Vous êtes sur la page 1sur 22

Journal of Constructional Steel Research 55 (2000) 267288


Recent research advances in cold-formed steel

J.M. Davies
Manchester School of Engineering, University of Manchester, Oxford Road, Manchester, M13 9PL,


Recent developments in the practical utilisation of cold-formed sections in building construc-

tion have taken place on three related fronts. There have been significant developments in the
technology which result in more complex shapes with a higher yield stress so that cold-formed
sections represent a particularly high-tech form of constructional steelwork. Developments in
technology would be of little consequence unless there were parallel developments in practical
applications and this is illustrated by the continual increase in the market share of cold-formed
sections. This, in turn, makes demands on design procedures and requires parallel development
in calculation models and design codes. In particular, sections have tended to become more
highly stiffened and this necessitates a more sophisticated treatment of local buckling, distor-
tional buckling and the interactions between them. The latest trend is to move from simplified
design models to design procedures based on whole section analysis. In this paper, recent
developments in technology and application are outlined and this is followed by more detailed
consideration of the related design procedures. 2000 Elsevier Science Ltd. All rights

Keywords: Beams; Buckling; Cold-formed; Columns; Distortion; Restrained; Steel; Thin-walled

1. Introduction

The market share of cold-formed structural steelwork continues to increase in the

developed world. The reasons for this include the improving technology of manufac-
ture and corrosion protection which leads, in turn, to the increased competitiveness
of the resulting product as well as new applications. This paper first reviews the

* Fax: +44-(0)161-9045929.
E-mail address: jmdavies@fs1.eng.man.ac.uk (J.M. Davies).

0143-974X/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 9 7 4 X ( 9 9 ) 0 0 0 8 9 - 9
268 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

technical developments and some of the resulting new applications. It is shown that
as a result of current trends, cold-formed sections are becoming more slender and
more highly stiffened and therefore more prone to local and distortional buckling.
It then follows that there is a consequential demand for improved design procedures.
These generally take the form of calculation models suitable for incorporation in
design codes. The available approaches to the design of cold formed sections are
reviewed in some detail and some possible improvements in design codes are antici-

2. Developments in cold-formed section technology

As a high-tech product, a cold-formed steel section is more open to development

than its more mundane hot-rolled counterpart. This section of the paper reviews some
recent developments. Some of these developments have previously been summarised
by Pekoz [1].

2.1. Trend to higher quality steels

The trend to higher quality steel is primarily seen in an increased yield stress and
there has been a continuing increase in the yield stress of typical mass-produced
products such as purlins, sheeting and decking in recent years. The steel used for
such applications now typically has a yield stress in the range 280 to 550 N/mm2.
This trend of increasing yield stress is likely to continue. There are also some steels
available with ultra-high yield stress. These are generally only economic for special
applications such as lightweight bridges for the military. In this context, Hoglund
[2] discusses the welding of quenched and tempered (QT) steel produced in Sweden
with a yield strength up to 1100 N/mm2. The application of steels with a very high
yield stress is often limited by considerations of stiffness and Nippon Steel in Japan
are reputed to have produced steels with a 30% increase of Youngs modulus in the
direction of rolling though with a corresponding decrease in the transverse direction.
Steels with better fire resistance, including less creep at elevated temperature, are
also being introduced.

2.2. Trend to more complex section shapes

The use of higher strength steels is inevitable accompanied by a reduction in

thickness and considerations of local stability lead logically to the development of
highly stiffened sections with more folds and rolled in stiffeners. This trend can be
conveniently illustrated by the development in sections for purlins. Early purlin
shapes were simple lipped channels and Zed sections. As shown in Fig. 1, the simple
channel has evolved into Multibeam Marks I, II and III, the latter having a compound
lip. A similar trend can be seen in the evolution of the Zed into the Zeta and
Ultrazed shapes.
A similar evolution can be observed in cassette sections with stiffened flanges and
J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 269

Fig. 1. Evolution of cold-formed purlin sections.

webs, the (perforated) uprights in pallet racks, slotted wall studs and the 2nd and
3rd generation decking profiles. The use of high strength steels in thinner, highly
stiffened sections inevitably leads to more demanding design requirements. The
consequences of this are considered in more detail later in this paper.

2.3. Better corrosion resistance

Undoubtedly, one of the primary reasons for the increase in the practical usage of
cold-formed sections is improved corrosion resistance. This is the result of improved
galvanising and other coating technology, bearing in mind that the protective coating
system is undamaged by the cold-forming process. There is also an increasing use
of cold-formed stainless steel. A 12% chromium steel with no nickel has been used
for housing in Japan. This steel does not have the shiny surface usually associated
with stainless steel but is less prone to rust.

2.4. Improved rolling and forming technology

Modern rolling lines are generally computer controlled from the design office so
that not only can highly accurate complex shapes of precise lengths be produced to
order but also holes, perforations and slots (e.g. web openings for services) can be
punched in precise positions during the rolling process. A significant recent develop-
ment is the automatic end forming of beams at the time of rolling, as shown in Fig. 2.

Fig. 2. End forming of beams (Ward Building Components).

270 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

The rolling of transverse stiffeners in the wide top flange of 3rd generation decking
profiles, as shown in Fig. 3, has been available for some time. Similar technology
allows the rolling of vertical web stiffeners in beam sections. Even more complex
sections may now be formed in two or more parts in which the individual parts are
formed and assembled using high frequency welding in a single operation. This
may involve the combination of different material qualities and thicknesses in a
single component.

2.5. Improved connection technology

Because of the comparative thinness of the material, connection technology plays

an important role in the development of structures formed using cold-formed sec-
tions. The conventional methods of connection, such as bolting and arc-welding are,
of course, available but are generally less appropriate and the emphasis is on special
techniques more suited to thin material. Long-standing methods for connecting two
elements of thin material are blind rivets and self-drilling, self-tapping screws. Fired
pins are often used to connect thin material to a thicker supporting member. More
recently, clinching technology [3] which is very quick, requires no additional
components and causes no damage to the galvanising or other metallic coating
has been taken from the automotive industry and applied to building construction.
An even more recent and significantly stronger innovation is the Rosette [4,5].

3. Developments in applications

This section is concerned with practical applications of cold-formed sections in

structural engineering, particularly building construction, with which the author is
familiar. Historically, the main applications have been such elements as: purlins and
sheeting rails and associated components; cladding and decking; pallet racking and
shelving etc. The developments in technology outlined above are all applicable to

Fig. 3. 3rd generation decking profile.

J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 271

components of this type and their evolution will undoubtedly continue. However,
the main current and foreseeable developments are in a rather different sphere.

3.1. Applications in residential and other low-rise buildings

Although this is not yet so evident in Europe, in other parts of the developed
world there is an extremely rapid development in the use of cold-formed framing
systems for houses and other low-rise construction. This development is being led
by the USA where the rise may be said to be exponential. Pekoz [1] states that,
in the USA, about 500 homes were built in light gauge steel in 1992. This number
rose to 15 000 in 1993 and 75 000 in 1994. A further five-fold increase is estimated
by 2002. The primary framing elements for this form of construction are cold-formed
steel wall studs and floor joists. Light gauge steel roof trusses or rafters are also
used. There is a vast amount of literature on the subject, a good deal of which is
summarised in the Residential Steel Framing Manual published by the American
Iron and Steel Institute [6].
Surprisingly, this major evolution has been accompanied by little development in
cold-formed section technology. The majority of steel framing systems have evolved
from timber framed solutions and have merely involved replacing timber sections by
cold formed steel sections of similar overall size connected together by conventional
methods such as bolting and welding. More recently, advantage has been taken of
the clinching connection methods discussed in Section 2.5, particularly where factory
prefabrication is used. A recent important development is the use of wall studs with
perforated webs [7,8]. These studs have been developed in the Scandinavian coun-
tries in order to reduce thermal bridging in external walls. The web perforations take
the form of arrays of slots which, of course, significantly weaken the web in trans-
verse bending and shear. This has several effects, including a reduced resistance to
distortional buckling, and thus leads to rather complex design problems.
Another significant development is the use of cassette walls [9]. These have the
advantage of providing a weather proof wall as well as a structural frame and avoid
many of the stability problems of stud wall construction. They also act as a shear
diaphragm with regard to horizontal (wind) load and thus avoid the necessity of
providing bracing systems in the plane of the walls. This leads to much simpler
detailing and more rapid construction.

3.2. Factory assembly of prefabricated units

Closely associated with the trend outlined above is the related trend to maximise
prefabrication in the factory. Using jigs and appropriate jointing technology such as
clinching and Rosettes, cold-formed steel components can be rapidly and accurately
assembled into complete structural units for delivery to the site. Such units can take
the form of wall and floor panels and roof trusses. Even more dramatic is the increas-
ing factory prefabrication of complete building modules such as complete hotel
rooms. These may be built and fitted out with bathroom units, furniture and fittings
before being shipped to the site complete and requiring only the connection of exter-
272 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

nal services [10]. In the UK, most out of town hotels and motels are now built in
this way.

3.3. Interaction of cold-formed steel with other materials and components

It is inherent in the nature of thin-walled steel construction that this material is

often seen at its best when it interacts with other materials and building components.
Familiar applications of this principle are: purlins stabilised by cladding; wall studs
stabilised by sheathing; composite decks where embossed metal decking acts together
with in-situ concrete; and sandwich panels where two thin metal faces interact with
a lightweight insulating core. Stressed skin (diaphragm action) offers another
instance of this important principle in which the thin metal cladding of a building
can act to stabilise the framing members and even provide the primary stability of
the complete structure. Various forms of frameless construction take this principle
a stage further.
There are a number of current developments which utilise the interaction of cold-
formed steel members with other materials. These include filled cassettes, which are
substantial C-shaped sections used for wall construction which may be filled with
rigid thermal insulation in order to provide improved stability and resistance to
denting. Another important application is in dry flooring systems in which cold-
formed steel floor joists (or profiled steel decking) act together with a built-up walk-
ing surface to provide enhanced structural performance as well as meeting the
requirements of building physics (acoustic, fire, vibration etc). It is significant that
these two examples are both particularly applicable to residential and low-rise
steel construction.

4. Developments in design procedures for cold-formed sections

A consequence of the various developments described above is that design pro-

cedures are becoming more complicated. Although this may not be to the taste of
all practitioners, it is an inevitable consequence of a high-tech product which is being
put to more demanding use. Those of us who propose design procedures and write
design codes have to find an acceptable compromise between rigorous
(mathematical) design and the aspirations of practical designers who demand sim-
ple design procedures. This section summarises the tools that are available for
research and development. The paper then demonstrates how these are related to
practical design.

4.1. Buckling modes for cold-formed sections

Steel sections may be subject to one of four generic types of buckling, namely
local, global, distortional and shear. Local buckling is particularly prevalent in cold-
formed sections and is characterised by the relatively short wavelength buckling of
individual plate elements. The term global buckling embraces Euler and lateral-
J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 273

torsional buckling of columns and lateral buckling of beams. It is sometimes termed

rigid-body buckling because any given cross-section moves as a rigid body without
any distortion of the cross-section. Distortional buckling, as the term suggests, is
buckling which takes place as a consequence of distortion of the cross-section. In
cold-formed sections, it is characterised by relative movement of the fold-lines. The
wavelength of distortional buckling is generally intermediate between that of local
buckling and global buckling.
It is a consequence of the increasing complexity of section shapes that local buck-
ling calculations are becoming more complicated and that distortional buckling takes
on increasing importance. These two generic modes of buckling may, of course,
interact with each other as well as with global buckling. Whereas much has been
written about local and global buckling, and codes of practice for the design of
structural steelwork contain the relevant clauses, distortional buckling is less well
known and less well documented. This, however, should not be taken as an indication
of its lack of importance. In many cold-formed sections, distortional buckling is at
least as likely as local buckling and warrants similar consideration in design. It is
also likely in relatively slender hot-rolled sections.
Shear buckling may also be either local or global. Local shear buckling can occur
in the slender webs of members subject to bending or in the wide flanges of sheeting
and decking profiles subject to diaphragm action. Global shear buckling generally
only arises when relatively lightweight sheeting or lining profiles are subject to dia-
phragm action. There is a need for greater understanding of both local and global
shear buckling of stiffened tray and cassette sections subject to diaphragm action.

4.2. Tools for research and development into buckling phenomena

The research and development engineer has a range of tools available to him when
considering the design of a new range of cold-formed sections or new applications
of existing sections. Evidently, most cases of practical importance are covered by
modern design codes such as Eurocode 3: Part 1.3 [11]; The American Specification
[12], The Australian and New Zealand Standard [13] and the British Standard BS
5950: Part 5 [14]. However, cases may arise which are not adequately covered by
the available codes or where the codes are over-conservative and more efficient
designs can be obtained by an investment into a more detailed design. The available
design technologies for such situations may be summarised as follows.

4.2.1. Testing
Historically, design by testing has been more widely used for cold-formed sections
than for most other building components. There are a number of reasons for this,
both practical and economic. As analytical methods have improved, testing is now
used less frequently. However, it is still widely used for particular cases such as:
purlins clad with sheeting; perforated racking uprights; (slotted) studs with sheathing
attached; composite decks; sandwich panels etc. It is notable that almost all of these
examples involve the interaction of light gauge steel components with other materials
or parts of the structure as discussed above. These situations are, of course, notori-
274 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

ously difficult to analyse satisfactorily although much progress is being made in this
area. Other situations which are difficult to analyse and where testing may still be
appropriate are those involving significant holes or openings through the member
and those where economic design requires consideration of ultimate limit states in
which there is interaction between buckling and yielding.

4.2.2. Classical methods based on explicit solutions of the governing differential

Although the differential equations which govern the behaviour of thin-walled
metal sections have been known for many years, explicit solutions have been found
for relatively few practical situations. The emphasis, therefore, is on numerical sol-
utions which are considered later. The main application of explicit solutions is in
the analysis of the global buckling (lateral and lateral torsional) of beams and col-
umns. Explicit solutions are also useful in some simple cases of shear buckling.

4.2.3. The finite element method

The finite element method is the most general of the numerical methods which
provide solutions of the governing differential equations. In principle, all of the
required phenomena can be modelled using appropriate finite element techniques.
However, this method is generally considered to be too cumbersome for practical
design and therefore may be seen as the primary method available to researchers.
The primary building-block for the analysis of cold-formed sections is the second-
order thin shell element which can accommodate the full range of section shapes
and buckling phenomena. If a non-linear stress-strain relationship is incorporated
into the analysis, such elements can also model yielding and elastic-plastic buckling.
Contact elements, connection elements and large deflection theory add to the huge
range of facilities that are available to the analyst so that all of the relevant practical
problems can be solved using the finite element method. The disadvantage is, of
course, the considerable cost. This is mainly a consequence of the time spent in data
preparation and post-processing although non-linear problems may take a long time
to run. It should be noted that great care is needed in formulating the correct bound-
ary conditions and this can also consume a considerable amount of time.

4.2.4. The finite strip method

The finite strip method falls into the category of numerical methods that are
specifically designed for prismatic members. Cold-formed sections are generally pris-
matic and the finite strip method has the advantage over the finite element method
of requiring less computer time and memory as well as less data preparation.
From the practical point of view, the second-order finite strip method is parti-
cularly important because bifurcation buckling solutions may be obtained relatively
easily using simple half sine wave displacement functions. This provides whole
section solutions for the full range of buckling phenomena, leading to relatively
new design procedures which will be considered later. Several well-known
researchers (e.g. Hancock, Pekoz, Rhodes) have prepared user-friendly computer
software for this calculation which is available to practical designers. That of Pekoz
J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 275

and Schafer is particularly significant because it is in the public domain:


4.2.5. The 7 degree of freedom prismatic member finite element

The seven degree of freedom prismatic finite element was developed by Barsoum
and Gallagher [15]. It adds the first derivative of rotation about the longitudinal axis
to a conventional six degree of freedom, second-order beam-column element and
thus introduces consideration of lateral torsional buckling. This provides a rational
approach to rigid body buckling when the available solutions based on explicit
solutions of the governing differential equations are not sufficient.

4.2.6. Generalised beam theory (GBT)

Generalised Beam Theory is also applicable to prismatic members and has been
compared to the finite strip method. However, it is much more than an alternative
method and more in the nature of a new theory. It is applicable to all three
generic modes of buckling as well as their interaction. It is, therefore, considered in
some detail in the next section of this paper.

5. The application of generalised beam theory to buckling problems

Generalised Beam Theory (GBT) has been described elsewhere [16,17] and space
precludes a full description here. For our purposes, it is important to note that the
essential concept of GBT is the separation of the behaviour of a prismatic member
into a series of orthogonal displacement modes. It is a particular strength of GBT
that these modes may be considered separately or in any combination in order to
investigate different aspects of the structural response. In GBT, each mode has an
equation and, in second-order format, neglecting the shear deformation terms, the
equation for mode k is:

n n

E kC kV////G k D k V//k B k V ijk

(iWj V)k q for k1,2,%n (1)

where the left superscript k denotes the mode and: kC is the generalised warping
constant; kD is the generalised torsional constant; and kB is the transverse bending
These are the generalised section properties which depend only on the cross-sec-
tion geometry. In addition:
are second-order section properties which relate the cross-section deformations
to the stress distributions,
E and G are the modulus of elasticity and the shear modulus respectively,
V and kW are the generalised deformation resultant and stress resultant,
q is the uniformly distributed load and
n is the number of modes in the analysis.
276 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

For simple cross-sections, the section properties and the ijk values may be calcu-
lated manually but this is cumbersome. In general, this task is best carried out by
computer and a general program has been prepared by the author and his colleague
Dr C. Jiang and has been made available in the public domain.
If the right hand side terms kq of the equation system (1) are zero, the solution
gives the critical stress resultant iW. In general, this requires the solution of an eigen-
value problem in which the analyst is free to choose which modes to include in his
analysis. Various mode interactions can be studied in this way for prismatic members
of varying lengths.
When a constant stress resultant, such as an axial load or a constant bending
moment, is applied along a member which is assumed to buckle in a half sine wave
of wavelength l, GBT allows some particularly simple results to be obtained. Thus
the critical stress resultant for single mode buckling is [17]:

p2 k
l2 k
B (2)

As the wavelength is varied, the minimum critical stress resultant is


(2 E kCBG k D) (3)

and the corresponding half wavelength is

E kC

Eqs. (2)(4) allow a particularly simple examination to be made of any individual

buckling mode, including the distortional modes. No other method known to the
author allows the distortional modes to be isolated in this way. In particular, it can
be seen that critical stress resultant for either local or distortional buckling in mode
k is dependant only on readily calculable section properties. The second-order coup-
ling term ikk is dependant on the load mode i but the buckling half wavelength
depends only on the cross-section properties kC and kB which are independent of
the type of load (e.g. axial or bending). The design expressions based on simplified
models, which will be discussed later, have the same pattern.
There are three different applications of second-order GBT which may be used
to investigate buckling phenomena in cold-formed sections. These are: (1) the above
explicit expressions for single-mode buckling; (2) simple mode interaction based on
the half sine wave assumption which gives rise to elementary eigenvalue problems;
and (3) the general case, without any assumed displacement function, which gives
rise to larger eigenvalue problems based on a finite difference solution of the coupled
differential equations. These may be illustrated by considering a particular example.

5.1. The application of GBT to the buckling of a column section under axial load

A particular class of member for which distortional buckling may be important is

the upright section used in the construction of pallet racking. These members usually
J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 277

contain regular arrays of perforations in order to allow beams to be fixed at heights

which are not pre-determined. However, this feature will be ignored in the follow-
ing illustration.
Fig. 4 shows a typical upright section which has previously been analyzed by the
authors [18]. Here, for simplicity, it is idealised with sharp corners. By including an
artificial node in the middle of the long side, local buckling of this face of the section
is included in the analysis. Local buckling of the other faces is clearly much less
critical. Additional (optional) nodes are also included at the ends of the lips. These
detect a pure local buckling mode in the lips. There are, thus, a total of 11 nodes
in the cross-section and these give rise to 11 orthogonal modes of deformation in
the GBT analysis. Fig. 4 also shows the modes of distortion which are determined
by GBT. Mode 1 is axial shortening, modes 2 and 3 are bending about the two
principal axes and mode 4 is pure torsion. These are the rigid-body modes with
no distortion or local buckling of the cross-section. The remaining modes all involve
either distortion or local buckling. Modes 5 to 8 are the distortional modes, mode
9 is local buckling of the long face and modes 10 and 11 are local lip buckling modes.
Eq. (4) shows that, if kB is not equal to zero, a given mode k has a clearly
defined buckling half-wavelength and, therefore, from Eq. (3), a clearly defined mini-
mum critical stress. All of the modes except the rigid-body modes fall into this
category and the critical wavelengths and stresses are given in Table 1.
It may be noted here that the modern tendency to introduce more folds in order
to reduce the impact of local buckling has the effect of introducing more distor-
tional modes!
Fig. 5 shows the buckling behaviour in some of the individual modes according
to Eqs. (2)(4) together with the lateral torsional (rigid-body) buckling curve and
lower bound curves when all modes are considered together. There are some entirely
typical features which can be clearly seen from these results:

The local and distortional buckling modes are clearly identified and their buckling

Fig. 4. Section RA17 with mode shapes in the order 1 to 11.

278 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

Table 1
Buckling behaviour in the individual modes

Mode Buckling half-wavelength (cm) Buckling stress (kN/cm2)

5 52.1 31.2
6 41.3 52.4
7 29.2 309.4
8 25.6 486.8
9 6.31 38.7
10 5.46 61.4
11 5.45 61.3

Fig. 5. Buckling curves for section RA17.

stresses may be readily calculated based on consideration of the complete cross

The three local buckling modes have very short wavelengths and, for the purposes
of design, they can be treated by the usual effective width approaches
The critical distortion mode is a symmetrical rotation of both flanges about the
junction with the upper web. Distortional buckling modes have intermediate
For this type of problem (singly symmetric section subject to axial load), there
is little interaction between distortional and lateral-torsional (rigid-body) buckling
and it is sufficient to consider distortional buckling in isolation.
J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 279

In general, with GBT, the critical distortional buckling mode is symmetrical and
can be easily identified. Eq. (3) then provides a simple design expression based on
the calculated section properties. However, at the present time, GBT is not readily
available to structural designers and it is appropriate to consider other simplified
approaches. Not surprisingly, as discussed in Section 6.2 of this paper, other authors
have attempted to model the symmetrical mode 5 as shown in Fig. 4, though without
a formal evaluation of the alternative modes revealed by GBT.

5.2. The application of GBT to the buckling of a beam

The buckling behaviour of beams bent about the major axis differs from that of
columns in a number of respects. Fig. 6(a) shows a typical cold formed section beam.
Ignoring considerations of local buckling, which do not add anything to the argument
here, the section has 6 natural nodes and therefore there are 6 orthogonal modes of
buckling. These are shown in Fig. 6(b) and are 4 rigid-body modes and 2 distor-
tional modes.
When the beam is bent about the major axis, it is well known that individual
lateral and torsional modes have no significance and the only rigid-body buckling
mode is a combination of modes 3 and 4, namely lateral torsional buckling. In the
same way, the distortional modes 5 (symmetrical) and 6 (antisymmetrical) have no
individual significance and the only distortional mode is a combination of the two
such that most of the distortion takes place in the compression flange with the flange
in tension playing a minor role. Assuming again that the bucking mode is a half
sine wave, GBT again allows a simple calculation for the case of pure bending. The
relevant buckling curves are shown in Fig. 7.
The above results are typical and the general conclusions are similar to those
for columns:

Fig. 6. Buckling modes for a lipped channel section beam.

280 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

Fig. 7. Buckling curves for a lipped channel section beam in uniform bending.

The critical distortion mode is a rotation of the compression flange about its junc-
tion with the web. Distortional buckling modes have intermediate wavelengths.
For this type of problem (channel section beam bent about the minor axis) there
is little interaction between distortional and lateral-torsional (rigid-body) buckling
and it is sufficient to consider distortional buckling in isolation.

5.3. The application of GBT to the buckling of a restrained beam

Unrestrained beams are, of course, relatively rare. It is more usual for cold-formed
section beams to be at least partly restrained by the floor or roof that they support.
An important practical case is that of a purlin supporting a lightweight roof. In this,
and other similar cases, the purlin receives both lateral and torsional restraint to one
flange. It is an important feature of GBT that such restraints can be incorporated
into the member properties [19]. Consider the channel section from above with the
upper flange restrained in position laterally and restrained torsionally by a spring
with a specified stiffness cq. This is a typical representation of a purlin. Fig. 8 shows
the orthogonal modes of deformation, computed with the aid of GBT, for the purlin
section shown in Fig. 6 and for the typical case of cq =1.0 kNcm/radian per cm.
Mode 1 is axial strain which is not relevant for a purlin in bending and the remain-
ing modes are each associated with a buckling mode. Mode 3 is global buckling by
bending about the major axis which is unaffected by the restraints. Due to the exist-
ence of the restraints, the mode of buckling by bending about the minor axis of the
section is eliminated and the mode of pure torsion about the shear centre is turned
into a combination of torsion and distortion (mode 2). Modes 4 and 5 are the distor-
tional modes.
Under downward load, such a purlin usually fails in distortional buckling with a
J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 281

Fig. 8. Orthogonal deformation modes for a restrained purlin.

combination of modes 4 and 5. However, for the important case of uplift load, there
is a transition from distortional buckling to the combined torsional-distortional mode
as the span increases. The point of transition depends on the amount of restraint
which is received from the cladding and, if this is sufficiently great, the torsional
distortional mode may not occur at all, regardless of the span. GBT can, of course,
deal rather elegantly with this complexity.

6. Current design models and their deficiencies

6.1. Models for local buckling

Local buckling of thin-walled sections has been known for many years and has
been well researched. The basic phenomenon is illustrated in Fig. 9 which shows
local buckling in a cassette column with an intermediate stiffener in the wide flange.

Fig. 9. Local buckling in a cassette section column.

282 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

As shown, plate elements which are adequately stiffened along both longitudinal
edges tend to buckle into approximately square waves. For economic design, it is
essential to consider the post-buckled condition and, in this context, the primary
building block for cold formed section design is the concept of effective width
beff illustrated in Fig. 10.
The usual effective width formula is the semi-empirical formula due to Winter

beffrb where if lp0.673; r1.0 if lp0.673; r 1.0

0.22 1
lp lp (5)

in which the plate slenderness lp is given by:

s Ek
seff bp seff
l p 1.052 (6)
cr t s

where seff is the maximum compressive stress in the plate element; scr is the critical
stress for elastic buckling of the plate element; E is the Youngs modulus; ks is the
buckling factor which = 4.0 for a simply supported plate in uniform compression
and 0.43 for an outstand plate element with one edge free.
Elements stiffened on one side only are much less stable and can also be treated
by an analogous effective width approach. However, such unstiffened elements rarely
arise in practice because the free edges are usually restrained by lips. The modern
tendency is to use compound lips as illustrated in Fig. 1 (Multibeam Mark III), Figs.
2 and 9.
Eurocode 3: Part 1.3 [11] gives some comprehensive rules for the determination
of effective widths under different stress conditions. In principle, the effective widths
of the individual plate elements may be combined to give an effective section and
member design completed using conventional techniques. However, this apparent
simplicity conceals a number of difficulties. As illustrated by Fig. 9, individual plate
elements do not buckle in isolation but interact with each other. Although some
codes (e.g. BSI 1987 [14]) give an approximate treatment of this phenomenon, it
can only be dealt with accurately by a whole-section analysis.

Fig. 10. Effective width beff of plane element stiffened along both edges.
J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 283

Furthermore, Fig. 9 suggests that both the intermediate stiffener in the wide flange
and the compound lip stiffeners are fully effective so that they remain straight
during local buckling of the plate elements. With modern, highly stiffened sections,
this is often not the case. Stiffeners may be partially effective so that stiffener buck-
ling interacts with local plate buckling. Highly stiffened decking profiles are a special
case for which some complex but accurate design procedures have been derived
[20]. Eurocode 3: Part 1.3 [11] gives some design rules for more general situations
but these are complicated to use and not particularly accurate [21]. Evidently, this
is another situation where design based on an analysis of the whole section is to
be preferred.

6.2. Models for distortional buckling

Recently, distortional buckling has received the attention of a number of

researchers. The research has now reached the point where design procedures suitable
for inclusion in codes of practice are beginning to emerge. Both columns and beams
will be considered in turn. Mention will also be made of the special case of beams
(purlins or sheeting rails) which are partially restrained by the cladding that they
support as this is an important practical case.

6.2.1. Model for distortional column buckling

The considerations discussed above in the context of GBT point the way to an
appropriate model from which simplified analytical expressions for the distortional
elastic buckling stress may be derived. Such a model, which was originally developed
by Lau and Hancock [22], is shown in Fig. 11. In contrast to GBT, in which the
whole cross-section is considered, their analytical expressions were based on a flange
buckling model in which the flange was treated as a compression member restrained
by a rotational and a translational spring. The rotational spring stiffness kf represents
the torsional restraint from the web and the translational spring stiffness kx represents
the restraint to translational movement of the cross section.
In Lau and Hancocks analysis [22], it is shown that the translational spring stiff-
ness kx does not have much significance for the buckling mode under consideration

Fig. 11. Analytical model for distortional column buckling.

284 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

and the value of kx was assumed to be zero. The key to evaluating this model is to
consider the rotational spring stiffness kf and the half buckling wavelength l, while
taking account of symmetry. Lau and Hancock [22] give a detailed analysis in which
the effect of the local buckling stress in the web and of shear and flange distortion
were taken into account in determining expressions for kf and l. These are too
complex to reproduce here but a direct analogy with Eq. (4) may be noted.
Davies and Jiang [23] have carried out a detailed calibration of this model against
the more accurate whole-section analysis offered by Generalised Beam Theory
(GBT). They noted the sensitivity of this model to the value of the spring stiffness
kf and were able to propose a significant improvement.

6.2.2. Model for distortional buckling in beams

Analytical expressions for the distortional buckling of thin-walled beams of gen-
eral section geometry under a constant bending moment about the major axis have
been developed by Hancock [24]. These analytical expressions were based on the
simple flange buckling model shown in Fig. 12 (together with an improvement pro-
posed by Davies and Jiang [25]) in which the flange was treated as a compression
member with both rotational and translational spring restraints in the longitudinal
direction. The rotational spring stiffness kf and the translational spring stiffness kx
represent the torsional restraint and translational restraint from the web respectively.
In his analysis, Hancock again chose the translational spring stiffness kx to be zero.
These beam models are, of course, directly analogous to the column model shown
in Fig. 11. The only significant difference lies only in the stiffness of the rotational
spring and the necessary modifications to the design expressions for the rotational
spring stiffness kf and the buckling length l are given in Hancocks paper [24]. This
then leads to the similar equations for the critical stress for distortional buckling.
When comparing the results given by Hancocks expressions and GBT for the
distortional buckling of channel section beams [25], it was found that good results
were obtained for small web depths but that, for deeper sections, Hancocks
expression could become unsafe due to the neglect of web local buckling and the
assumption of fixity at the bottom (tension) end of the web. In order to improve

Fig. 12. Comparison of design expressions with the results given by GBT for channel section beams.
J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 285

the accuracy of the design expressions, modifications were proposed by authors by

introducing a reduction coefficient into the equation for kf and by assuming the
tension end of the web to be pinned. It was shown that the modified design
expressions give a better estimate of the critical bending moment than Hancocks
original expressions. This again emphasises the sensitivity of models of this type to
the value assumed for kf and the preference for whole-section analysis.

6.2.3. Models for the buckling of restrained beams

Evidently, as discussed in Section 5.3, when one flange is elastically restrained,
the behaviour is more complex than for an unrestrained section and there are two
distortional modes to consider in design. Lack of space precludes a full discussion
of this topic but it is sufficient to say that Pekoz and Soroushian [26] have described
an adequate model for the combined torsional-distortional mode 2 and that Hancocks
distortion buckling model in Fig. 12 can be readily modified further [27] to include
the additional torsional restraint from the cladding. Both models must be considered
and the more critical of the two gives a satisfactory design expression for a
restrained purlin.

7. Design using whole-section models

At a number of points in the preceding discussion, the desirability of using whole-

section analysis rather than simplified design models has been emphasised. Whole
section models are believed to be particularly appropriate for dealing with local and
distortional buckling and the interaction between them. This is now considered to
be a practical proposition for designers because of the ready availability of finite
strip or GBT bifurcation analyses based on half sine wave displacement functions.
However, bifurcation analyses only give the critical stress for elastic buckling.
For design purposes, it is generally necessary to consider the interaction of buckling
and yielding. Furthermore, for economic design, it also necessary to consider the
beneficial effect of post-buckling. There are two complementary approaches which
are available to deal with this situation.

7.1. The AyrtonPerry equation

The AyrtonPerry equation is used in most modern codes to combine yielding

and global buckling. Solutions of practical accuracy can generally be obtained by
combining the theoretical load (or stress) for elastic (bifurcation) buckling with the
corresponding yield load (or stress) using the equation:
c but c1 with f0.5[1a(l0.2)l2] (7)
f+[f l2]0.5

where c is the reduction factor for buckling with respect to the unbuckled capacity;
a is an imperfection factor and l is the relative slenderness in the relevant buck-
ling mode.
286 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

In Eurocode 3: Part 1.3 [11], this equation is applied to the flexural buckling of
columns and to the lateral torsional buckling of beams with l equal to P/Pcr and
M/Mcr respectively. Local buckling may be allowed for by basing the yield load
(or stress) on the effective cross section. Up to the present time, this part of the
calculation has been based on conventional (effective width) approaches with the
problems discussed in some detail above.
In principle, Eq. (7) is equally valid when used to allow second-order elastic GBT
solutions to be used to give reliable estimates of the failure loads for both beams
and columns in a wider range of practical situations. However, the AyrtonPerry
approach does not include any allowance for post-buckling so that it is likely to be
over-conservative when applied to situations where local buckling is significant. It
is with this in mind that Schafer and Pekoz [28] have recently proposed the direct
strength approach.

7.2. Direct strength design of cold-formed sections

The direct strength method recognises that the available models for local and
distortional buckling design are far from simple and have significant limitations. It
also recognises that the current trend is to increase the complexity of section shapes
and also increases the complexity of the required mathematical models. It therefore
proposes a formal design procedure based on elastic bucking solutions for the com-
plete cross-section. The particular characteristic of this procedure is that it also
recognises that, for economic design, it is necessary to take advantage of the post-
buckling strength. It therefore takes the conventional effective width Eqs. (5) and
(6) and applies these to the complete cross-section. The method has been initially
expressed in terms of flexural members but it is clearly also applicable to members
under axial compression. Thus, the design moment capacity is given by:

M 0.673; r1.0 if l0.673; r1.0 l l

Mcr 0.22 1
if l (8)

The elastic buckling moment Mcr for local or distortional buckling may be readily
obtained from either the finite strip method or GBT using half sine wave displace-
ment functions. The proposals of Schafer and Pekoz [28] offer two possible
additional refinements to the basic procedure described above. Noting that there may
be decreased post-buckling capacity in the case of distortional buckling, a reduction
factor is suggested for this case. Alternatively, a modified equation may be used for
r in order to obtain better agreement with the experimental results.
The above proposal, and the possible refinements, have been calibrated against
the AISI [12] specification for a total of 574 test results for unrestrained beams
obtained by 17 researchers. It is shown that, as well as being simpler, the initial
J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288 287

form of the direct strength method is conservative and at least as accurate as the
AISI specification. The reduction factor for distortional buckling does not improve
matters but the second proposed improvement results in a distinct improvement on
the AISI design rules.

8. Conclusions

Cold-formed sections have been enjoying an exciting period of development and

this is likely to continue for the foreseeable future. This development has been a
combination of improvements in technology and developments in applications. It has
placed researchers under some pressure to find adequate practical design procedures
for increasingly complicated section shapes. Research workers have responded to
this challenge in two ways. Viable design models have been developed for the local
and distortional buckling, and the interaction between them, for most of the sections
of interest to the designers of building structures. However, these models are far
from simple and a more interesting possibility arises from the ready availability of
bifurcation buckling analyses of the complete cross-section using either the finite
strip method or Generalised Beam Theory with a half sine wave displacement func-
For cold-formed steel columns and beams with the proportions typically used in
practice, distortional buckling is often critical. Generalised Beam Theory (GBT) pro-
vides a particularly appropriate tool with which to analyze distortional buckling in
isolation and in combination with other buckling modes. It also provides a yardstick
with which other simplified methods may be assessed.
The results of such bifurcation analyses need to be combined with a yielding
limitation in a semi-empirical fashion. There are two complementary procedures
available. The AyrtonPerry equation is well-established for situations where the
predominant buckling mode is global (lateral or lateral-torsional). An allowance for
post-buckling behaviour may be made in the calculation of the yield load or stress.
The direct strength method makes a more formal allowance for post-buckling and
is evidently more appropriate when local buckling is significant. Combining the two
procedures gives the best available design procedure when members with a range
of length or span need to be considered.
At the present time, the author understands that the direct strength approach is
being seriously considered for inclusion in certain National design standards. The
author believes that this is a positive step forward which should be encouraged.


[1] Pekoz T. Possible future developments in the design and application of cold-formed steel. Keynote
lecture, ICSAS99, 4th Int. Conf. Light-weight Steel and Aluminium Structures. Espoo, Finland
2023 June 1999.
[2] Hoglund T. Strength of butt welds in extra high strength steel. In: Proc. 3rd Int. Conf. on Steel and
Aluminium Structures. Istanbul, 1995.
288 J.M. Davies / Journal of Constructional Steel Research 55 (2000) 267288

[3] Davies R, Pedreschi R, Sinha BP. The shear behaviour of press-joining in cold-formed steel struc-
tures. Thin-Walled Structures 1996;25(3):15370.
[4] Makelainen P, Kesti J. Advanced method for lightweight steel joining. Jnl of Constructional Steel
Research 1999;49:10716.
[5] Pasternak H, Komann S. Shear behaviour of rosette joints Tests and finite element analysis. In:
Proc. Conf. Eurosteel 99. CVUT Prague, 1999:47982.
[6] AISI. Residential Steel Framing Manual. American Iron and Steel Institute. Chestertown, MD 21690,
USA, undated.
[7] Hoglund T. Design of light gauge studs with perforated webs. In: Nordic Steel Conference,
Bergen, 1998.
[8] Kesti J, Makelainen P. Design of gypsum-sheathed perforated wall studs. In: Second World Conf.
on Steel in Construction. San Sebastian, 1998.
[9] Davies JM. Light gauge steel cassette wall construction. In: Nordic Steel Construction Conference.
Bergen, 1998:42740.
[10] Davies JM, Leach P, Kelo E. The use of light gauge steel in low and medium rise modular buildings.
In: Proc. 3rd Int. Conf. on Steel and Aluminium Structures, ICSAS 95. Istanbul, 1995.
[11] CEN. Eurocode 3: Design of steel structures. Part 1.3: General rules, Committee European de Nor-
malisation, ENV 1993-1-3, 1996.
[12] AISI. Specification for the design of cold-formed steel structural members. Washington: American
Iron and Steel Institute, 1996.
[13] AS/NZS. Cold formed steel structures, Australian and New Zealand Standards, AS/NZS 4600: 1996,
Sydney, 1994.
[14] BSI. Structural use of steelwork in building: Code of practice for the design of cold formed sections,
BS 5950: Part 5. London: British Standards Institution, 1987.
[15] Barsoum R, Gallagher R. Finite element analysis of torsional and torsional-flexural stability prob-
lems. Int Jnl of Numerical Methods in Engineering 1970;2:33552.
[16] Davies JM, Leach P. First-Order Generalised Beam Theory. J Construct Steel Research
[17] Davies JM, Leach P, Heinz D. Second-order generalised beam theory. J Construct Steel Research
[18] Davies JM, Leach P. Some applications of generalised beam theory, In: 11th Int. Speciality Conf.
on Cold-Formed Steel Structures. St. Louis, Missouri, 1992:479501.
[19] Davies JM, Jiang C, Leach P. The analysis of restrained purlins using Generalised Beam Theory,
In: 12th Int. Speciality Conf. on Cold-Formed Steel Structures. St. Louis, Missouri, Oct. 18-19
[20] Davies JM. Sheeting and decking. In: Rhodes J, editor. Design of cold-formed steel members, chapter
10. Elsevier Applied Science, London and New York 1991. pp. 33960.
[21] Kesti J, Davies JM. Local and distortional buckling of thin-walled short columns. Thin-Walled
Structures 1999;34:11534.
[22] Lau SCW, Hancock GJ. Distortional buckling formulas for channel columns. Journal of Structural
Division, ASCE 1987;113(5):106378.
[23] Davies JM, Jiang C. Design of thin-walled columns for distortional buckling, In: 2nd Int. Speciality
Conf. on Coupled Instabilities in Metal Structures, CIMS 96, Liege, 1996:16572.
[24] Hancock GJ. Design for distortional buckling of flexural members, In: Proc. 3rd Int. Conf. on Steel
and Aluminium Structures, Istanbul, 1995. (also in Thin-Walled Structures 1997;27:312.
[25] Davies JM, Jiang C. Design of thin-walled beams for distortional buckling, In: 13th Int. Speciality
Conf. on Cold-Formed Steel Structures, St. Louis, Missouri, 1996:14153.
[26] Pekoz T, Soroushian P. Behaviour of C- and Z-purlins under wind uplift, Report No. 81-2, Depart-
ment of Civil Engineering, Cornell University, Ithaca, NY, 1982.
[27] Davies JM, Jiang C. Design of thin-walled purlins for distortional buckling, In: TWS Bicentenary
Conf. on Thin-Walled Structures, Strathclyde, Glasgow, 1996.
[28] Schafer BW, Pekoz T. Direct strength prediction of cold-formed steel members using numerical
elastic buckling solutions. In: Thin-Walled Structures, Research and Development. Proc. 2nd Inter-
national Conf. on Thin Walled Structures. Singapore, Dec. 1998, Elsevier, 1998:13744.