.
3.1. Test specimen and loading protocol

3. Pilot experiments in torsion The specimens were cut from a single 6.7 m long punched
400S162-54 [50 ksi] stud (nomenclature per AISI-S200 [15]). This
A small experimental pilot study was conducted to explore cross section was chosen based upon material availability, but
torsion on the behavior of cold-formed steel lipped channels represents a common depth, 102 mm (4 in.), thickness, 1.37 mm
dominated by warping torsion, and loaded past initial yield. The (0.054 in.), and grade 345 MPa (50 ksi), for load-bearing wall studs
tests employed an MTS tension-torsion rig with hydraulic grips. in cold-formed steel light frame construction. A tensile coupon
The specimens consisted of cold-formed steel lipped channel was taken from the web of the stud and testing conducted per

Fig. 6. Cross-sectional maximum stress distribution along the member under torsion: (a) warping-fixed boundary condition and (b) warping-free boundary condition.
 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577 569

centroid. Based on as-built measurements, error in this alignment

Loading Machine was within 2.5% of ideal.
Specimens were monotonically loaded to a twist of 45° at a rate
(Fix part) of 0.025° per second. Once a 45-degree twist was achieved, the
specimen was unloaded at a rate of 0.75° per second. Basic test
parameters are summarized in Table 1. Tests b–d, identical save for
the load rate in test c, establish a core set of results for torsional
response of a section with approximately rigid end conditions.

3.2. Experimental results

End
Results for the experimental testing are shown in Fig. 10, and
plate the moment at maximum torsion and its corresponding rotation
are provided in Table 1. All of the specimens are able to undergo
large twist rotations (at least  30°) prior to reaching their peak
Specimen

torsional moment. The specimens with thicker (25.4 mm [1 in.])

end plates (b–d) had stiffer response than that with thinner
(6.4mm [0.25 in.]) end plates. Nominally identical specimens (b–d)
share a common elastic stiffness, similar rotations at failure, but
modestly different peak torsional moments. Observed failure
modes for representative specimens are shown in Fig. 11. The
End specimen with a thin end plate (6.4 mm [0.25 in.]) experienced
significant end plate deformations as a result of warping de-
plate formations generated by the specimen in torsion (Fig. 11(a)). The
warping boundary condition for the thin end plate is semi-rigid.
Specimens with 25.4 mm (1 in.) end plates experienced inelastic
lip local buckling past yield (Fig. 11(b) and ultimately, failed in the
Grip Shaft weld between the channel and the end plate (Fig. 11(c)). This is
considered indicative of a warping-fixed end boundary condition.
The tested specimens were all of a relatively short length and
Loading Ma
Machine dominated by warping response.
(Rotating p
part)
4. Shell finite element model of experiments
Fig. 7. Torsion testing of cold-formed steel lipped channel: untwisted specimen in
rig. Shell finite element models motivated from the tests were
developed and conducted in ABAQUS [19] to further evaluate the
test results and provide an estimation of the internal stresses
developed under torsion. Specifically, this section provides the
basic model characteristics, comparison with the conducted ex-
periments, and the spread of plasticity through the section under
twist.

4.1. Model characteristics

To model the cold-formed steel stud the ABAQUS S4R shell

element was used. When applicable the end plate and shaft were
modeled with the C3D8R solid element. Typical finite element
meshes are shown in Fig. 15b.
End boundary conditions play a particularly important role in
the response, and three distinct models are explored in this re-
gard: (i) test model, essentially true to the testing apparatus; (ii)
fully fixed model, providing an idealized upper-bound response;
and (iii) idealized warping-fixed model, creating conditions con-
sistent with classical warping torsion theory. As shown in Fig. 12,
in the test model, Fig. 12(a), both the end plate and loading shaft
Fig. 8. Experimental tensile stress–strain relationship for coupon cut from are added to the stud. For the fully fixed model, Fig. 12(b), one end
400S162-54 [50 ksi].
of the stud is coupled to a reference point with all six degrees of
freedom fixed, while the other end of the stud is coupled to a
ASTM [16] with the result shown in Fig. 8, the measured yield reference point with five degrees of freedom fixed (only the tor-
stress is 373 MPa (54 ksi) per the 2% offset method. sion/twist degree of freedom where the twist action is applied is
As detailed in Fig. 9 the typical 304.8 mm (12 in.) long speci- free). For the idealized warping-fixed model, as shown in Fig. 12
mens, had steel end plates, 6.4 mm (0.25 in.) or 25.4 mm (1 in.) (c), the ends of the stud are tied to a reference point at the cen-
thick, welded to the ends. In addition, to connect to the grips of troid, but only torsional moment is allowed to develop all other
the torsion rig, 25.4 mm (1 in.) diameter steel shafts were welded forces/reaction are released (this is accomplished by allowing
to the end plates. The shafts were aligned with the cross-section longitudinal deformations associated with axial deformation and
570 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577

Fig. 9. Typical test specimen, plan view showing (a) basic designation, and (b) size and location of end plates and (c) elevation view of stud, end plate, and shaft.

Table 1
Test details and basic response for torsion tests.

IDa Length End plate Τu θ@Tu

(mm) (mm) (kN m) (deg)

a 304.8 6.4 0.634 45.0b
b 304.8 25.4 0.62 34.0
cc 304.8 25.4 0.64 31.6
d 304.8 25.4 0.70 29.2

a
400S162-54 [50 ksi] stud, nomenclature per [16,18].
b
Maximum rotation tested ¼ 45°.
c
Loaded at faster rate: see [17] for detail.

d
c
b

a
Lip
Buckling
Fracture

Fig. 10. Load-displacement of test results.

major- and minor-axis movements, but not torsion warping de-

formations). The idealized warping-fixed model is the subject of
Section 5, while the test and fully fixed model are compared with Fig. 11. Observed deformations and failure (a) test a, significant distortions in end
plates, (b) test b, inelastic lip local buckling, (c) test c, inelastic lip local buckling
the experiments in this section.
followed by fracture along the opposing weld.
Material nonlinearity was accounted for using classical metal
plasticity theory based on von Mises yield criterion and isotropic
 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577 571

Ux=Uy=Uz=0 z
Rx=Ry=0 Fz
θ
Mx, Fx My, Fy
θ
Ux=Uy=Uz=0

x y
Rx=Ry=0
θ

Ux=Uy=Uz=0
Rx=Ry=Rz=0 Ux=Uy=Uz=0
T T Rx=Ry=Rz=0

Ux=Uy=Uz=0
Rx=Ry=Rz=0
T
Fig. 12. Typical FE model and depiction of end boundary conditions for: (a) tested stud; (b) fully fixed; (c) idealized warping-fixed using multi-point coupling to the centroid
and releasing all actions except torsion.

hardening. For direct comparison with the tests, the measured, impact/sensitivity to errors when welding the loading shaft to the
engineering stress and strain values as given in Fig. 8 were em- end plate a þ/  6.4 mm offset was considered. To vary the stiff-
ployed. For implementation in ABAQUS the engineering values ness of the loading shaft its diameter was varied from 12.7 mm (1/
were converted to true stress and strain. 2 the tested diameter, or 1/16th the actual J) to 63.5 mm (2.5 times
Imperfections and residual stresses were not considered in the the tested diameter, or 39 times the actual J).
models. Initial geometric imperfections play a modestly reduced
role in the response of the tested specimens due to the large ro- 4.2. Comparison with experiments
tations ( 410°) that exist prior to local buckling initiation. Residual
stresses, both from forming and welding, deserve future study. The The torsion-twist response of the developed finite element
model does not consider fracture. models are compared with the tests with the thick end plate
Two variations on the tested specimens were examined to (25.4 mm, Test a) and the thin end plate (6.4 mm Test c) in Fig. 13.
understand model sensitivity: (i) offset in the loading shaft, and The “fully-fixed” model is unrealistically stiff and strong indicating
(ii) torsional stiffness of the loading shaft. To understand the the end plate and loading shaft (though stiff) contribute

25.4 mm endplate 6.4 mm endplate

Fig. 13. Torsional moment-angle relationship for:(a) 25.4 mm endplate model (b) 6.4 mm endplate model.
572 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577

D= 63.5 mm
+ 6.4 mm offset
D= 31.8 mm
- 6.4 mm offset
D= 25.4 mm
test

D= 12.7 mm

Fig. 14. Loading shaft effect on FE result: (a) offset effect; (b) diameter effect.

meaningfully to the as-measured response. The “test model” with yielding in the modeled section under twist to determine the
the end plate and loading shaft explicitly modeled gives results limits of these common approaches. Consider the torque at which
that are most comparable to the testing prior to fracture (o30° for first yield occurs: in the developed shell finite element model one
the thick end plate, and o 40° for the thin end plate). can determine the first integration point in an element in the stud
The impact of shaft offset and torsional rigidity is provided in in which the von Mises stress max(seff)¼Fy. The shell finite ele-
Fig. 14 (a) and (b) respectively in comparison with Test 6. Including ment model includes the full three-dimensional state of stress
a small shaft offset increases the accuracy of the initial stiffness resulting from warping, shear, and even local plate bending and
prediction, and provides a small measure of scatter in the torsional end effects developed in response to torsion. In the model first
moment. In effect the offset is a useful, small imperfection in the yield occurs at the end cross section due to warping restraint at
model. The studied shaft diameters provide a variation in shaft the ends.
torsional rigidity (J) from 0.0625 to 39 times the actual shaft J. The To examine yielding, consider the torsion-plastic strain re-
response thus ranges from nearly fully dominated by shaft twist to sponse for the “test model” and “fully fixed” model as provided in
the fully fixed case, as shown in Fig. 14(b). One must take care with Fig. 15. For comparison, max(|sw|) ¼Fy ¼ 373 MPa at T¼ Tw ¼
using the shaft diameter as a simple end torsional rigidity proxy 0.2 kN m by classical theory. The end conditions have a clear im-
since the model uses solid elements for the shaft and thus influ- pact on the observed torsional moment and plastic strain. How-
ence the end plate rigidity. ever, for all cases the torsional moment, even for a small limiting
peak plastic strain of e.g. 1%, is nearly double the torsional moment
4.3. Characterization of cross-section yielding and plasticity at which yielding initiates in the model. Significant torsional in-
elastic reserve is observed. Fig. 15(b) shows a prediction of the
The torsion-twist response is initially elastic, but large de- developed plastic strain and deformed shape, deep into the in-
formations and yielding lead to a softening in the response. It is elastic reserve in torsion.
common in design to use elastic stress predictions and first yield To more fully explore the inelastic reserve, consider yielding in
criterion for torsion, here we explore the extent of observed the cross-section as opposed to just the peak plastic strain. First,

0.06
FE model 6.4 mm endplate
FE model 25.4 mm endplate
FE model fully fixed
0.05 T by hand calculation
wy
Maximum plastic strain

0.04

0.03

0.02
εplastic > 0
0.01 εplastic = 0

0
0 0.2 0.4 0.6 0.8 1
Torsional moment (KN-m)
Fig. 15. Plastic demands in models (a) peak plastic strain – torsional moment relationship; (b) plastic strain locations in deformed geometry of “test model” with 25.4 mm
end plate.
 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577 573

Fig. 16. Plastic cross section ratio vs. (a) rotation and (b) torsional moment for “test model” with 25.4 mm end plate.

consider a simple indicator function for yielding in element j: of 10°. The deformed shape and plastic strain for the model pro-
vided in Fig. 15(b) is shown as  in Fig. 16(a) and (b).
⎧ εp > 0
⎪ 1
Ij = ⎨
⎪
⎩ 0 εp = 0 (7)
5. Idealized warping-fixed shell finite element model for
where, εp is the effective plastic strain at mid-thickness of the
torsion
shell element. Thus, the yielding must be through the thickness
not just on the surface. This partially mitigates the influence of The testing model required the addition of end plates and a
local plate bending on the measure. Plastification of a cross-section loading shaft to the specimens. The result, as depicted in Fig. 12(a),
is then defined by: is that secondary bending moments and shears can develop as the
N
∑ j=1 I j member is twisted. In the “fully fixed” limit, Fig. 12(b), the warping
P= fixity is complete, and the secondary bending moments and shears
N (8)
are maximized. In this case the secondary stresses are large, and
where, Ij, is the plasticity indicator for element j in a cross even for small deflection, the elastic stress response is in poor
section; and N is the number of elements around the cross section. agreement with classical Vlasov torsion unless all secondary
The plastic cross section ratio, P, is 0 when the cross section is stresses are explicitly included. The objective of this work is to
elastic and 1 when the cross section is fully plastic. Fig. 16 shows investigate torsion in isolation, so a third “idealized warping-
predicted P for the end and middle cross-sections as a function of fixed” model, Fig. 12(c), is pursued in this section and used for
(a) rotation, and (b) torsional moment. Plasticity initiates at the subsequent parametric studies. This model allows bending to oc-
ends and continues to be greatest at the ends throughout the cur, but restricts warping, by restraining the member ends to re-
twist. Significant twist is required to approach full plastification of main in a rigid plane, but allowing that plane to twist (thus re-
the section; however 75% of the end section is plastified by a twist leasing the end moments, except for torsion).

174.4 MPa 185.6 MPa

191.7 MPa 178.1 MPa
193.1 MPa 257.9 MPa 205.0 MPa 249.4 MPa

373.0 MPa 373.0 MPa

Ty = 0.20 KN-m Ty = 0.21 KN-m

Fy = 373.0 MPa Fy = 373.0 MPa

373.0 MPa 373.0 MPa

193.1 MPa 205.0 MPa

257.9 MPa 249.4 MPa
174.4 MPa 185.6 MPa
191.7 MPa 178.1 MPa

Fig. 17. Longitudinal stress distribution at yielding torque (Ty ¼ 0.2 kN m) from (a) classical Vlasov theory and (b) idealized warping-fixed shell FE model at the end cross-
section.
574 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577

The elastic longitudinal stresses (sz) that develop in the idea- finite element model, or approximated with other methods as
lized warping-fixed shell finite element model are compared with discussed later in this section. For the 304.8 mm long 400S162-54
the classical Vlasov warping theory in Fig. 17. The sz from the shell at Fy ¼ 373 MPa, Ty ¼ 0.21 kN m as reported in Fig. 17, and Tcr for
finite element model are extrapolated to the nodes at the mid- selected modes of the shell finite element model are shown in
thickness of the shell elements. The sw for the classical theory are Fig. 19.
determined via Eq. (2) which is implemented by finding the bi-
moment at the end from a beam finite element model [7,8] and 6.1. Parametric study
numerically determined ω and Cw [20]. The basic stress distribu-
tion is similar in both models, and the peak stress location is the To explore the relationship between torsional slenderness, λT
same in both models, but the stresses in the shell finite element and strength, a small parametric study is conducted. In the study
model are generally slightly lower (reflecting the additional flex- λT is varied from 0.25 to 2.5 for four different physical lengths all
ibility of the non-rigid cross-section) resulting in a slightly higher completed with the 400S162-54 cross-section, by varying the yield
prediction of the torque at which first yield occurs. stress as shown in Table 2. The basic torsion-twist results, for the
The torsion-twist response for the idealized warping-fixed studied sections are shown in Fig. 20. By varying the physical
model of the 400S162-54 with a length of 304.8 mm is compared length the elastic stiffness is varied and by varying the yield stress
with the fully fixed model and test results in Fig. 18(a). The idea- the ultimate torsional strength is varied. The end result is that a
lized warping-fixed model provides the most conservative esti- wide variety of strength responses from elastic, to nearly fully
mation of the response: Tu ¼ 0.48 kN m and θ ¼ 5.4° at Tu. The de- plastic, are realized in the studied models.
formation and developed plasticity at Tu are shown in Fig. 18(b). For each of the studied shell finite element models Ty and Tcr
Based on these results, it is included that the idealized warping- are determined. Ty is the applied torque at which the first yielded
fixed model is most appropriate for exploring the strength in element occurs in the shell finite element model and Tcr is the first
(isolated) torsion. positive buckling moment from eigen-buckling analysis of the
shell finite element model. The observed peak strength (Tu) is
examined as a function of torsional slenderenss in Fig. 21 for the
6. Development of direct strength method for torsion studied sections. The results indicate that significant torsional in-
elastic reserve is common, but lipped channels which are globally
The Direct Strength Method of cold-formed steel member de- slender in torsion have only limited elastic post-buckling.
sign [5,21] has shown that slenderness, with due consideration of
both member and cross-section buckling modes, may be used to 6.2. Design expressions
predict member strength in axial, bending, and shear actions. Si-
milar relationships do not exist for torsion, and are investigated As provided in Fig. 21, two possible design expressions are
herein. The idealized warping-fixed shell finite element model is postulated: a two-part expression (Prediction I) that is quadratic in
used to provide “exact” predictions of ultimate torsional capacity. the inelastic reserve regime and asmytpotes to the elastic bucking
Torsional slenderness is then used to develop “approximate” Di- solution for large global slenderness regime; and a two-part ex-
rect Strength predictions appropriate for use in design. pression (Prediction II) that is linear throughout the inelastic re-
Torsional slenderness is defined as: serve regime and then asymptotes to the elastic buckling solution.
The expressions are as follows:
λT = Ty/Tcr (9) Prediction I: two-part with quadratic inelastic reserve
where Ty is the torque at first yield and Tcr is the critical elastic For λT ≤ 1
torsional buckling moment, considering plate buckling. Ty is most Tn/Ty = 2 − λT 2 (10)
rigorously defined as the T when seff ¼ Fy; however, for cross-
sections dominated by warping torsion response Ty may be ap- For λT > 1
proximated by Twy, i.e. the T when max(|sw|)¼ Fy. The torsional
Tn/Ty = 1/λT 2 (11)
elastic buckling moment Tcr may be determined from the shell

εplastic > 0
εplastic = 0

Fig. 18. (a) Torsional moment-angle curve for fully fixed model and idealized warping-fixed model; (b) Idealized warping-fixed FE model in ABAQUS.
 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577 575

1st mode 2nd mode 3rd mode 4th mode 5th mode
Tcr/Ty= 7.47 Tcr/Ty = 7.76 Tcr/Ty = 9.08 Tcr/Ty = 10.45 Tcr/Ty =10.53
Fig. 19. Buckling modes and critical torque for idealized warping-fixed shell finite element model.

Table 2 3
Summary of parametric variation for 400S162-54 section. Abaqus
Prediction I
α¼ Ty 2.5 Prediction II
DSM local buckling
L Tcr ECw/(GJL2)a Min Max
DSM dist. buckling
2
(mm) (kN m) (kN m) (kN m)
152 4.92 21.4 0.31 30.7
Tu/Ty

229 2.92 9.5 0.18 18.2 1.5

305 1.90 5.3 0.12 11.9
457 1.26 2.4 0.08 7.9
1
a
consistent with Fig. 4, for α, L is defined here as twice the model length.

0.5
7
L= 152.4 mm
L= 228.6 mm 0
6 L= 304.8 mm 0 0.5 1 1.5 2 2.5
L= 457.2 mm 0.5
(T /T )
Torsional moment (KN-m)

y cr
5
Fig. 21. Direct strength prediction curve for members under torsion, exact solution
for Ty and Tcr.
4
The test-to-predicted ratio (P) for Prediction I is 1.19 with a
3 coefficient of variation (VP) of 0.19. Based on the reliability pro-
cedures outlined in Chapter F of AISI-S100 [5] for a reliability index
2 β ¼2.5 this implies an LRFD resistance factor ϕ ¼0.84 or and ASD
safety factor of Ω ¼1.9. Prediction II has a P of 1.13 with a VP of 0.21
and therefore a ϕ ¼ 0.83 or Ω ¼1.9. For the studied sections either
1
method is acceptable in design. Prediction II is consistent with the
simplified method adopted for inelastic reserve in beams in AISI-
0 S100 [5]; however Prediction I has a smaller COV and places a
0 5 10 15 20 25 30 35
more conservative limit on the maximum torque, and is generally
Angle (degrees) recommended.
Fig. 20. Torsional moment-angle curve for idealized models with different tor-
sional slenderness and strengths. 6.3. Implementation in design

Prediction II: two-part design expression with linear elastic To implement the prediction method the engineer must be able
reserve to quickly provide Ty and Tcr. Although approximate, it may be
For λT ≤ 1 sufficiently accurate to determine classic Vlasov warping stresses
due to torsion (sw) and set Ty as the torque at which max(|sw|)¼Fy.
Tn/Ty = 2.5 − 1.5λT (12) The warping stresses may be found using classical methods [3,4,5]
or numerical methods [7,8,20]. If a shell finite element model is
For λT > 1 available it may be used for determining Tcr, in the absence of such
a model we have investigated the use of the computationally ef-
Tn/Ty = 1/λT 2 (13) ficient semi-analytical finite strip method (FSM) as implemented
576 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577

3
Abaqus
Prediction I
2.5 Prediction II
DSM local buckling
DSM dist. buckling
2

Tu/Ty
1.5

0.5

0
0 0.5 1 1.5 2 2.5
0.5
(Ty/Tcr)

Fig. 23. Direct strength prediction curve for members under torsion employing
Fig. 22. Signature curve in CUFSM for member under torsion. approximate solution for Ty and Tcr to evaluate the slenderness.

in CUFSM [22]. end conditions is inconsistent and may be presumed as a coarse

For the 400S162-54 the warping stresses that cause first yield design approximation at best. The last 4 columns of Table 3 provide
in the section are generated and applied as the reference stress for the FSM solutions for clamped, warping-fixed, end conditions. The
an elastic buckling analysis. The signature curve results are shown model still assumes constant longitudinal stress, so even this re-
in Fig. 22. Three buckling modes are postulated by the signature mains conservative, and Tcr is further elevated by these conditions.
curve analysis: lip local buckling, flange/web local buckling, and The last assumption, ignoring the shear stress, is unconservative
distortional buckling. The first two modes are observed in the shell and recent work provides a means to include this aspect in the
finite element model, but at different buckling torques. These two context of the semi-analytical FSM [23].
buckling modes are compared to the shell finite element solutions If Ty is defined by Vlasov’s warping stress, and Tcr by the lip
of a 304.8 mm long model with idealized warping-fixed end local buckling mode from a signature curve finite strip analysis
boundary conditions in Table 3. The FSM model presumes constant then the parametric study results may be revisited. For these ap-
longitudinal stress, while in reality the longitudinal warping stress proximate solutions the strength as a function of slenderness is
varies linearly along the length of the member with maximum slightly modified, as shown in Fig. 23 and the statistics for Pre-
values of opposing sign at the member ends. As a result of this diction I and II are also modified. Prediction I has P ¼1.26, VP ¼ 0.22,
variation the distortional buckling mode identified in the signature and ϕ ¼0.84 for β ¼1.9, while Prediction II has P ¼1.26, VP ¼ 0.22,
curve is not relevant since it has a long buckling length. and ϕ ¼0.84 for β ¼1.9. The approximate methods for Ty and Tcr
The implementation employed for the semi-analytical finite determination are adequate for design.
strip method [21] contains three serious simplifications for use in
this application: (1) constant longitudinal stress (2) warping-free 6.4. Limitations and future work
end conditions, and (3) no influence from shear stress. Assuming
constant longitudinal stress is conservative, and one must consider The work presented herein is a limited investigation of torsion in
the length of the stress gradient vs. the length of the buckling mode cold-formed steel members. Significant work remains to develop a
to determine the severity of this simplification. Typically local robust means of handling torsion. Although thin-walled members
buckling is unaffected by stress gradients – and this is largely true are dominated by warping torsion, additional cases (e.g. due to
even in this case; however, other buckling modes generally require thickness, end boundary conditions, etc.) where the contribution of
some consideration of stress gradient. Assuming warping-free end St. Venant torsion is non-negligible need further study. The poten-
conditions when the applied stresses are based on warping-fixed tial longitudinal variation of warping stresses (i.e. variation in

Table 3
Comparison of elastic buckling solutions.

Shell FEa Finite strip methodb (constant longitudinal stress)

Warping-free ends Warping-fixed ends

L¼ 305 mm Signature curve 152 mm 229 mm 305 mm 457 mm

Tcr/Ty Lcr Tcr/Ty Tcr/Ty Tcr/Ty Tcr/Ty Tcr/Ty
(mm)

Lip Local Buckling 7.47 28 7.21 7.38 7.27 7.24 7.22

Flange/Web Local Buckling 7.76 56 5.90 6.63 6.20 6.06 5.97
Distortional Buckling 4 10.53c 307 3.26 19.41 10.46 7.17 5.15

Ty, classic torsion Fy ¼373 MPa.

a
Idealized warping-fixed” model.
b
CUFSMv4.
c
Not in first 5 modes.
 G. Bian et al. / Thin-Walled Structures 98 (2016) 565–577 577

bimoment) is significant in structural applications, e.g. a floor joist (AISI), ClarkDietrich, Steel Stud Manufacturers Association, and the
loaded away from its shear center has a much different longitudinal Steel Framing Industry Alliance, The views expressed in this work
warping stress distribution than the twisted member studied here. are those of the authors and not NSF, AISI, or any of the partici-
Systematic study of the impact of this variation (i.e., stress gradient) pating companies or advisors. Testing was performed at the
is needed to understand the impact on buckling modes and on Structural Engineering and Earthquake Simulation Laboratory
yielding. Simplified methods are needed for predicting torque un- (SEESL) at the University at Buffalo. The authors would like to
der partial or full plastification. The lack of a simplified calculation thank the SEESL staff for their expertize and cooperation.
for plastic torque in thin-walled members leads to the type of gross
simplifications provided in the currently developed prediction
methods. Focused tests and models varying torsional slenderness in
each of local, distortional, and global buckling are needed to fully References
understand the complete torsional strength. Torsion in combined
loading should be revisited with the goal of investigating limit- [1] V.Z. Vlasov. Thin-walled Elastic Beams. Jerusalem Israel Program for Scien-
tific Translations, 1961.
states based strength interaction equations to replace the stress- [2] S.P. Timoshenko, J.M. Gere, Theory of Elastic Stability, McGraw-Hill-Kogakusha
based expressions in current use in design. Further testing, mod- Ltd, Tokyo, 1961.
eling, and analytical developments are all needed. [3] W.W. Yu, R.A. LaBoube, Cold-Formed Steel Design, John Wiley & Sons, Inc, New
York, 2000.
[4] AISC-Design Guide 9. Torsional Analysis of Structural Steel Members. Chicago,
IL: American Institute of Steel Construction, 2003.
7. Conclusions [5] AISI-S100-12, North American Specification for the Design of Cold-Formed
Steel Structural Members, American Iron and Steel Institute, Washington (DC),
2012.
Torsional response is fundamental to understanding thin-walled [6] B.W. Schafer, R. Madsen, AISI Torsion Provisions and Torsional Analysis of
cold-formed steel members; but beyond classical work on elastic Cold-formed Steel Members. Structures Congress, Massachusetts, Boston,
response and prescriptive guidance little is available to engineers March 2014.
[7] W. McGuire, R. Gallagher, R. Ziemian, Matrix Structural Analysis, John Wiley &
that must consider either torsion or buckling modes undergoing
Sons, Inc, New York, 2000.
torsion. Classical analysis of cold-formed steel members indicate [8] MASTAN2 v3.3. Tutorial for MASTAN2. 〈www.mastan2.com〉 .
that they often have high values of the non-dimensional constant [9] B.M. Put, Y.L. Pi, N.S. Trahair, Bending and torsion of cold-formed channel
ECw/(GJL2) and may be often dominated by warping response. A beams, J. Struct. Eng. 125 (5) (1999) 540–546.
[10] B.P. Gotluru, B.W. Schafer, T. Peköz, Torsion in thin-walled cold-formed steel
small series of torsion tests on a cold-formed steel lipped channel beams, Thin-Walled Struct. 37 (2) (2000) 127–145.
demonstrates that cold-formed steel members can: undergo large [11] Z. Ye, R. Kettle, L. Li, Analysis of cold-formed Zed-Purlins partially restrained by
twist rotations prior to failure, exhibit significant post-yield (in- steel sheeting, Comput. Struct. 82 (9-10) (2004) 731–739.
[12] L. Vieira, M. Malite, B.W. Schafer, Simplified models for cross-section stress
elastic reserve) strength, and are sensitive to end conditions.
demands on C-section Purlins in uplift, Thin-Walled Struct. 48 (1) (2010)
Complementary shell finite element models demonstrate the large 33–41.
extent of plastification in a common cold-formed steel member [13] T. Gao, C. Moen, Extending the direct strength method for cold-formed steel
undergoing torsion and provide further insight on the sensitivity to design to through-fastened simple span girts and Purlins with laterally un-
braced compression flanges, J. Struct. Eng. vol. 140 (2014) 6.
end boundary conditions. The end boundary conditions in the tor- [14] EN 1993-1-3. Eurocode 3: Design of steel structures —Part 1-3: General Rules
sion testing allow secondary shears and moments to develop at the Supplementary rules for cold-formed members and sheeting. European
fixed member ends, therefore an alternative idealized warping- Standard, CEN, 2004.
[15] AISI-S200-12, North American Standard for Cold-Formed Steel Framing -
fixed boundary condition is developed in the shell finite element
General Provisions, American Iron and Steel Institute, Washington (DC), 2012.
models. This idealized warping-fixed model generates longitudinal [16] ASTM E8/E8M-13. Standard Test Methods for Tension Testing of Metallic
warping stresses consistent with classic Vlasov torsion theory and is Materials, 2013.
utilized to develop a small parametric study on torsional strength. [17] K.D. Peterman, G. Bian, B.W. Schafer, Experimental and computational analysis
of direct torsion in cold-formed steel lipped channels, in: Proceedings of
For sections dominated by warping torsion, the study demonstrates Structural Stability Research Council Annual Stability Conference. Toronto,
that torsional slenderness may be used to predict ultimate torsional Canada, March 2014.
strength, in a manner consistent with Direct Strength Method de- [18] Steel Stud Manufacturers Association. Product Technical Information. 〈http://
www.ssma.com〉 .
sign expressions previously developed for axial, shear, and bending
[19] ABAQUS. ABAQUS/CAE User’s Manual Version 6.12. Simulia, 2012.
actions. Simplified methods for calculating torsional slenderness are [20] Li Z., Schafer, B.W. Buckling analysis of cold-formed steel members with
explored and shown to be adequate for design. Significant addi- general boundary conditions using CUFSM: conventional and constrained fi-
tional work remains to generalize the results and develop a fully nite strip methods. In: Proceedings of the 22nd international specialty con-
ference on cold-formed steel structures, St. Louis, MO. November, 2010, pp.
limit-states based approach to torsional strength in cold-formed 17–32.
steel members, but the completed work is intended to provide [21] B.W. Schafer, Review: the direct strength method of cold-formed steel
proof that a clear path is possible and available. member design, J. Constr. Steel Res. 64 (7) (2008) 766–778.
[22] Schafer B.W., Ádány S. Buckling Analysis of Cold-Formed Steel Members Using
CUFSM: Conventional and Constrained Finite Strip Methods. In: Proceedings
of the 18th International Specialty Conference on Cold-Formed Steel Struc-
Acknowledgments tures. St. Louis, MO. Octomber, 2006.
[23] G.J. Hancock, C.H. Pham, Development in the finite strip buckling analysis of
plates and channel Section under Localized Loading. In: Proceedings of the
The authors would like to thank the National Science Founda- 22nd international specialty conference on cold-formed steel structures. St.
tion (NSF-CMMI #1041578), American Iron and Steel Institute Louis, MO, November, 2014. pp. 295–309.