You are on page 1of 19

J. Construct. Steel Res. Vol. 36, No. 2, pp.

101-119, 1996
Copyright 1996 ElsevierScienceLtd
Printed in Great Britain.All rights reserved
0143-974X/96 $15.00+0.00
0143-974X(95)00014-3
ELSEVIER

The Lateral-torsional Buckling of Unrestrained


Steel Beams in Fire

C. G. Bailey, a I. W. Burgess a & R. J. P l a n k b


a Depm'tment of Civil and Structural Engineering, University of Sheffield, Sheffield,
S1 4DU, U K
School of Architectural Studies, University of Sheffield, Sheffield, S1 4DU, UK
(Received 10 October 1994; revised version received 27 February 1995;
accepted 21 April 1995)
ABSTRACT

A three-dimensional computer model which is capable of predicting the structural


behaviour at elevated temperatures of skeletal steel frames has been used to
invest;igate the ultimate behaviour of uniformly heated unrestrained beams. A series
of different sections and spans has been studied for different loading patterns and
load ratios. The predicted results indicate failure by lateral-torsional buckling in all
cases. A comparison has been made with the limiting temperatures which are given
in BS5950 Part 8, and in most cases the model predicts lower failure temperatures
tha does the code. This is in contrast to the extremely accurate limiting temperatures
which the Code specifies for restrained beams. A similar finding is obtained when
the results are compared with limiting temperatures obtained from EC3 Part 1.2. It
can be seen that failure temperatures generally depend on the ambient-temperature
ratio of lateral-torsional buckling resistance moment to in-plane moment capacity,
with t~e last slender cases failing at the lowest temperat,',es for any given load ratio.

NOMENCLATURE

For design to BS5950 Parts I and 8


MCx Moment capacity of section about major axis
Mb Buckling resistance moment (lateral torsional)
n Slenderness correction factor
m Equivalent uniform moment factor
Le Effective length
Mf Applied moment at the fire limit state
R Load ratio
Perry coefficient (imperfection parameter)
2 Slenderness ratio

I01
102 C. G. Bailey, I. IV. Burgess, R. J. Plank

For design to EC3 Parts 1.1 and 1.2


Me,Rd Design moment of resistance of the cross-section
Mb,Rd Design buckling resistance moment (lateral-torsion)
XLT Reduction factor for lateral-torsional buckling
Factor based on the class of cross-section
Wply Plastic section modulus about major axis ( Y - Y )
f, Yield strength
7ml Partial safety factor
2-LT Non-dimensional slenderness
Mer Elastic critical moment for later torsional buckling
E Modulus of elasticity
G Shear modulus
Iz Second moment of area about minor axis
It Torsion constant
Iw Warping constant
k, kw Effective length factor
Cl Load case factor
(XLT Imperfection factors
a,Cl" Critical steel temperature
Rfi,d,t Design value of resistance at time t
Efi,d Design effect of the action in fire situation
/10 Degree of utilisation at time t = 0
Mb,fi,t,Rd Design buckling resistance moment in fire at time t
ky,0,max Reduction factor for yield strength of steel at maximum
steel temperature 0max

1 INTRODUCTION

To comply with regulatory requirements I all buildings require a minimum


degree of fire resistance, primarily to ensure that premature collapse will
not occur during evacuation and fire fighting. The stated periods of fire
resistance quoted often exceed those required for life-safety alone, due to
the provision for property protection. In the case of steel-framed buildings
the most familiar method of providing this fire resistance period is based
on the use of some form of insulating material to keep the steel at a
sufficiently low temperature to retain enough strength and stiffness for
structural safety. In traditional design procedures the maximum permis-
sible temperature of a steel member tended to be prescribed at values
based on limiting the assumed reduction of yield stress so that the effect
The lateral-torsional bucklino of unrestrained steel beams in fire 103

of fire was simply to remove the built-in safety margin on permissible


stresses. This approach fits reasonably well with the philosophy of permis-
sible-stress design, but is inadequate in that it does not take account of
load lewfls or the fact that stress-strain curves become highly curvilinear
at elevated temperatures so that the concept of a yield stress itself becomes
very arb:itrary. The recent design codes BS5950 Part 82 and EC3 Part 1.23
have introduced procedures for predicting the behaviour of steel members
at elevated temperatures which encourage designers to treat fire as one of
the basic design limit states. They allow different methods of predicting
member behaviour based either on testing or on calculation. The simplest
calculation approach presented is known as the Limiting Temperature
Method, in which the maximum permissible temperature of a member is
based oil its type, its loading characteristics and load ratio. This method
presents little difficulty in implementation, and is clearly the approach
which will recommend itself to practising engineers.
The development of limiting temperatures in BS5950 Part 8 has been
based on a large number of fire tests 4 and supplementary computer model
predictions. For members in bending, standard fire tests 5 have been
carried out on beams which support concrete and composite floors, the
major variables being applied load, structural configuration and beam
size. A more comprehensive experimental study of the parameters which
affect beams' response is hardly feasible due to the very high cost of
structural furnace testing and the size limitations of typical furnace. It is
not surprising that the beams tested have represented the systems in most
common use; floor beams carrying precast slabs on the top flange or on
shelf-angles, slim-floors with the slab filling the space between the flanges,
and composite beams using metal decking and slabs tied through shear
connectors to the top flange. Not only are these systems the most common
in current steel building construction, but all have full compression flange
restraint, and therefore present no threat to the integrity of the test furnace
which n:tight be caused by lateral buckling failure.
In the context of present-day building design beams without lateral
restraint to the compression flange may be considered as of secondary
importance. However, in some situations, such as industrial buildings and
plant areas, beams may not have continuous lateral restraint. This also
applies in some office construction systems, notably the 'Parallel Beam
Approach '6 in which the primary (spine) beams are unrestrained and
support composite secondary (rib) beams on their top flanges, thus
providing service spaces in both directions within the depths of these
beams, lit is necessary to calculate limiting temperatures for the unprotec-
ted beams either to determine the required protection thickness or, as part
104 C. G. Bailey, I. W. Burgess, R. J. Plank

of a fire engineering approach, to set the required section size, load ratio
and possibly partial protection. BS5950 Part 8 deals with unrestrained
beams firstly by using a load ratio based on lateral-torsional buckling
failure at ambient temperature, and then by giving limiting temperatures
reduced by about 65C from those for restrained beams. An underlying
problem is that the limiting temperatures for beams in furnace testing are
based on a definition of failure when the maximum in-plane deflection
reaches span/20, or a maximum rate of deflection when span/30 is
exceeded. 5 It has been shown at ambient temperature 7 that the lateral
instability of beams is a complex problem, which will become even more
complicated at elevated temperatures due to degradation of strength and
stiffness, and to the progressively more curvilinear nature of the stress-
strain curves. In the work reported here computer modelling has been
used to investigate the failure temperatures of unrestrained beams under
various loading conditions and comparisons have been made between
these and the limiting temperatures given by the Code.
The finite element software used is based on a non-linear spread-of-yield
program originally written by E1-Zanaty and Murray 8 to study the
two-dimensional behaviour of steel frames at ambient temperature. This
was firstly extended by Saab and Nethercot 9 to include elevated tempera-
ture effects, and has more recently been extensively re-formulated by
Najjar 1 and Najjar and Burgess 11 to analyse three-dimensional frame
behaviour, including warping behaviour, in fire conditions. The program
uses two-noded one-dimensional line elements with eight degrees of
freedom in local coordinates. Every effort has been made in the formula-
tion to retain higher-order terms, resulting in a very accurate treatment of
geometric non-linearity, together with the capacity to allow for consider-
able variation in cross-sectional properties. It has been shown to be very
accurate in large-deflection problems, both at ambient and elevated
temperatures. In the course of the present work this software has been
developed further to include flexural shear stresses, so that it is now
capable of accurately predicting lateral-torsional buckling effects, for loads
placed at any level on a cross-section, and the results presented below were
largely generated as part of this development process. Other extensions,
which are not relevant to the subject matter of this paper, include the
introduction of semi-rigid connection characteristics which degrade at
elevated temperatures, a proper treatment of strain reversal in cooling and
the inclusion of slabs within the building model. The overall objective of
the developments has been to facilitate the modelling of full steel-framed
buildings subject to local and spreading fire scenarios in order to produce
more comprehensive guidance both for fire engineering design and for
assessment of repairability after a fire.
The lateral-torsional buckling of unrestrained steel beams in fire ! 05

2 VALIDATION OF THE ANALYSIS AT


AMBIENT TEMPERATURE

The program's capacity to predict lateral-torsional buckling has been


checked at ambient temperature against theoretical bifurcation solutions
in the elastic range, and against experimental and computational results
in the inelastic range. For comparison with theoretical bifurcation sol-
utions 12 beams were assumed to remain elastic, to be initially unstressed
and perfectly straight. The critical loads obtained, using four elements to
model the behaviour of simply-supported beams of various serial sizes and
spans, with a point load acting at the shear centre and at mid-span, are
compared against the theoretical values in Table 1. It can be seen that the
error involved, even with this relatively low number of elements per
member, is almost negligible. The range of the minor-axis slenderness ratio
in these examples is from 152"7 to 288"5. Other structural configurations
have in fact been considered, including different loading conditions and
destabilising load positioning, producing an order of error which was very
consistent with those shown in Table 1.
To validate the model for inelastic lateral-torsional buckling, a compari-
son was made with full-scale tests carried out by Kitipornchai and
Trahair, :t3 who also made finite element predictions. Four tests were
carried out on different spans of a single serial size, which produced a
range of minor axis slenderness ratio from 68.14 to 170.34. The compari-
son between experimental and predicted results is shown in Table 2. The
present results compare well with the previous theoretical solutions 14 with

TABLE 1
Ambient-temperature Elastic Critical Loads Predicted by the Present Model, Compared
with Theoretical Critical Loads, 12 for a Simply Supported Beam with a Point Load at
Mid-span

Beam serial size Beam span Compression Predicted Theoretical Error


(m) flange elastic critical load (%)
slenderness critical load (kN)
ratio 2 = l/ry (kN)

838 292 x 176UB 12"0 203"4 280"80 280'90 0"04


686 254 x 125UB 8"0 152"7 394"00 390"41 0"92
533 210 x 82UB 8"0 182'7 154"50 154"20 0"19
406 178 x 54UB 6"0 155'8 137"00 136"77 0"17
406 x 140 x 39UB 6"0 207"6 55~70 55-35 0"63
305 102 x 28UB 6"0 288"5 23"65 23"53 0"51
254 102 x 28UB 6"0 270"3 27"00 26"78 0"82
106 C. G. Bailey, I. W. Bur#ess, R. J. Plank

TABLE 2
Comparison Between Computer Model Predictions of Inelastic Lateral-torsional Buckling
Loads and Kitipornchai and Trahair 13'14 Experimental and Computed Results

Beam serial size Beam span Slenderness Exprimental Present Kitipornchai


(m) ratio ~ = l/ry critical load analysis & Trahair
(kN) (kN) analysis (kN)

10UB29 2"44 68"1 234"96 259"199 252-5


3"05 85"2 185"12 204"998 198"3
3"66 102"2 145"07 148-399 152"6
6"10 170"3 56"96 53-599 55"0

The load was placed 219 mm above the shear centre.

both sets of predictions being within about 10% of the experimental


results. Kitipornchai and Trahair attributed the error between their
predicted and experimental results to initial geometric imperfections. The
analyses differ slightly in that initial residual stresses, for which an
assumed pattern was included in the original, were ignored in the present
work. Within the validation study all imperfections were assumed to be
represented by an equivalent intial out-of-plane bow, consistent with the
normal BS5950 Part 115 assumption. An inspection of the published
lateral displacements during the test suggests that this has underestimated
the actual initial imperfection.

3 T H E B E H A V I O U R O F B E A M S AT
ELEVATED TEMPERATURES

In order to set the findings for unrestrained beams into context the
prescribed limiting temperatures are first compared with analytical results
for beams with the compression flange continuously restrained against
lateral deflection. Uniform heating is assumed throughout. A Ramberg-
Osgood representation 16 of the BS5950/EC3 steel stress-strain data at
elevated temperatures was used in the analysis.

3.1 Beams with continuously restrained compression flange

This case has often been analysed, 16'17 particularly for beams supporting
concrete slabs on their top flange with uniformly distributed loading, and
it is known that the limiting temperatures given in BS5950 Part 8 and EC3
Part 1.2 for this three-sided heating case are an accurate lower bound. This
The lateral-torsional bucklin0 of unrestrained steel beams in fire 107

is a slightly different case from those analysed here, since it has a cooler
top flange and therefore bending occurs about a neutral axis above the
section's centroid. Limiting temperatures given in BS5950 Part 815 and
EC3 Part 1.218 for the four-sided heating case are lower than these since
more of the section is at a high temperature. Since no buckling can take
place the 'failure' criterion when the in-plane deflection reaches span/20
must be used. A comparison is shown in Fig. l(a) of the predicted limiting
temperatures for uniformly loaded beams of a series of different serial sizes
and spa:as giving different span:depth ratios. It has previously been
s h o w n 16 that the span:depth ratio is a controlling parameter for limiting
temperatures of restrained beams, and the effect of this parameter is
illustrated in Fig. 1. It can be seen that the limiting temperatures given by
BS5950 Part 8 are generally conservative compared with the computed
results. The yield stresses to be used in design at ambient temperature
differ slightly between BS5950 and EC3, and so it is not possible to plot
the EC3 limiting temperatures on the same figure since the load ratios
have slightly different meanings. However, these differences are small, and
the relationship between limiting temperature and load ratio is identical
in the two Codes, so it may be said that use of the limiting temperature
method on simple beams with continuous compression flange restraint
and uniformly distributed loading is vindicated by these results.

Failure Temperature C Failure Temperature C


688x254x125 UB
850 _ 6OOx254x125 UB 350 Vlspan = 6.0m)
[span = 5.Ore) I sparddepth = 8.9
800 span/depth = 8.9 800 IL /

750 838x292x194 UB
- (span = 9.Ore)
750 ~ (s~n = 9.Ore)
7O0 700 | / span/depth = 10.7
span/dept h =10,7 ~._~,,~ ~ | i- 457x191x67 UB
650 650 i ~ ' ~ ~ [ j (span = 5.Om)
6OO 600 ] ~ | span/depth = 13.2
Limiting temperatures /
55O 550 from BS.59501~.8
flx~rn BS.5O5O Pt.8
5OO I 5OO
356x127x39 UB
457x191x67 UB 45O
450
(span = 6.Ore) I spen/depth = 17.0 I
4(3O span/depth = 13.2 400 i
203x133x30 UB
350 35ox127x39 UB 350
(span = 6.0m) spen/depth = 2 4 2 I
3O0 3OO
span/depth = 17.0 i
250 25O 305x305x158 UC /
203x133x30 UB M M (span = O.Om)
200 (span = 5.0m) | 200 ~'__ ,) span/depth = 27.5
span/depth = 24.2
150 ~uII lateral restraint) 150 (full lateral restraint)
305x305x153 U
100 (span = 9 0 m ) 100
qT-/-/77F7
span/depth = 27.5
5O 50
0 i i i 0 I I I I ~ I
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.1 0.2 0.3 0,4 0,5 0.6 0.7 0.8
Load Ratio Load Ratio

(=) (b)

Fig. 1. Failure temperatures for simply supported beams with full lateral restraint to the
compression flange. (a) Uniformly distributed load, (b) Uniform moment.
108 C. G. Bailey, I. W. Burgess, R. J. Plank

When the form of loading is changed to the most onerous condition of


uniform bending it can be seen from Fig. l(b) that the limiting tempera-
tures lie within the cluster of computed results rather than below them.
For the higher span :depth ratios the Codes are no longer conservative,
although the discrepancies are relatively small and would not imply more
than a few minutes reduction in fire survival time. This might, of course,
make a considerable difference in the degree of protection required
because of the stepwise manner in which most regulatory authorities
specify fire resistance periods. The relationship between load ratio and
limiting temperature in both Codes is based simply on strength reduction
factors. If steel exhibited a sharp yield-point at high temperatures, and if
failure was defined by completion of the central plastic hinge, then the use
of strength reduction factors would produce a single failure temperature
for all simple beams at any given load ratio, without reference to the
loading case. Because a deflection criterion is actually used, with stress-
strain curves which are curvilinear at high temperatures, some spread of
failure temperatures occurs, as can be seen. However, this spread should
never be very great for restrained beams.

3.2 Unrestrained beams with load applied at the shear centre

The analysis was used to carry out an investigation into the behaviour of
uniformly heated, unrestrained simply supported beams, loaded at the
centroid (also the shear centre) of the cross-section. There are several
factors which affect the lateral-torsional buckling behaviour of steel
I-beams at ambient temperature, including cross-sectional properties,
span, support and restraint conditions, and the precise form of loading
employed. For any given case the ratio M b / M e x (in the terminology of
BS5950 Part 1is) is a convenient measure of the beam's propensity to
lateral-torsional buckling; low values tend to represent slender cases, and
values approaching 1.0 cases where plastic hinge failure is becoming the
preferred mode. Six different serial sizes were chosen, with spans giving a
wide range of values for Mb/Mcx. Two of the beams, 457 x 191 x 67UB and
838 x 292 194UB, were chosen to give exactly the same value of Mb/Mox.
The range of the ratio Mb/Mcx was chosen so that at ambient temperature
both elastic and inelastic lateral-torsional buckling failure modes were
represented. The actual value of this ratio varies with the loading arrange-
ment, and is shown on the figures. An initial out-of-plane imperfection was
assumed which gave an imperfection parameter value at mid-span equal
to r/=0-0032. The support conditions allow both flanges to rotate and
warp, but prevent twisting about the beam's axis. Thus, the effective length
of the beam is its actual span, in the absence of intermediate points of
restraint.
The lateral-torsional buckling of unrestrained steel beams in fire 109

The loading conditions were chosen to give a wide range of bending


moment distributions. The most severe case in terms of its effect on lateral
stability is that of uniform moment, for which the slenderness correction
factor n i,; set at 1"0 in BS5950 Part 1. This may be considered as the basic
case for lateral-torsional buckling. Other less severe patterns of moment
are assigned lower values of n in BS5950 according to the severity of the
effect of the bending moment distribution on buckling capacity. Each of
the six beams was subjected to different loading conditions, and to load
ratios M/Mb (now in the terminology of BS5950 Part 8) of 1.0, 0.7, 0.6, 0.5,
0-4, 0.3, 0.2 and 0-1. Four elements were found to be adequate to model
each beam for all loading conditions, except that of uniformly distributed
load for which the number of elements was increased to eight. The results
obtained are shown in Figs 2-7, in which the Limiting Temperatures
presented in table 5 of BS5950 Part 82 are also shown for comparison.
It is of interest to compare the computed failure temperatures with both
BS5950 ~Lnd EC3 limiting temperatures, at least for the basic case of
uniform naoment. Specimen design calculations of limiting temperature
according to both Codes are given in the Appendix. Both design codes
essentially use the same concept of a load ratio, with EC3 using a
reduction factor in its 'Load Domain' calculation and a degree of utilisa-
tion in the 'Temperature Domain'. For equivalent values of the load ratio
very similar failure temperatures are obtained from the two codes. How-

Failure Temperature C 356x127x3g UB Failure Temperature C Critical temperatures


85O- -(span = 6.0 m) 850 35Ox127x39 UB
', ~-EC3 Pt. 1.2 ~-(span = 6.0 m)
800- Limiting Mb/Mcx = 0.25 80O :~', / / Mb'ad/M='' = 0"29
~-emperet . . . . 457x191x67 UB
750- 750
"~ ~presented in (
~.=e.0m) ~', / i 457x191x87 UB
700- 700 N ~ [- (span = 6.0 m)
~-,~\_ ~ I /Mb.nd/IMc,Rd=0,45
~--~'. 838x292x194 UB
650 -
600-
~ . r(s..=o__o.;~ 65O
6O0

~------~ /
203xl 33x30 UB
r(span=5.0m)
550 -
500 -
450 -
550 -
500-
450 -
~ 535x20~i~4uB j f'.~
b.Rd/Mc.Rd= 0.53

(span =9.0 m) ~
400 400-
M b,RdlMc,Rd= 0.46
350 - 350-
300 3OO 5U
25O 250 -
Mb,Rd/Mc.Rd= 0.62
20O 200.
305x305x158 UC
150 M M 914x305x253 UB~ 150 - M M (span = 9.0 m)
(span = 5.0 m) Mb.Rd/Mc,Ra= 0.79
100 Mb/M= = 0.74
50 T ~ / ~ - 5o.
0 0 I I I I I I
0:, 012 0:3 0:4 0:5 0:0 017 05 0 01 o, 03 0,4 0.5 0.6 0.7 0.5
Load Ratio Degree of utilisation

(=) (b)

Fig. 2. Failure temperatures for unrestrained simply supported beams with uniform
moment. (a) Comparison with BS5950 Part 8, (b) Comparison with EC3 Part 1.2.
110 C. G. Bailey, I. W. Burgess, R. d. Plank

Failure Temperature *C
356x127x39 UB
850 Rrfrnm
L i m i t itemperatures
~KaKn
ng ~ (span = 6.0 m)
800 ~ ~ I " v. ' " ' vk" v v ' ~ " Mb/Mcx = 0.28
_457x191x67 UB
750 (span = 6.0 m)
700 M b/Mcx = 0.43
_838x292x194 UB
650 (span = 9.0 m)
600 M b/Mcx = 0.43
550 ~ "--.

450 203x133x30 U B ~ ~
(span = S.0 m) t"
400 Mb/Mcx = 0.5 /
/
350 686x254x125 UB_ /
(span = 6.0 m) |
300 Mb/Mcx = 0.56
250 305x305x158 UC_J
(span = 9.0 m)
200
Mb/Mcx = 0.78
150 "~
100
5O
0 I I I I I I I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Load Ratio

Fig. 3. Failure temperatures for unrestrained simply supported beams under uniformly
distributed load placed at the shear centre.

ever, the essential difference in the processes is in calculation of the


ambient-temperature buckling resistance moments. The calculations in the
Appendix shows that a simply supported 6 m beam of 686 254 125UB
section in Grade 43 steel has an ambient-temperature buckling resistance
of 549 kNm according to BS5950 Part 1, and 649.2 kNm according to EC3
Part 1.1. This apparently increased capacity using EC3 is then reduced in
Part 1.2 by the use of an 'empirical correction factor' which effectively
divides the buckling moment by 1.2, reducing it in the fire calculation to
541 kNm. Extrapolation from this may be dangerous, since the discrepan-
cies between the two codes calculations of ambient-temperature buckling
capacity is rather inconsistent. It must therefore be expected that load
ratios according to the two codes will be different in general. For the
example presented in the Appendix the failure temperatures predicted by
both code calculations are identical at 555C, which does not compare
well with the analytical prediction of buckling at 487C. Fig. 2(a,b) shows
the full comparisons between limiting temperatures according to both
codes and the analysis, for all six beams over the full range of load ratios.
The lateral-torsional buckling of unrestrained steel beams in fire 111

Failure Temperature *C
.356x127x39 UB
850 limiting temperatures (span = 6.0 m)
from BS5950 Pt.8 Mb/Mcx = 0.32
800 ~k F -457x191x67 UB
~, | (span = 6.0 m)
75o = 049

700 N~. r 838x292x194 UB


650 "~'.~ ] [ (span = g.o m)
600 - "''''"'- = "
550

500- 203x133x30 UB_]


450. (span = S.O m) I [
400- Mb/Mcx=0"5S J [
688x254x125 UB_I I
350- (span = 6.0 m) I
300 - Mb/Mcx = 0.62 I
250- 305x305x158 UC/
(span = g.o m)
200, /rP M b/M cx= 0.82
150- ,,~ m
U2 U2
100 - " ~
50.

0 I I I t I I i
0 0.1 0.2 0:3 0.4 0.5 0.6 0.7 0.8
Load ratio

Fig. 4. Failure temperatures for unrestrained simply supported beams under mid-span
point load placed at the shear centre.

The essential difference between the two figures lies in the way in which
the load :ratios have been calculated. It is interesting to note from Fig. 2(a)
that the beams 457x 191 67UB and 838 x292 x 194UB, which have
spans set to produce identical Mb/Mc,, values for uniform moment
according to BS5950, have identical failure temperatures throughout the
range of load ratio.
The predicted failure temperatures for various other loading conditions
are shown graphically in Figs 3-7, but are only compared with BS5950
limiting temperatures. These show a similar general pattern of results to
those for uniform moment, except that the range of variation in failure
temperatures for various Mb/Mc,, values is not as high.

3.3 Unrestrained beams with load applied on the top flange

The destabilising effect of placing the load on the top flange was consider-
ed for the case of uniformly distributed load. The same six beams were
analysed as before, but with different spans. Once again the 457 x 191 x
112 C. G. Bailey, I. W. Burgess, R. J. Plank

Failure Temperature C
Umiting temperatures 356x127x39 UB
850
800

750
700
650
~[ from BS5950 Pt.8 -(span = 6.0 m)
MI~/M~ = 0.38
457x191x67 UB
- (span = 6.0 m)
M b/Mcx = 0.56
- 838x292x194 UB
(span = 9.0 m)
M b/Me= = 0.56
6O0
550

500
450 203x133x30 UB
400 (span = 5.0 m)
Mb/M= = 0.62
350
686x254x125 UB
300. (span = 6.0 m)
250 Mb/Mcx = 0.69

200 305x305x158 UC
M (span = 9.0 m)
150 ~" M b IMcx = 0.87
100

50-
0 I I I I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Load Ratio

Fig. 5. Failure temperatures for unrestrained simply supported beams under moment
applied at one end.

67UB and 838 x 292 x 194UB spans were chosen to give exactly the same
value of Mb/Mox to allow a comparison to be made.
As can be seen from the results shown in Fig. 8, the failure temperatures
are higher compared with the predicted results with the load placed at the
shear centre. This is due to the very conservative nature of the BS5950
Part 1 design calculation is for the ambient-temperature buckling resis-
tance m o m e n t under destabilising load. The method uses a slenderness
correction factor of 1.0, irrespective of the m o m e n t pattern, and also a
'blanket allowance' of an assumed effective length equal to 1.2 times the
actual span. Load ratios are therefore artificially increased in these cases.
EC3 Part 1.1 adopts a more analytical approach to the calculation of
ambient-temperature load capacities under destabilising load, in which the
height of the load application point is specifically used, and this again
produces higher buckling capacities than does BS5950 Part 1. In the
examples used here the ambient-temperature buckling capacities for uni-
formly distributed loading with the load at the top flange are increased
The lateral-torsional buckling of unrestrained steel beams in fire I 13

Failure TemperatureC
358x127x39UB
850 Limiting temperatures -(span= 6.0 m)
~ .from BS5950Pt.8 Mb/M~ = 0.49
000- 457x101)(87UB
750 / (spen ,, e.o m)
/

Mb/Mcx= 0.67
700-
| 838X292X194UB
650 / -(span= 9.0 m)
~ ~ Ulo/M. =0'67
600.
550
580,
450,
400,
350.
300. (span = 6.0 m)
Mb/Mex= 0.79
250
305x305x158UC
200 (span = g.o m)
150 ~.,,.~ M..~ Mb/M~ = 0"95
100.
50.
0 I I I

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8


Load ratio

Fig. 6. Failure temperatures for unrestrained simply supported beams under equal end-
moments.

generally by about 20%, the exception being the Universal Column


section for which the increase is only 9%, although this increase in
capacity is reduced by use of the 'empirical correction factor' of 1.2 in the
high-temperature calculation.

4 DISCUSSION OF THE RESULTS

It can be seen that, at least for the higher range of load ratios, the failure
temperatures for unrestrained beams with any consistent loading pattern
vary with the value of Mb/Mcx, with instability occurring at progressively
lower temperatures as this increases. It may be agrued that this is because
inelastic buckling loads in general are controlled by tangent modulus
values oll the local stress-strain curves of the members affected just prior
to instability occurring. The stress-strain curves of steel at elevated
temperatures are continuously curvilinear, tending towards zero slope at
114 C. G. Bailey, I. IV. Burgess, R. J. Plank

Failure Temperature *C
_ 356x127x39 UB
850 Limiting temperatures (span = 6.0 m)
from BS5950 Pt.8 M b/Mcx = 0.28
800
_ 457x191x67 UB
750 (span = 6.0 m)
M b/MCX = 0,43
700
838x292X194 UB
650 (span = 9.0 m)
600 M b/Mcx = 0.43

550
500

450
203x133x30 UB
400 (span = 5.0 m)
M b/Mcx = 0.5
350
686x254x125 UB_
300 (span = 6.0 m)
25O Mb/M~ = 0.56
P P
305x305x158 UC
200 ~t J/=== (span = 9.0 m)
150 u4 L/4 Mb/Mcx = 0.78
100
8o X/////
0 I I I I I i i
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Load Ratio

Fig. 7. Failure temperatures for unrestrained simply supported beams with point loads at
the quarter-points, placed at the shear centre.

ultimate stress levels which degrade with rise in temperature. At any given
load ratio a beam with higher value of Mb/Mcx will be more highly
stressed than one with a lower value. Its compression flange is therefore
subject to a lower tangent modulus in resisting any infinitesimal disturb-
ance, and as the temperature rises this tangent modulus rapidly declines.
This is shown schematically in Fig. 9. As load ratios decrease the failure
temperatures of all beams can be seen to converge to a single value
depending only on the load ratio. Failure for these cases is essentially by
elastic buckling, which takes place when increase of temperature has
reduced the initial (Young's) modulus to a proportion of its initial value
approximately equal to the initial load ratio.
An effort was made to represent the failure temperatures for unre-
strained beams at elevated temperatures in a normalised fashion for any
given load ratio, independent of the loading pattern, similar to the
treatment of different load patterns at ambient temperature. This does not
seem to be possible due to the different rates of degradation of stiffness for
The lateral-torsional buckling of unrestrained steel beams in fire 1! 5

Failure temperature *C
Limiting temperatures ,56x127x39 UB
850 from BS5950 Pt.8 - (span = 5.0 m)
M b/Mc~ = 0.25
800
~ I 457x191x67 UB
750 %~ / r/ M.,M
(span = 5.0 m)
= 03,
700
~ . I ~ 83ex292x194 UB
650
"~',~ I
I (span = 7.5 m)
~ I M ~ = 0.39
600
550
500
450
203x133x30 UB
(span = 5.0 m)
~ I T"
400 Mb/Mcx = 0'47 I I

350
,w sasx2S4x125 UB_j I
(span = 6.0 m) -- I
300 MblMcx = 0"52 I
250 (Loaded at 305x305x158 UCJ
top flange) (span = 8.0 m)
200 Mb/Mcx = 0.72
150
100
5O
0 I I I I I I I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Load Ratio
Fig. 8. FailLure temperatures for unrestrained simply supported beams loaded at the top
flange (destabilising position) with uniformly distributed load.

Stress
T1
f " ~ Temperaturevalues
j ~ / ('I"2> T1)
T2

E1
ess Level

L_ E 2 ......._L_ow__S!ressLevel

Strain

Fig. 9. Illu=~tration of the different rates of degradation of tangent modulus with tempera-
ture for members at high and low load levels.
116 C. G. Bailey, I. W. Burgess, R. J. Plank

different loading arrangements. For example, considering a beam under


uniform moment the stress distribution through the cross-section is the
same along the length of the beam, and so the reduction of stiffness due
to the rise in temperature is constant along its length. However, in the case
of moment gradient the rate of reduction of stiffness in the more highly
stressed zones of the beam will be greater than in the zones of low stress.
This can be seen by comparing the cases of uniformly distributed load and
of two point loads placed at the quarter points of the span (see Figs 3 and
7). Both of these loading arrangements carry a slenderness correction
factor of 0-94 and thus have similar buckling capacities at ambient
temperature. However, the two point loads produce peak moment over
the greater length of beam. Since the degradation of stiffness is greater in
the zones of higher stress the beam with the point loads fails at the lower
temperatures of the two. Unfortunately there seems no obvious method of
rationalising the results presented for lateral-torsional buckling.

5 CONCLUSIONS

The results presented indicate that both BS5950 Part 8 and EC3 Part 1.2
overestimate the limiting temperatures for unrestrained simple beams in
fire resistance calculations. This is in contrast to the very accurate
predictions given for the case, which is very common in simply designed
building frames, of uniformly loaded beams with full lateral restraint from
concrete slabs. The less conservative predictions for unrestrained simple
beams are for those load cases which have high moment over the longest
portions of the beam in the midspan region. Thus, a moment profile
which changes sign at midspan produces the best prediction and uniform
moment the worst.
Clearly more work is needed before an all-embracing approach to the
design of unrestrained beams for fire resistance can be proposed. At
present, however, the indications are that for a given bending moment
pattern the locus of change of buckling temperature with load ratio is
governed by the value of Mb/Mcx(or Mb.Rd/Mc.Rdin the notation of EC3),
and that limiting temperatures decrease as this increases, at least within
the range of cases examined in this study.
Despite the fact that the more extreme cases studied here give very low
failure temperatures which are considerably below the prescribed values,
it must be recognised that beams in real structures are very rarely subject
to the idealised conditions used. Support conditions often provide con-
siderable restraint and continuity which is not allowed for in simple design
approaches. Even after buckling has occurred there is every prospect that
The lateral-torsional bucklino of unrestrained steel beams in fire 117

beams will progressively carry more of their load by catenary action,


mobilising the tying strength of their connections, as deflection increases,
so that catastrophic collapse should in general be averted.

ACKNOWLEDGEMENT

The work presented in this paper was performed in the course of a


research project sponsored by the Engineering and Physical Sciences
Research Council of Great Britain, whose support is gratefully acknow-
ledged.

REFERENCES

1. The Buildino Reoulations, Approved Document B: Fire Safety, Department of


the Environment and the Welsh Office. HMSO, 1991.
2. BS5950: 1990, Structural Use of Steelwork in Buildinff. Part 8: Code of Practice
for Fire Resistant Desion, BSI, London (1990).
3. Eurcode 3: Desiffn of Steel Structures. Part 1.2: Structural Fire Desion (Draft),
European Committee for Standardisation (1993).
4. British Steel Corporation, Compendium of UK Standard Fire Test Data on
Unprotected Structural Steel, Report RS/RSC/S10328/1/87/B, 1987.
5. BS476:1987 Method for Determination of the Fire Resistance of Elements of
Construction: Part 20, BSI, London, 1987.
6. Brett, P. and Rushton, J., Parallel Beam Approach - A Design Guide. Steel
Conlstruction Institute, 1990.
7. Trahair, N. S. and Bradford, M. A., The Behaviour and Desion of Steel
Structures. Chapman & Hall, 1991.
8. E1-Zanaty, M. H. and Murray, D. W., Non-linear finite element analysis of
steel frames. ASCE J. Struct. Div., 109 (ST2) (1983) 353-368.
9. Saab, H. A. and Nethercot, D. A., Modelling steel frame behaviour under fire
conditions. Enono. Struct., 13 (1991) 371-382.
10. Najjar, S. R., Three-dimensional analysis of steel frames and subframes in fire,
PhD Thesis, University of Sheffield (1994).
11. Najjar, S. R. and Burgess, I. W., A non-linear analysis for three-dimensional
steel frames in fire conditions. Enono. Struct., (in press).
12. Timashenko, S. P. and Gere, J. M., Theory of Elastic Stability, Mc.Graw-Hill,
1961.
13. Kitipornchai, S. and Trahair, N. S., Inelastic buckling of simply supported
steel I-beams. ASCE J. Struct. Div., 101(ST7) (1975) 1333-1347.
14. Kitipornchai, S. and Trahair, N, S., Buckling of inelastic I-beams under
moment gradient. ASCE J. Struct. Div., 101(ST5) (1975) 991-1004.
15. BS 5950: 1990, Structural Use of Steelwork in Buildino: Part 1. Code of
Practice for Desion in Simple and Continuous Construction: Hot Rolled
Sections, BSI, London, 1990.
118 C. G. Bailey, I. W. Burgess, R. J. Plank

16. Burgess, I. W,, E1-Rimawi, J. and Plank, R. J., Studies of the behaviour of steel
beams in fire. J. Constr. Steel Research, 19 (1991) 285-312.
17. Lawson, R. M. and Newman, G. M., Fire Resistant Design of Steel Struc-
tures--A Handbook to BS5950: Part 8. Steel Construction Institute, 1990.
18. Eurocode 3: Design of Steel Structures. Part 1.1: General Rules and Rules for
Buildings (Draft), Commission of the European Communities, 1993.
19. Steelwork Design Guide to BS5950: Part 1: 1990, Volume 1, Section Proper-
ties, Member Capacities: 3rd Edition, Steel Construction Institute, 1993.

APPENDIX. DESIGN ESTIMATION OF


LIMITING TEMPERATURES

This example illustrates the procedures for estimation of limiting tempera-


tures for a simple beam of serial size 686 254 125UB, Grade 43, with
6.0 m span and uniform bending m o m e n t of 329.4 k N m at the fire limit
state.

Using BS5950 Parts 115 and 82


Buckling capacity calculation from Part 1:
Slenderness correction factor n = 1.0 [table 20]
Effective length Le = 6.0 m [table 9]
Buckling resistance moment19 Mb = 549 kNm.
Limiting temperature calculation from Part 8:
Equivalent uniform m o m e n t m = 1.0
factor
Applied m o m e n t at the fire Mt = 329.4 kNm,
limit state
Load ratio mMf
R- =0.60 [cl. 4.4.2.2]
Mb
Hence, limiting temperature =555C.

Using EC3 Parts 1.11 a and 1.23

Buckling capacity calculation from Part 1.1:


Since E = 210,000 N / m m 2, G = 80769-2 N / m m 2,
Iw = 4.79 x 1012 m m 6, I z = 4 3 8 0 x 104 m m 4, It = 116 x 104 m m 4,
k = 1.0, kw = 1-0, C1 = 1.0

Elastic critical m o m e n t Mcr


=CI(-KL-~) E -~z -t- ~2EIz J

= 965.2 k N m
The lateral-torsional bucklin# of unrestrained steel beams in fire 119

Yield , ; t r e s s fy = 275 N/mm 2 [table 3.1]


For a class 1 section /~w = 1"0 [table 5.3.1]
Normalised slenderness ,~LT r_
7/t~.wo,,j,/o.,
... _

= 1.07
4'LT ----0"5 [ 1 + ~,~T(:rL, -- 0"2) + I-~T]
=1"16
1
XLT -- ~ + [#,L~-
2 ~-~T]
2 0"5

= 0.62
UK partial safety YM,1 = 1.05
factor
Buckling resistance Mb,Rd =XLlrBw Wvlyfy [cl. 5.5.2]
~M,1
moment
= 649.2 kNm.
Limiting: temperature calculation from Part 1.2 (temperature domain):
Design effect of the Eei,d = 329.4 kNm,
actions in fire
Load capacity at time Rfi,d,o = 649"5/1-2 = 541"25 kNm
t=0
(where 1.2 is an empirical correction factor)
Degree of utilisation /~0 = Efi,d/Rfi,d,O
at time t = 0
=0.61
Critical (limiting) 0a,er = 39.191n(0.96741-bto3833 1)+482
temperature
= 551C. [cl. 4.2.2.5]
Alternative limiting temperature calculation from Part 1.2 (load domain):
Design buckling Mb,fi,t,Rd ----3 2 9 " 4 kNm,
resistance at time t
Reduction factor for
yield strength of steel
at temperature 0ma, ky,0,max _ l'2Mb,fi,t,Rd
Mb,Rd

=0"609
From table 3.1 0m,, =555C. [cl. 4.2.2.2]
The limiting temperatures of 555C (BS5950), 551C and 555C (EC3) may
be compared with the computed failure temperature of 487C for this case.

You might also like