www.elsevier.com/locate/tws
Abstract
The companion paper has presented results of elevated temperature tests on 52 cold-formed
thin-walled channels under compressive load. This paper presents the results of theoretical
studies using a number of different calculation tools, these including simple design calculations
based on modifying a few current design methods and a commercial finite element package
ABAQUS. The design methods considered in this paper include the British standard BS5950
Part 5, Eurocode 3 Part 1.3 and the American Specification AISI. Modifications of the current
design equations are made to enable them to include distortional buckling, the effects of service
holes and elevated temperatures. To enable BS5950 Part 5 and Eurocode 3 Part 1.3 to predict
the ultimate strength of thin-walled columns with a service hole, the AISI (1996) design
method is introduced. To extend the capacity of these design methods to deal with distortional
buckling failure mode, the method of Young, Kwon and Hancock for calculating distortional
buckling capacity is introduced in these codes. Finally, the ambient temperature design
methods are modified to take into account changes in the strength and stiffness of steel at
elevated temperatures. From extensive comparisons between the results of tests, code predic-
tions and numerical analyses, it may be concluded that by adopting the aforementioned modi-
fications, the current code design methods can be easily modified to consider these advanced
modes of behaviour.
For finite element analyses, both geometrical and material non-linearities are taken into
account. The high temperature stressstrain relationships of steel are determined according to
Eurocode 3, Part 1.2 or Outinen et al.
2003 Elsevier Science Ltd. All rights reserved.
Corresponding author. Tel.: +44 161 2008968; Fax: +44 161 2004646.
E-mail address: yong.wang@man.ac.uk (Y.C. Wang).
0263-8231/03/$ - see front matter 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0263-8231(03)00003-X
572 M. Feng et al. / Thin-Walled Structures 41 (2003) 571594
Keywords: Local buckling; Distortional buckling; Design methods; Effective width; Fire resistance; Mech-
anical properties; Elevated temperature; Service hole
1. Introduction
Fig. 2. Strength retention factors at the 0.5% and 1.5% total strains (from Lawson [8]).
Fig. 3. Limiting temperatures (C) of different types of cold-formed thin-walled steel structures.
M. Feng et al. / Thin-Walled Structures 41 (2003) 571594 575
Eurocode 3 Part 1.2 [5] has recently introduced cold-formed, thin-walled steel
structures. By introducing the strength retention factors in Fig. 1 into the design
equations in Eurocode 3 Part 1.2 for hot-rolled steel structures, the scope of Eurocode
3 Part 1.2 is extended to include cold-formed thin-walled steel structures. The AISC
method [14] does not consider cold-formed thin-walled steel structures in fire.
Thus, the first objective of this paper is to use the experimental results in the
companion paper to assess and improve current design methods. Since these tests
were on short columns under uniform temperatures, this paper will only address
local and distortional buckling. Future studies will also consider overall buckling,
the interaction between different modes of buckling and non-uniform temperature
distributions.
In all design methods, local buckling is dealt with using the effective width method
and the effective area of a column cross-section is assumed to be the summation of
the effective width multiplying the thickness of all individual plates, i.e.
Pu fy beff,iti (1)
where fy is the yield stress of steel, beff,i and ti are the effective width and thickness
of the ith plate of the cross-section.
The Winter [16] expression has been accepted in the AISI [14] method and in
Eurocode 3 Part 1.3 [4] for the analysis of local buckling. The effective width beff
of a plate is:
Y (10.22 Y )
beff pcr pcr
(2)
b s s
in which b is the plate width, pcr is the elastic critical buckling stress of the plate
and Ys is the maximum edge stress of the plate, which may be taken as the yield
stress of steel fy. The elastic critical buckling stress pcr of a plate depends on such
factors as the boundary condition along the plate edges and stress distributions in
the plate, which is given as:
12(1 n2)
b
pcr (3)
t p2Ek
in which t is the plate thickness, E is the elastic modulus of steel, v is the poissons
ratio and k is a buckling factor. The buckling factor depends on the edge condition
and stress distribution of the plate. Eurocode 3 Part 1.3 and AISC give slightly
different values of k for uniformly compressed element with an edge stiffener.
A slightly different approach to deal with local buckling is adopted in the British
Standard, BS5950 Part 5 [3], which gives:
p 0.35) ]
beff fc
[1 14( 4 0.2
(4)
b cr
where, fc is the compression stress on the effective element, which may be taken as
576 M. Feng et al. / Thin-Walled Structures 41 (2003) 571594
the yield stress of steel fy and pcr is the elastic critical buckling stress of the plate,
which can be calculated according to eq. (3).
Apart from the design method of Lawson [8], all other methods need some modi-
fications to enable them to analyse the test results in the companion paper. The
required modifications are as folows.
f f
beff,de sde sde
(10.22 ) for l0.673 (6)
b y y
where, beff,de is the effective width of the flange plate for distortional buckling and
the elastic distortional buckling stress sde may be calculated according to Lau and
Hancock [7]. The flange plate slenderness for distortional buckling is calculated
using:
s
fy
l (7)
de
Of the three different design methods, only the AISI method [14] considers plates
with a service hole. Referring to Fig. 5, the AISI recommendations are:
According to Ranby [12], when predicting the column failure load under a uniform
elevated temperature, the reduced 0.2% proof stress and the reduced elastic modulus
should be used.
Thus, the specific modifications for different design method are as follows:
For using BS 5950 Part 5: distortional buckling, service hole and elevated tem-
peratures.
Fig. 6. Comparison between test failure loads and modified BS5950 Part 5 predictions.
For using Eurocode 3 Part 1.3: distortional buckling, service hole and elevated
temperature.
For using the AISI method: distortional buckling and elevated temperatures.
The various design methods with aforementioned modifications have been used
to calculate the strength of the short columns reported in the companion paper. Figs
68 show comparisons between the test failure loads with the calculated values. Each
figure compares the test results with calculations using the retention factors for the
Fig. 7. Comparison between test failure loads and modified Eurocode 3 Part 1.3 predictions.
M. Feng et al. / Thin-Walled Structures 41 (2003) 571594 579
Fig. 8. Comparison between test failure loads and modified AISI predictions.
0.2% proof stress of Outinen [911] under steady state test condition and the strength
retention factors of Eurocode 3 Part 1.2 [5]. It can be found that all predicted failure
loads are close to the test failure loads, suggesting the suitability of using the three
design methods as basis of calculations and the validity of the previously described
modifications. Furthermore, Fig. 9 suggests that there is very little difference in
results obtained from the three different methods. Since the tests were carried out
under steady state condition, it appears that the steady state strength retention factors
of Outinen [911] give slightly better agreement between the calculated and test
failure loads.
In the companion paper, it was mentioned that ratios of failure loads of different
Fig. 9. Comparison between test failure loads and predictions of modified BS5950 Part 5, Eurocode 3
Part 1.3 and AISI, using Outinen properties.
580 M. Feng et al. / Thin-Walled Structures 41 (2003) 571594
types of columns at the same elevated temperature were close. The ratio of the failure
load of a column at elevated temperature to that at ambient temperature is the load
ratio of the column and the test temperature is the limiting temperature. Thus, the
closeness of different test strength values at the same test temperature suggests that
the limiting temperature method may be used. However, only the test temperature
550 C is within the range of application of Lawsons method [8]. For this particular
temperature and the case of column in walls, which is appropriate for the tested
columns, Lawsons method gives a load ratio of 0.425. Table 1 indicates that this
appears to be a good approximation of the average value of test results.
Finite element modelling using a commerical program ABAQUS [1] has also been
performed. The purpose of this study is to check the suitability of this software and
to determine appropriate modelling parameters.
All cross-sections were modelled using the measured centre line dimensions. One
rigid plate was attached to each end of the column to simulate the end support. The
rigid plate was only allowed to deform in the axial direction with all other modes
of deformation prevented.
Initial imperfections used for the non-linear FE analyses were scaled from eigen-
vectors obtained from the results of linear eigenvalue analyses. Since the distortional
buckling mode is in general more sensitive to initial imperfections than the local
buckling one, initial imperfections have been chosen so that the final failure mode
is consistent with the observed failure mode even though in some cases the elastic
distortional buckling stress may be higher than the elastic local buckling stress. For
example, Table 2 shows initial deflection profiles for each specimen used in FE
analyses at ambient temperature. The maximum initial deflection was assumed to be
the same as the thickness of the cross-section.
The results of tensile coupon tests indicate that the stressstrain relationship of
cold-formed steel exhibits a gradual yielding behaviour followed by a considerable
period of strain hardening. Hence the elastic-plastic model of hot-rolled steel cannot
be used. Fig. 10 shows the s curve used in FEM analyses based on the tensile
coupon test results. In this model, the elastic stressstrain behaviour is represented
by a linear segment up to a limit stress fp = 0.7fy, where fy is the 0.2% proof stress.
The slope of this linear segment equals to the measured elastic modulus. The gradual
yield behaviour is idealised by using a bilinear representation with tangent modulus
E1 and E2 between the proportional limit fp and the yield strength fy with an inter-
mediate point at a stress of 0.875fy. The strain-hardening behaviour is represented
by a linear segment with a tangent modulus E3. The value of E1, E2 and E3 are
80 000 N/mm2, 20 000 N/mm2 and a proportion of the initial elastic modulus.
The stressstrain relationship at high temperatures used in finite element analyses
Table 1
Comparison between load ratios of tests and Lawson [8]
Section type Temp. (C) Test ID Test results Lawsons method [8] Load ratio at
fire limit
Table 2
Initial imperfection shape and magnitude for ambient temperature analysis
are based on the models in Eurocode 3 Part 1.2 and Outinen [9,10], but modified
by the appropriate measured yield stress and elastic modulus. Eurocode 3 Part 1.2
gives mathematical equations for the stressstrain relationships of steel at elevated
temperatures and can be directly used in this study if the yield stress is based on
the stress at 2% total strain. Since the retention factors of 0.2% proof stress have
only been given in Eurocode 3 Part 1.2, a suitable modification has been done in
order that the mathematical equations for the stressstrain relationships of steel at
elevated temperature in Eurocode 3 Part 1.2 can be used directly. The modification
M. Feng et al. / Thin-Walled Structures 41 (2003) 571594 583
is to select a value for the stress retention factor at 2% total strain by trial and error
so that the retention factors for the 0.2% proof stress obtained from the resulting
stressstrain relationship equal to the Eurocode 3 Part 1.2 value for the 0.2% proof
stress. To use the stressstrain relationships of Outinen [9], since the ambient tem-
perature strengths of steel from this study are different from those used by Outinen,
modifications are necessary. The final stressplastic strain curves to be used in
ABAQUS at different temperatures are given in Fig. 11. To obtain these curves, the
Fig. 11. Input stressstrain relationships according to Outinen [9] (a) for lipped channel 100 54
15 1.2 (b) for lipped channel 100 56 15 2 (c) for unlipped channel 104 63 1.5.
584 M. Feng et al. / Thin-Walled Structures 41 (2003) 571594
plastic strains were extracted from the stresstotal strain curves reported in Outinen
[9]. At the same plastic strain, the stress of steel used in ABAQUS simulations was
obtained by multiplying the stress in the Outinen stressstrain curve by the ratio of
the measured 0.2% proof stress to the 0.2% proof stress of Outinen.
Fig. 12. Loadaxial displacement curves for lipped channel 100 54 15 1.2 with different element
sizes.
M. Feng et al. / Thin-Walled Structures 41 (2003) 571594 585
As far as the type of elements is concerned, Fig. 13 indicates that these tests may
be modelled by using ABAQUS elements S8R, S8R5 and S4R. S8R is eight-noded
doubly curved thin shell element using six degrees of freedom per node; S8R5 is
8-noded doubly curved thin shell element using five degrees of freedom per node
and S4R is 4-noded doubly curved thin shell element using six degrees of freedom
per node. Obviously, when using S4R elements, more elements should be used.
In ABAQUS, as in many other FE analysis programs, integration is carried out
numerically by summing up contributions at a number of integration points. The
more integration points are used, the higher the accuracy of simulation and compu-
tational demand. Fig. 16 shows that the difference in results of using five and three
integration points in the thickness direction of the element is less than 5%. Using
more integration points than five would increase computational effort without gain-
ing accuracy.
Fig. 13. Load-axial displacement curve for lipped channel 100 54 15 1.2 with different element
types.
586 M. Feng et al. / Thin-Walled Structures 41 (2003) 571594
shape close to the observed failure mode. This may be different from the deflected
shape of the lowest eigenvalue. Numerical analysis of lipped channel 100 56
15 2 has been chosen for this sensitivity study since predominantly distortional
buckling failure mode was observed for this section, but the lowest eigenvalue is
from local buckling. Nevertheless, simulation results in Fig. 14 suggest that neither
the loadaxial displacement response nor the ultimate strength of the column was
affected by the profiles of initial deflections. In addition, Fig. 15 compares the test
results with simulations using different maximum imperfections. It can be seen that
the smallest imperfection, at 1% of the web thickness, gives a very sharp loss of
stiffness when approaching the failure load of the column of about 64.41 kN. As
the magnitude of initial imperfection is increased, the predicted column behaviour
becomes smoother, as observed in the test. Nevertheless, the predicted ultimate loads
are in general close to each other, less than 5% different, and also agree well with
the experimental results.
Thus, depending on the input initial deflection profile, any failure mode may be
obtained. The loadstrain and loadlateral deflection responses would have been dif-
ferent. However, the column strength would have not been noticeably affected. Also,
the maximum initial imperfection may be taken as the thickness of the cross-section.
Fig. 14. Axial loaddisplacement curves for lipped channel 100 56 15 2 (without hole) with
different initial imperfection curves.
M. Feng et al. / Thin-Walled Structures 41 (2003) 571594 587
Fig. 15. Load-strain curves for lipped channel 100 54 15 1.2 with different initial imperfections.
Fig. 16. Load-lateral displacement curve for lipped channel 100 56 15 1.2 (without hole) with
different strain hardening modulus.
588 M. Feng et al. / Thin-Walled Structures 41 (2003) 571594
To check the accuracy of the ABAQUS model, all the tests have been simulated.
Based on the results of the sensitivity studies, the following values have been used
in all ABAQUS simulations:
Element type: S8R5 (eight-noded doubly curved thin shell element using five
degrees of freedom per node).
1
Element size: the smaller of 15mm or lip width;
2
Stressstrain relationship: as in Fig. 10 for ambient temperature simulations; based
on Eurocode 3 Part 1.2 or Fig. 11 for elevated temperature simulations.
Integration points: 5.
Profile of initial deflection: the lowest eigenvalue that gives a deflection shape
similar to the observed failure mode.
Maximum value of initial deflection: thickness of cross-section.
Fig. 17. Loadstrain curves for lipped channel 100 54 15 1.2 (no hole) with different elastic modulus and different yield stress.
589
590 M. Feng et al. / Thin-Walled Structures 41 (2003) 571594
Fig. 18. Loadaxial displacement curves for lipped channel 100 54 15 1.2 with different pro-
portional stresses.
appear to be more accurate. However, the results of simulations using the Eurocode
model are also acceptable and may be more suitable for realistic structures which
tend to be under transient state condition in fire.
7. Conclusion
This paper has presented the results of a theoretical study of the strength of cold-
formed thin-walled steel short columns under uniform high temperatures, using a
number of design calculation methods and a commercial finite element package
ABAQUS. All analysis results have been compared to the test results. The following
conclusions may be drawn:
channel 100
56 15 2 without
hole
Lip2d1 108.97 As above 125.3 As above 125.3 As above
250 Lip2b225 123.69 As above 109.8 As above 113.8 As above
Lip2b325 123.23 As above 109.8 As above 113.8 As above
400 Lip2C140 101.87 As above 87.18 As above 93.65 As above
Lip2C240 101.56 As above 87.18 As above 93.65 As above
(continued on next page)
591
Table 3 (continued) 592
Acknowledgements
References
[1] ABAQUS Standard Users Manual. Volumes IIII. Version 5.8.1.5 Hibbit, Karlsson & Sorensen
Inc, 1998.
[2] Ala-Outinen T, Myllymaki J. The local buckling of RHS members at elevated temperatures, VTT
Research Notes 1672, Technical Research Centre of Finland, VTT, Espoo, Finland, 1995.
[3] British Standard Insitituion (BSI 1987), BS5950: Structural Use of Steelwork in Building. Part 5:
Code of practice for design of cold-formed sections. British Standards Institution, London.
[4] CEN. Eurocode 3: Design of Steel Structures, Part 1.3: General rules, Supplementary rules for cold
formed thin gauge members and sheeting, Commission of the European Communities, Brussels,
1993.
[5] CEN. Pre EN 1993-1-2, Eurocode 3; Design of steel structures. Part 1.2: General rules, Structural
fire design. British Standards Institution, 2001.
[6] Kirby BR, Preston RR. High temperature properties of hot-rolled structural steels for use in fire
engineering studies. Fire Safety Journal 1988;13(1):2737.
[7] Lau SCW, Hancock GJ. Distortional buckling tests of cold-formed columns. J Struct Engrg ASCE
1987;113(5):106378.
[8] Lawson RM. Building design using cold formed steel sections: Fire protection, Publication P129.
The Steel Construction Institute, 1993.
[9] Outinen J. Mechanical properties of structural steels at elevated temperature. Licentiate thesis, Hel-
sinki University of Technology, Laboratory of Steel Structures, Espoo, 1999.
[10] Outinen J. A study for the development of the design of steel structures in fire condition. In: Proceed-
ings of the First International Workship on Structures in Fire, Copenhagen. 2000. p. 26981.
[11] Outinen J. Effect of high temperature on mechanical properties of cold-formed structural steel. In:
Puthli R, Herion S, editors. Tubular Structures IX. 2001.
[12] Ranby A. Structural fire design of thin walled steel sections. J Construct Steel Res 1998;46:3034.
[13] Sidey MP, Teague DP. Elevated temperature data for structural grades of galvanized steel. British
Steel (Welsh Laboratories) Report, 1988.
[14] AISI. Specification for the design of cold-formed steel structural members, 1996 edition, American
Iron and Steel Institute, 1996.
[15] Wang YC, Davies JM. Design of thin-walled steel channel columns in fire using Eurocode 3 Part
1.3. In: Proceedings of the First International Workship on Structures in Fire, Copenhagen. 2000.
p. 18193.
[16] Winter G. Strength of thin steel compression flanges. Cornell Univ Eng Exp Stn Reprint No. 32,
1947.
[17] Young B, Kwon TB, Hancock GJ. Tests of cold-formed channels with local and distortional buck-
ling. J Struct Engrg ASCE 1992;117(7):1786803.