Sie sind auf Seite 1von 18

Thin-Walled Structures 42 (2004) 515–532

www.elsevier.com/locate/tws

Minimum weight of cold-formed steel


sections under compression
Y.S. Tian, T.J. Lu 
Department of Engineering, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, UK
Received 20 October 2002; received in revised form 3 November 2003; accepted 23 December 2003

Abstract

This paper presents a combined theoretical and experimental study on the minimum
weight and the associated optimal geometric dimensions of an open-channel steel section
with given length subjected to a prescribed axial compressive load. Sections both with and
without lips are analyzed. The results obtained using a nonlinearly constrained optimization
method are compared with those estimated from a simple-minded optimization procedure
that assumes the simultaneous occurrence of all failure modes in a minimum weight struc-
ture. The types of failure mode considered include yielding, flexural buckling, torsional–
flexural buckling, and local buckling. The failure criterion is based purely on compressive
strength, with other possible design constraints (e.g. bending stiffness, minimum gauge and
cost) ignored. The effects of end support conditions and restraint on torsional buckling are
examined. The load capacity of a C-section calculated according to the 1998 British Stan-
dard Institution’s specifications on Structural Use of Steelwork in Building is used to check
the validity of theoretical predictions. Finally, two new C-sections with lips were designed
and manufactured based on the optimal results, and tested. Test results confirm the analytical
predictions, with the optimal C-sections performing much better than the existing ones.
# 2003 Elsevier Ltd. All rights reserved.

Keywords: Cold-formed steel; C-channel section; Optimization; Minimum weight design; Experimental
measurement

1. Introduction

Cold formed sections are found in many applications such as window reinforce-
ments, metal track and stud for partitioning systems, metal framed buildings,


Corresponding author. Tel.: +44-1223-766316; fax: +44-1223-332662.
E-mail address: tjl21@cam.ac.uk (T.J. Lu).

0263-8231/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tws.2003.12.011
516 Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532

mezzanine flooring etc. Because of its high strength but good forming properties,
the material generally used is galvanized mild steel which is initially supplied in
large coils and later slit to appropriate strip widths. One of the major growth areas
for cold formed steel sections has been in the structural sector, particularly track
and stud for plasterboard partitioning support. Steel track and stud is seen as an
environmentally friendly, recyclable alternative to the timber that is traditionally
used for studding and roofing trusses. The replacement of timber with steel has
been more prevalent in areas where timber resources are scarce, and also in com-
mercial or community applications where other advantages such as speed of
assembly and fire retardance are more important.
Strength-based minimum weight has always been an important aspect in the
design and processing of lightweight structures [1–10]. Because a high percentage
(>50%) of the final product cost of cold-formed sections is in the raw material, the
amount of steel in the section must be minimized for it to become a viable alterna-
tive to timber. As a result, the most successful companies in the industry are those
who have developed novel ‘rigidizing’ techniques to improve their section proper-
ties whilst reducing the steel content. In addition to this, it is vital to optimize
manufacturing processes to ensure high throughput on mills and low downtime.
Nevertheless, there appears to be significant opportunities to improve upon the
technical foundation. Traditionally, design optimizations have been empirical and
there has been little evidence of application of advanced development tools that are
based on analytical modeling and numerical simulation (e.g., the method of finite
elements). Although this is widely used in the design of buildings etc., it is not cur-
rently used to optimize the sections themselves on a smaller structural scale. The
objective from the product perspective is to reduce the material content of the sec-
tions which will further accelerate the trend from timber to steel in the construction
industry. Even rough empirical developments in this area have resulted in savings
of more than 10% in raw material content. This study will explore scientific ways in
which section design might be optimized to achieve enhanced properties without
increasing the material content.
Despite its obvious practical importance, only a few previous studies have
focused on the optimal design of cold-formed steel C-sections. Based on the design
specifications of the American Iron and Steel Institute [11], Seaburg and Salmon
[12] studied the minimum weight of hat-shaped sections by using the direct and
gradient search techniques, although only one example was given due to the com-
plexity of the numerical problem. Dinovitzer [13] optimized the lip dimension of a
C-section according to the provisions of the Canadian Standards Association
(CAN/CSA-S136); the optimization procedure was not based on strength, but on
increasing the moment resistance of the section while minimizing its cross-sectional
area. Adeli and Karim [14] developed a neural network model for nonlinear opti-
mization problems, and applied it to optimize simply-supported steel beams sub-
jected to transverse loading. Beams with hat-, I-, and Z-shaped cross-sections were
optimized according to the AISI allowable stress design specifications. Al-Mosawi
and Saka [15] included in their optimization procedure the stresses resulting from
warping of thin-walled sections and obtained the optimal cross-sectional shapes of
Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532 517

symmetrical channels, unsymmetrical channels and Z-sections subjected to a uni-


form transverse loading. Surprisingly, our literature survey has failed to locate a
single journal article on the minimum weight design of cold-formed C-sections sub-
jected to axial compression, the most common type of load experienced by steel
track and stud that connect the plasterboards of a partition wall.
To find the minimum weight of a compression structure is a challenging prob-
lem, as the total number of variables and constraints is usually large for compli-
cated sections and structures. For a C-section under compression, its minimum
weight depends upon the section height, flange length, lip length and wall thickness,
subject to the provisos that anywhere in the section there is no yielding, local buck-
ling, global buckling, flexural buckling, and torsional–flexural buckling [13–17].
Here, local buckling becomes increasingly more important as the thickness of the
section decreases, whereas torsional–flexural buckling needs to be accounted for
whenever the shear center of the section is not coincident with its centroid.
This paper studies the minimum weight of a C-channel steel section, either with
or without lips (Fig. 1a,b). A nonlinearly constrained optimization procedure
based on the sequential quadratic programming (SQP) algorithm is used, and its
predictions on the minimum weight and the corresponding cross-sectional shape
are compared with those obtained by using a simpler, albeit approximate,
approach proposed initially by Gerard [7] for a hollow square tube but adapted
here for open-channel sections. The load capacity of a steel C-section calculated
according to BS 5950 [18] is used to check the validity of the optimized results. The
optimization is restricted to C-sections subjected to axial compression, with factors
other than strength neglected, e.g. cost, minimum gauge, and longitudinal/bending
stiffness. Based on the optimized results, C-sections with lips are fabricated and tes-
ted to further validate the present design methodology.

Fig. 1. Geometry and notations of a C-channel section: (a) With lips; (b) without lips.
518 Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532

2. Statement of problem

Consider a C-channel steel section of span length L, depth b1 , width b2 and


thickness t, either with or without lips of length b3 (Fig. 1a,b). The Young’s modu-
lus, Poisson ratio, uniaxial yield strength and weight density of the material are
denoted by E, m, ry and q, respectively. Optimal designs of the section subjected to
an axial compressive force P will be considered, with the section weight minimized
under the constraints on the maximum stress that none of the failure modes listed
below occurs:

. Yielding
. Overall column buckling
– Flexural buckling: bending about the principal axis
– Torsional buckling: twisting about the shear center
– Torsional–flexural buckling: simultaneous bending and twisting

. Local buckling of individual elements (web, flange and lip).

The weight of the section is given by:


W ¼ qtLðb1 þ 2b2 þ 2b3 Þ ð1Þ
The axial stress acting on the section is
P
r¼ ð2Þ
ðb1 þ 2b2 þ 2b3 Þt

2.1. Yielding

Yielding of the section is avoided if


r  ry ð3Þ
where ry is the yield stress of steel. The range of ry is relatively large for steel
(100 MPa  ry  350 MPa), with little variation in material cost. Consequently, ry ¼
350 MPa will be assumed below. Furthermore, the behavior of the section after initial
yielding will not be considered in this study, and hence no strain-hardening effect is
accounted for in subsequent calculations.
2.2. Flexural buckling

The elastic flexural buckling load Pe for a long column is given by the Euler for-
mula, as:
p2 EI
Pe ¼ ð3Þ
ðKLÞ2
where I is the moment of inertia and K is the effective length factor, with K ¼ 1 for
simply supported ends and K ¼ 0:7 for clamped ends [18]. For a C-channel section,
Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532 519

buckling along two principal directions—denoted separately by the x and y axes,


with x taken as the axis of symmetry—needs to be considered [17], i.e.

p2 EIx
Pex ¼ ð5aÞ
ðKLÞ2
p2 EIy
Pey ¼ ð5bÞ
ðKLÞ2
where
b32 t
Ix ¼ ½2ðb1 =b2 Þ3 þ 6ðb1 =b2 Þ2  ðb1 =b2  2b3 =b2 Þ3 ð6aÞ
12
b3 t b2 þ 4b3 þ b1 ð2 þ 6b3 =b2 Þ
Iy ¼ 2 ð6bÞ
3 b1 þ 2b2 þ 2b3
are the moments of inertia about the x and y axes, respectively (Fig. 1). Flexural
buckling is dominated by the smaller of Pex and Pey . Upon substituting Ix ¼ Arx 2
and Iy ¼ Ary 2 into Eqs. (4) and (5), the constraints on the axial sectional stress r

P=A due to Euler buckling are obtained as:

p2 E
r  rex
ð7Þ
ðKL=rx Þ2
p2 E
r  rey
 2 ð8Þ
KL=ry
pffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffi
where rx ¼ Ix =A and ry ¼ Iy =A are the radii of gyration about the x and y
axes, respectively.
2.3. Torsional–flexural buckling

A closed section will normally not buckle under torsion due to its relatively large
torsional rigidity. For an open thin-walled section, because its shear center does
not always coincide with its centroid, three modes of buckling failure are possible:
flexural buckling, torsional buckling, and torsional–flexural buckling. When an
open section column buckles in the torsional–flexural mode, bending and twisting
of the section occur simultaneously. With reference to Fig. 1, the distance from the
centroid to the mid-line of the web is given by:
b2 ðb2 þ 2b3 Þ
xc ¼ ð9Þ
b1 þ 2b2 þ 2b3
the distance from the shear center to the web mid-line is:
 
tb2 b3 b21 1 b2 2b2
e¼ þ  32 ð10Þ
Ix 2 4b3 3b1
and the distance from the centroid to the shear center is x0 ¼ xc þ e.
520 Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532

The critical torsional–flexural buckling load Ptf of an open section column as


illustrated in Fig. 1 is given by [17]:
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
Ptf ¼ ðPex þ Pt Þ  ðPex þ Pt Þ2 4bPex Pt ð11aÞ
2b
Alternatively, a more conservative estimate can be obtained using the following
equation:
Pt Pex
Ptf ¼ ð11bÞ
Pt þ Pex
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where b ¼ 1  ðx0 =r0 Þ2 , r0 ¼ r2x þ r2y þ x20 þ y20 is the polar radius of gyration
about the shear center (x0 ,y0 ), and Pt is the torsional buckling load given by
" #
1 p2 ECw
Pt ¼ 2 GJ þ : ð12Þ
r0 ðKt LÞ2

Here, J ¼ ðt3 =3Þðb1 þ 2b2 þ 2b3 Þ is the St. Venant torsion constant, G ¼ E=2ð1 þ mÞ
is the shear modulus, Cw ¼ b22 tð4b33 þ 3b21 b3 þ 6b1 b23 þ b2 b21 Þ=6  Ix e2 is the warping
constant of torsion, and Kt is the effective length factor for torsional buckling; for
simply supported ends, Kt ¼ 1, whereas for clamped ends, Kt ¼ 0:5. For an open-
channel section, Kt ¼ 0:707 is recommended by BS 5950 [18].
Finally, the design constraint on minimum weight due to torsional–flexural
buckling can be expressed as:
rt rex
r  rtf
ð13Þ
rt þ rex
where rt ¼ Pt =A is the critical torsional buckling stress.
2.4. Local buckling

For a C-section with lips, the local buckling of three relatively long plates of
equal length L but different widths b1 , b2 and b3 need to be considered. Each plate
is supported differently along its edges. To avoid local buckling, the axial stress of
the section r must not exceed the smallest of three critical buckling stresses, namely
r  minðrlb1 ; rlb2 ; rlb3 Þ ð14Þ
where

ki p2 Eðt=bi Þ2
rlbi ¼ ; i ¼ 1; 2; 3: ð15Þ
12ð1  m2 Þ
Here, ki is the plate bucking coefficient, with k1 ¼ k2 ¼ 4 for a stiffened element
supported either by a web or flange on both longitudinal edges and k3 ¼ 0:425 for
an element unstiffened at least at one of its edges (i.e. the lip). If no lips are
present, the flange is taken to be unstiffened with k2 ¼ 0:425.
Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532 521

In practice, the C-section can support a load higher than that dictated by the
local buckling stress rlb . By introducing the concept of effective width, the load
capacity of the section after local buckling has occurred can be estimated by using
the procedure described in Appendix A.
It should be pointed out here that the focus of the present paper is on illustrat-
ing how to apply simple optimization procedures in section design and hence, for
simplicity, among the failure modes described above no consideration is given to
residual stresses, interaction between local and overall buckling modes, as well as
inelastic and post buckling effect, which have some influence on the final failure
load of the section. However, to evaluate the reliability of the present optimization
procedure, the predictions will be compared with those calculated by BS 5950 [18]
in which inelastic effect is considered as well as experimental measurements on
non-optimized and optimized steel sections.

3. Optimization procedure

The maximum load, P, and the length of the section, L, are assumed to be speci-
fied. Define the dimensionless load parameter as P ¼ P=EL2 . For convenience, the
following nondimensional geometric design variables are introduced:


b1 ¼ b1 =L; 
b2 ¼ b2 =L; 
b3 ¼ b3 =L; t ¼ t=L: ð16Þ

Notice that the lips shown in Fig. 1a are introduced to a C-section mainly to
reinforce the flange and hence increase the flexural stiffness and stability of the
whole stud [13,16,18]. It has been established that the flange stiffening increases
initially with increasing lip length b3 until local buckling stress is reached in the lip,
whereas flange stiffening decreases if b3 is further increased [13]. In subsequent cal-
culations, unless otherwise stated, the recommendation by BS 5950 [18] that b3 ¼
b2 =5 will be adopted. As a result, there are only three independent geometric vari-
ables: b1 , b2 and t.

3.1. Non-linear optimization (SQP method)

With 
b3 ¼ 
b2 =5, the dimensionless weight is:

 
 ¼ W ¼ 
W b1 þ 2:4
b2 t: ð17Þ
qL3

The optimization aims at minimizing W  min with respect to the geometric


b1 , 
variables  b2 and t subject to the constraints of (3b), (7), (8), (13), (14) which, in
522 Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532

nondimensional form, are given by:

P E L2
1 0 ðyieldingÞ ð18aÞ
EL2 ry A
 
P L4 K 2
1 0 ðflexural buckling about x-axisÞ ð18bÞ
EL2 Ix p
 
P L4 K 2
1 0 ðflexural buckling about y-axisÞ ð18cÞ
EL2 Iy p
    2  2 
P L2 K 2 ðG=E Þ J=L4 þ ðp=Kt Þ Cw =L6 þ ðp=K Þ r2x Ar20 =L6
1 0
EL2 r2x p ðG=E ÞðJ=L4 Þ þ ðp=K Þ2 ðCw =L6 Þ
ðtorsional-flexural bucklingÞ ð18dÞ
   
12 1  m2 P L2 b1 2
1 0 ðlocal buckling of webÞ ð18eÞ
k1 p2 EL2 A t
   
12 1  m2 P L2 b2 2
1 0 ðlocal buckling of flangeÞ ð18fÞ
k2 p2 EL2 A t
   
12 1  m2 P L2 b3 2
1 0 ðlocal buckling of lipÞ: ð18gÞ
k3 p2 EL2 A t
The above nonlinearly constrained optimization problem is solved by using the
SQP method with Matlab. It is emphasized that the optimization is local, in the
sense that it is carried out for C-sections having the same cross-sectional shape as
that illustrated in Fig. 1a (or Fig. 1b). The effect of varying cross-sectional shapes
on the minimum weight is not considered in this study. Also, for simplicity, the
effect of initial geometrical imperfections on column buckling are ignored.
3.2. Simple-minded optimization procedure

As a first approximation, the minimum weight of a compression structure may


be taken as that corresponding to the simultaneous occurrence of all failure modes.
Historically, this simple-minded concept has been dubbed the ‘‘naı̈ve optimization’’
[10], because the structure would be over-designed if there exists a residual margin
of safety with respect to the secondary failure mode(s) when primary failure
occurs. When the naı̈ve optimization procedure is adopted, a knockdown factor is
normally introduced to account for the strong nonlinear interaction amongst differ-
ent failure modes [19].
For an open-channel C-section under compression, the optimal design in the
elastic regime is dominated by flexural buckling stress about the y-axis, rey , tor-
sional-flexural buckling stress, rtf , and local buckling stress of the web, rlb1 .
According to the naı̈ve optimization, the minimum weight of the section is
obtained by solving the following simultaneous equations:
r ¼ rey ¼ rtf ¼ rlb1 ð19Þ
Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532 523

where r, rey , rtf and rlb1 are given by Eqs. (2), (7), (10) and (15), respectively. With
b3 ¼ b2 =5, Eq. (19) can be non-dimensionalized as

f1 ðt; 
b1 ;   Þ ¼ 0 ðflexural buckling about y-axisÞ
b2 ; P ð20aÞ

f2 ðt; 
b1 ;   Þ ¼ 0 ðtorsional-flexural bucklingÞ
b2 ; P ð20bÞ

f3 ðt; 
b1 ;   Þ ¼ 0 ðlocal buckling of webÞ
b2 ; P ð20cÞ
where f1 , f2 and f3 are dimensionless functions. For a prescribed non-dimensional
 , the non-dimensional geometric variables t, b1 and b2 are obtained by solv-
load P
ing Eqs. (20a–c) and then substituted into (17) to calculate the corresponding
dimensionless weight.

4. Results and discussion

The following material properties are used in all calculations: E ¼ 205 GPa,
m ¼ 0:3, q ¼ 7800 kg=m3 and ry ¼ 350 MPa. The predicted results on the mini-
mum weight of the C-section and its associated geometric dimensions from both
the naı̈ve optimization procedure and the SQP method will be presented, and
compared with those calculated by using the BS 5950 [18] and experimental
measurements.

4.1. Minimum weight

Fig. 2a plots the non-dimensional minimum weight W =qL3 of a C-section as


a function of the dimensionless load P=EL2 for selected values of the torsional
effective length factor Kt , with simply supported ends assumed for the section
(i.e. K ¼ 1). Compared with the non-linear optimization results, the predictions
from the naı̈ve optimization procedure show the correct trends, but are slightly
conservative. The minimum sectional weight is sensitive to the torsional
mobility of the section, decreasing with decreasing Kt , with the lowest minimum
weight achieved when torsion is completely restrained. In general, torsional
buckling dominates over other types of failure mode, and hence the torsional
buckling stress is the most important strength parameter to be considered in
the minimum weight design of an open-channel section within the elastic
regime.
With the torsional effective length factor kept at Kt ¼ 0:7, Fig. 2b plots the
dimensionless minimum sectional weight as a function of P=EL2 for different types
of support boundary condition: K ¼ 1 (simply supported), K ¼ 0:85, and K ¼ 0:7
(clamped). The effective length factor is seen to have a fairly weak influence on the
minimum weight, in contrast to that caused by the torsional effective length factor
Kt (Fig. 2a).
524 Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532

Fig. 2. Non-dimensional weight of an optimal C-section. (a) effect of torsional effective length factor Kt
for K ¼ 1; (b) effect of effective length factor K for Kt ¼ 0:7. Model 1: SQP mrthod; Model 2: Naı̈ve
optimization.

4.2. Optimal configuration of the C-section with lips

The non-dimensional geometric dimensions of the section associated with its


minimum weight (Fig. 2a) are presented in Fig. 3 as functions of P=EL2 for the
case of K ¼ 1. Although the predictions from the naı̈ve optimization are not
included in Fig. 3, it has been established that the thickness predicted by this
method is nearly twice that shown in Fig. 3.
For a given set of Kt and K, the ratio of web depth b1 to flange width b2 is a
constant (Table 1). If torsional buckling exists, the ratio b1 =b2 varies from 1.47 to
1.89 such that, for an optimally designed C-section with b2 ¼ 5b3 , its web depth is
always larger than its flange width. On the other hand, if torsion is completely
restrained, b1 =b2 ¼ 0:86 is obtained, i.e. the web depth is less than the flange
Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532 525

Fig. 3. Non-dimensional geometric dimensions of a C-section optimized by the SQP method for K ¼ 1:
(a) Thickness; (b) web depth; (c) flange length.

Table 1
Configurations of optimally-designed section with lips (b3 ¼ b2 =5)
Kt K b1 =b2
1 1 1.61 (1.58)a
0.7 1 1.89
0.5 1 1.47
No torsion 1 0.86
0.7 0.85 1.87
0.7 0.7 1.84
a
Prediction from the naïve optimization procedure.
526 Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532

length, indicating again that torsional buckling plays a profound role in the design
of open-channel steel sections.
4.3. C-section without lips

For a C-channel section without lips (b3 ¼ 0), the objective function is:
 
W ¼ W ¼  b1 þ 2
b2 t: ð21Þ
qL3
The constraints are identical to those given in (18) except that k2 ¼ 0:425 for the
unstiffened flange. The SQP method is used to minimize W .
For the case of K ¼ Kt ¼ 1, Fig. 4 compares the non-dimensional minimum
weight of a C-section without lips to that of a section with a fixed lip length of
b3 ¼ b2 =5. The two results are nearly identical, although the section with lips
weighs slightly less than the section without lips for a fixed load capacity.
The optimized dimensions of both sections are shown in Fig. 5 as functions of
the applied load. The thickness of the section without lips is seen to be significantly
larger than that of the section with lips, which may be attributed to its unstiffened
flanges in the absence of the lips. That is, to avoid local buckling, the unstiffened
flanges need to be thicker compared to the stiffened flanges in the section with lips.
On the other hand, because the minimum weight of the section remains (approxi-
mately) unchanged by the removal of the lips, the web depth and flange length of
the section without lips become considerably smaller than the corresponding

Fig. 4. Effect of lips on minimum weight for K ¼ 1 and Kt ¼ 1.


Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532 527

Fig. 5. Effect of lips on optimal sectional dimensions for K ¼ 1 and Kt ¼ 1: (a) Thickness; (b) web
depth; (c) flange length.

dimensions of the section with lips. The ratio of the web depth to flange length for
an optimal section without lips is 1.53.

4.4. Comparison with BS 5950

According to BS 5950 [18], the ultimate load capacity, Pc , of a column under


compression can be calculated as:

Pe Pcs
Pc ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð22Þ
/ þ /2  Pe Pcs
528 Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532

Here, Pe is the smaller of Pey and Ptf , Pcs is the local buckling load (Appendix A),
Pcs þð1þgÞPe
/¼ 2 , and
8
< 0; if aKL=r  20
g¼ ð23Þ
:
0:002ðaKL=r  20Þ; if aKL=r > 20
where r denotes the radius of gyration of the gross cross-section corresponding to
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pe and a ¼ Pey =Ptf .
When the section length L is fixed and the values of K and Kt are specified, the
optimal section dimensions for a prescribed load capacity P can be found directly
from Fig. 3. The optimal dimensions of a few selected sections of fixed length L ¼
2400 mm are listed in Table 2 for P ¼ 10, 20, 30 and 40 kN, respectively. Upon
substituting these dimensions into (22), the load carrying capacity of the optimized
section calculated according to BS 5950 [18] is obtained. The results are shown in
Table 2 as well as in Fig. 6, and compared with those obtained with the SQP opti-
mization procedure. The load capacity based on BS 5950 [18] is seen to be consist-
ently smaller than the SQP prediction, and hence is somewhat conservative. It
should be indicated here that in BS 5950 [18] Kt ¼ 0:7 is recommended, so in Fig. 6
only one curve is plotted, corresponding to the pinned end conditions.
4.5. Comparison with test results

Three different C-section with lips (Table 3) were fabricated and tested. Details
of the experimental procedure are reported in Tian et al. [20–23]. During tests,
pined end conditions for studs were arranged at both ends of the stud. The steel
grade of the tested studs is identical to that used in the optimal calculations.
CS9015 (Table 3) is the traditional wall stud used in construction. From our test
results and calculation, it is found that its failure load under uniaxial compression
is less than 20 kN. For an individual stud (2400 mm long), the industry wishes to
push for a stud load capacity of about 30–40 kN without increasing the raw
material content. Based on the present optimization results, the ratio of web depth
to flange length, b1/b2, should be a constant in the range of 1.47 to 1.89, depending
on the end conditions selected for torsional buckling. To design a the new stud, the
web depth (90 mm) is kept as the same as CS9015, and the ratio of web depth to
flange length is fixed at 1.5. According to Table 2 and the typical steel gages used

Table 2
Load capacity of minimum weight C-channel sections
P (N) W (kg) b1 (mm) b2 (mm) b3 (mm) t (mm) P (N) BS 5950
10,000 2.50 76.4 40.5 8.1 0.769 8591
20,000 3.78 87.8 46.4 9.3 1.01 16,441
30,000 4.82 95.2 50.3 10.1 1.19 23,974
40,000 5.73 101 53.2 10.6 1.34 31,168
L ¼ 2400 mm; K ¼ 1; Kt ¼ 0:7:
Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532 529

Fig. 6. Minimum weight predicted from the SQP method compared with that calculated by using
BS 5950 for K ¼ 1.

in the industry, two gages, 1.2 mm and 1.5 mm are selected. Finally, two new
studs, CDS9015 and CDS9012 were designed and fabricated, and their dimensions
are listed in Table 3. For each stud, three identical specimens were tested till failure
under axial compression, and the results summarized in Table 4 and plotted in

Table 3
Dimensions of tested studs
Stud type b1 (mm) b2 (mm) b3 (mm) t (mm) L (mm)
CS9015 88.5 37.5/40.5 7.65 1.5 2400
CDS9015 88.5 58.5 11.25 1.5 2400
CDS9012 88.8 58.8 11.4 1.2 2400

Table 4
Failure load of tested studs
Stud typea Test failure load (kN) Average failure load (kN)
CS9015 18.7, 19.2, 17.5 18.5
CDS9015 43.0, 42.9, 40.9 42.3
CDS9012 29.4, 29.5, 29.7 29.5
a
For each stud type, three nominally identical specimens were tested.
530 Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532

Fig. 6. It can be seen that the performance of the two new studs is much superior
than that of the non-optimized CS9015 stud. For example, studs CS9015 and
CDS9012 have similar weight, but the failure load of CDS9012 is 50% higher than
that of CS9015.

5. Conclusions

The minimum weight of cold formed steel C-sections under axial compression
are obtained using the SQP method and the naı̈ve optimization procedure. For a
given set of effective length factor K and effective torsional length factor Kt , the
ratio of web depth to flange length b1 =b2 is found to be constant. If torsional buck-
ling exists, the corresponding critical stress is the dominant parameter governing
failure. The weight of an optimal section decreases with decreasing K as well as
decreasing Kt , although the effect of K is relatively weak. If the section has no lips,
its geometric dimensions are considerably different from those of a section with
lips, even though there is no difference between their weights. The naı̈ve optimiza-
tion procedure is simple, but it underestimates the load carrying capacity of the
section, and hence needs to be used with caution. Similarly, for a prescribed com-
pressive load, the sectional weight calculated by using the BS 5950 is about 10–30%
bigger than that predicted from the SQP method.
New C-sections with lips were designed and fabricated based on the optimization
procedure. The experimentally measured axial load capacity of the new sections
agrees well with the prediction. The optimized section achieves a axial load
capacity 50% larger than that of the non-optimized section currently used by the
industry, at the same weight.

Acknowledgements

This work is supported partially by the UK Engineering and Physical Scientific


Research Council (EPSRC GR/M95936), and partially by the Banro Holdings
Ltd., UK. The authors wish to thank Dr. J. Wang of Cambridge University and
Dr. J. Evans of Banro Holdings Ltd for constructive discussion and assistance during
testing.

Appendix A. Load capacity after initial local buckling

In the absence of local buckling, the stress distribution over the cross-section of
a C-section under axial compression is uniform. When the axial stress exceeds the
local bucking strength of the web, the web can still support the load, but the stress
distribution is no longer uniform, as shown schematically in Fig. A1. Under such
circumstances, the effective width idea may be used to simplify the calculation of
the load capacity of the web. It is assumed that, after the occurrence of local buck-
ling, the compressive load is supported by a portion of the web over which the
Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532 531

stress distribution is uniform, with the rest of the web stress free. The web depth
beff over which the stress is non zero is termed the effective width of the web
(Fig. A1). The same concept can be used to define the effective length of the flange
or the lip.
The formula to calculate the effective width (or length) is mainly based on
experiments. For a web stiffened on both sides by the flanges (Fig. 1), the BS 5950
(BSI 1998) suggests that
2 ( )4 30:2
1=2
beff r
¼ 41 þ 14 0:35 5 : ðA1Þ
b rlb1

On the other hand, if the web is unstiffened, its effective width, beu , may be esti-
mated as (BSI 1998):

beu ¼ 0:89beff þ 0:11b: ðA2Þ

For the C-channel section shown in Fig. 1a, the effective width of each element,
namely, web, flange and lip, can be calculated separately, and then the effective
area Aeff of the section can be determined as:

Aeff ¼ ðbeff1 þ 2beff2 þ 2beff3 Þt ðA3Þ

where beff 1 is the effective web depth, beff 2 is the effective flange length, and beff3 is
the effective lip length. Finally, the local buckling load, Pcs , of the section is
obtained as:

Pcs ¼ rAeff : ðA4Þ

Fig. A1. Actual stress distribution.


532 Y.S. Tian, T.J. Lu / Thin-Walled Structures 42 (2004) 515–532

References
[1] Wagner H. Remarks on airplane struts and girders under compressive and bending stresses. Index
values. NACA Tech Memo 1992;500:241–7.
[2] Zahorski A. Effects of material distribution on strength of panels. J Aeronaut Sci 1944;11(3):
247–53.
[3] Schuette EH, Charts for the minimum weight design of 24S-T aluminium alloy flat panels with
longitudinal Z-section stiffners. NACA Technical Report. 1945;827.
[4] Farrar DJ. The design of compression structures for minimum weight. J Roy Aeronaut Soc
1949;53:1041–52.
[5] Shanley FR. Weight-strength analysis of aircraft structures. New York: McGraw-Hill; 1952.
[6] Catchpole EJ. The optimum design of compression surface having unflanged integral stiffeners.
J Roy Aeronaut Soc 1954;58(11):765–8.
[7] Gerald G. Minimum weight analysis of compression structures. New York: New York University
Press; 1956.
[8] Cox HL. The design of structures of least weight. Oxford: Pergamon Press; 1965.
[9] Weaver PM, Ashby MF. Material limits for shape efficiency. Prog Mat Sci 1997;41:61–128.
[10] Budiansky B. On the minimum weights of compression structures. Int J Solids Struct 1999;36:
3677–708.
[11] American Iron and Steel Institute. Specification for the design of cold-formed steel structural mem-
bers. Washington DC, USA, 1996.
[12] Seaburg PA, Salmom CG. Minimum weight design of light gauge steel members. J Struct Div
ASCE 1971;97(1):203–22.
[13] Dinovitzer AS. Optimization of cold formed steel C-sections using standard Can/CSA-S136-M89.
Can J Civ Eng 1992;19:39–50.
[14] Adeli H, Karim A. Neural network model for optimization of cold-formed steel sections. J Struct
Eng ASCE 1997;123(11):1535–43.
[15] Al-Mosawi S, Saka MP. Optimum shape design of cold-formed thin-walled steel sections. Adv Eng
Software 2000;31:851–62.
[16] Rhodes J. Design of cold formed steel members. London: Elsevier Applied Science; 1991.
[17] Yu WW. Cold-formed steel design. New York: John Wiley and Sons; 2000.
[18] British Standard Institution. British Standard: structural use of steelwork in building, Part 5, Code
of practice for design of cold-formed section, BS 5959, 1998.
[19] Thompson JMT, Lewis GM. On the optimal design of thin-walled compression members. J Mech
Phys Solids 1972;20:101–9.
[20] Tian YS, Wang J, Lu TJ, Barlow CY, Evans J. An experimental study on the load carrying
capacity of cold-formed steel studs and panels. 16th International Specialty Conference on Cold-
formed Steel Structure. October 17 and 18, 2002, Orlando FL, USA. 2002.
[21] Tian YS, Wang J, Lu TJ, Barlow CY. An experimental study on the axial behaviour of cold
formed steel wall studs and panels. Thin-Walled Structures 2003 (in press).
[22] Tian YS, Wang J, Lu TJ. Racking strength and stiffness of cold-formed steel wall frames. J Con-
struct Steel Res 2003 (in press).
[23] Tian, Y.S., Optimal Design of Cold-formed Steel Sections and Panels. Ph.D. thesis, Cambridge
University Engineering Department, 2003.

Das könnte Ihnen auch gefallen