Engineering Structures 22 (2000) 755768

www.elsevier.com/locate/engstruct

The influence of the thermal expansion of beams on the structural

behaviour of columns in steel-framed structures during a fire
*
C.G. Bailey
Building Research Establishment, Centre for Steel Construction, Garston, Watford WD2 7JR, UK

Received 18 December 1998; accepted 11 March 1999

Abstract

In the UK, new design guidance is currently being developed for the behaviour of steel-framed buildings when subjected to fire.
This is primarily based on recent research that considers the structural behaviour of all horizontal members, without applied fire
protection, acting as a complete entity within the building. This guidance assumes that columns designed to current design procedures
will always be adequate when used within this new design philosophy. For bare steel columns these existing design methods usually
consist of applying some form of passive fire protection.
Presented in this paper is an analytical investigation of the structural behaviour of columns when subjected to various structural
and fire scenarios. The results from this study do not endorse the view that current fire design methods for columns are adequate.
These design methods will require revision if instability of columns during a fire is to be avoided. In most cases this will result
in the need for additional passive fire protection to be applied to the steel columns. 2000 Published by Elsevier Science Ltd.
All rights reserved.

Keywords: Fire; Columns; Stability; Thermal expansion

1. Introduction iour of isolated members. Results from the tests sup-

Recent research [14] into the behaviour of steel-
framed buildings during a fire has begun to consider the
structural response of the building in its entirety instead
of considering structural members in isolation. In the UK
the nucleus of this research [1] is the recently completed
fire tests, conducted on a full-scale eight-storey steel-
framed building constructed at the Building Research
Establishment (BRE) Laboratory in Cardington,
Bedford.
A total of six compartment fire tests were carried out
[3], two by BRE and four by British Steel. The major
aim of the tests was to provide quality data to validate
and develop computer models, which will enable differ-
ent structural and fire scenarios to be investigated.
Before the Cardington tests, it was felt that existing
design guidance on horizontal spanning structural mem-
bers was too conservative, since it is based on the behav-
* Tel.: 44-1923-664619; fax: 44-1923-664096; e-mail:
baileyc@bre.co.uk
ported this theory, and together with supplementary
computer modelling design guidance is now being
developed based on whole building performance during
a fire.
The major aims of fire design, once ignition occurs,
is to contain the fire within the compartment of origin
and to limit structural damage so that overall collapse is
avoided. Before the Cardington fire tests were conducted
it was generally agreed, by the Research Organisations
[1] involved, that to meet these criteria the bare steel
columns would need to be designed to current practice
[58] (i.e., treated as isolated members). This involved
applying passive fire protection to the columns. This
effectively resulted in only the horizontal members being
tested. Since the limit in terms of load-carrying capacity
of these members in fire was unknown, it was decided
to protect the columns to a higher standard than that
given by current design methods, to ensure that the col-
umns did not fail first. This is based on the assumption
that columns designed to current UK practice will per-
form adequately in a fire, and therefore do not need to
be tested to their limit in the full-scale fire tests.

0141-0296/00/$ - see front matter 2000 Published by Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 1 - 0 2 9 6 ( 9 9 ) 0 0 0 2 8 - 0
756 C.G. Bailey / Engineering Structures 22 (2000) 755768
The design guidance, which is currently being
developed, based on the Cardington test results, is effec-
tively a hybrid design philosophy. The design of the
beams (which could possibly be unprotected) and slabs
is based on the behaviour of the structure as a whole,
whereas the columns are based on current design prac-
tice by treating them as a series of isolated members.
This paper investigates the validity of this approach and
raises questions about its safety in relation to the overall
stability of the structure.
2. Review of the full-scale fire tests conducted on
the Cardington frame
The full-scale eight-storey steel-framed test building
[9] was designed and constructed to resemble a typical
modern UK city centre eight-storey office development.
On plan (Fig. 1), the building covers an area of 21 m
45 m with an overall height of 33 m. There are five
equally spaced bays along the length of the building.
Across the width there are three bays spaced 6 m, 9 m
and 6 m. Placed centrally, on the footprint of the build-
ing, is a 9 m 2.5 m lift core with two 4.5 m 4.5
m stairwells placed at each end. The structure is
designed as a braced frame with lateral restraint provided
by cross-bracing around the three vertical access shafts.
The beams are designed as simply supported acting com-
positely (via shear studs) with the supported floor slab.
The composite floor slab is 130 mm deep and consists
of a steel trapezoidal deck, with lightweight concrete and
anti-crack mesh.
Six compartment fire tests were conducted on the
frame at various locations throughout the building. In all
the tests the columns were protected using ceramic fibre,
Fig. 1. Location of corner compartment fire tests.
with the steel beams and underside of the steel trap-
ezoidal deck unprotected. To highlight the behaviour of
the columns the results from two of the tests are briefly
presented, one conducted by BRE, the other by British
Steel. The locations of these tests are shown in Fig. 1
and are situated in the corner bays of the frame. The
following description of the tests concentrates on the
behaviour of the columns. A more detailed description
of the overall structural behaviour is presented else-
where [1,2].
2.1. BRE corner fire test
In the BRE corner fire test, situated between second
and third floor levels, the columns were extensively
instrumented, which allowed their behaviour during the
test to be investigated. Figs. 2 and 3 show the maximum
recorded test moments (calculated from the strain
gauges) about the minor and major axes at various pos-
itions along the length of the columns, between the first
and fourth floors. In the figures the notation (?) rep-
resents strain gauges that failed during the test. However,
it was possible in most cases to identify the trend of
the moments from the strain gauges at these locations,
which allowed the sign of the moment to be identified
and the full column moment pattern estimated.
From the test results it can be seen that the columns
are subjected to significant moments about both axes
during the fire test. The pattern of the recorded moments
suggests that the columns are displaced laterally at third
floor level in the minor axis direction and, for perimeter
columns (depending on their orientation), in the major
axis direction. This was caused by the thermal expansion
of the connected heated beams and has been proved by
 C.G. Bailey / Engineering Structures 22 (2000) 755768 757

Fig. 3. Maximum recorded major axis moments during the BRE cor-
Fig. 2. Maximum recorded minor axis moments during the BRE cor- ner fire test.
ner fire test.

previous analytical work [10] and validated by test

measurements.
One visual phenomenon of the structural behaviour of
the BRE corner test was that no local buckling of the
beams occurred. In all the other fire tests local buckling
occurred in a large number of beams in the proximity
of the beam-to-column connections. It is felt that this
local buckling was caused by the restraint to thermal
expansion of the beams together with the negative
moment caused by the rotational restraint of the connec-
tion [11]. To investigate if local buckling of the beams
reduces the column moments the corner compartment
fire test conducted by British Steel, where extensive
local buckling of the beams occurred, was investigated.

2.2. British Steel corner fire test

Fig. 4 shows the locations of beam local buckling in
the British Steel corner fire test, which was carried out
between first and second floors. The major axis moments
of column E/1 were recorded at a position 500 mm
above the second floor and are shown in Fig. 5. The
maximum recorded major axis moment of 343 kNm in
column E/1 is in good agreement with the maximum
recorded major axis moment of 316 kNm in column E/4
(Fig. 3). The horizontal displacement at floor level, in
Fig. 4. Position of local buckling in the steel beams during the British
Steel corner fire test.
relation to the temperature of the connecting beam, for
column E/1 about the major axis, is shown in Fig. 6.
Also shown in this figure is the horizontal displacement
for column E/4 (Fig. 1) recorded during the BRE fire
test. It can be seen that the horizontal displacement is
similar for both tests but the temperatures of the con-
necting beam are different. Although local buckling in
758 C.G. Bailey / Engineering Structures 22 (2000) 755768

Fig. 5. Recorded major axis moment for column E/1 during the British Steel corner compartment fire test (measured 500 mm above second floor).

Fig. 6. Recorded lateral displacement of column E/1 (British Steel test) and column E4 (BRE test).

the beams should, in part, alleviate the lateral displace-
ment of the columns, results from all the fire tests indi-
cate that local buckling does not occur until beam steel
temperatures in excess of 500C are reached. From Fig.
6 it can be seen that the difference in displacements
occurs at the start of the test and continued throughout
the duration of the test. This difference could be due to
either one of the following, or a combination of the two
(together with local buckling at higher temperatures):
1. The beam in the BRE test was only heated from one
side and was partially protected by the compartment
wall [2].
2. The restraint from the composite floor was different
in each test, due to the different timetemperature
relationship of the tests, as shown in Fig. 7.
Further analytical work is required to obtain definitive
explanations for the different thermal expansion of simi-
lar beams in the BRE and British Steel tests. However,
both tests have shown that high moments are induced
into the columns owing to thermal expansion of the con-
necting beams. This paper presents the possible conse-
quence of this thermal expansion on the behaviour of the
columns and considers the influence of different fire and
structural scenarios.
3. Behaviour of steel columns due to lateral
displacement caused by thermal expansion of the
connecting beams
The results from the Cardington fire tests have shown
that the internal and external columns are subjected to
high moments caused by expansion of the connecting
beams during a fire. If these moments were simply
included within the present member design procedure
 C.G. Bailey / Engineering Structures 22 (2000) 755768 759

Fig. 7. Maximum atmosphere temperatures recorded in the BRE and British Steel corner fire tests.

outlined in BS5950 Part 8 [6] or EC3 Part 1.2 [7], the

calculations would show that the columns would fail
during the fire owing to local plasticity. However, in the
case of a braced frame a large amount of redundancy
exists and it should be possible for the induced column
moments to cause high localised stresses, resulting in
plastic hinges at floor level, without causing overall col-
lapse. This assumes that the columns are restrained hori-
zontally at floor level. Owing to its good inherent fire
resistance the composite floor should provide this
restraint during a fire. Although it is assumed that local-
ised plasticity can be accommodated, the influence of
expanding beams on the overall stability of the column
must be checked during a fire.
The simple model shown in Fig. 8 was developed to
investigate the stability of columns, during a fire, which
are subjected to both static axial load and moments
caused by the expansion of the connected steel beams.
The following modelling assumptions apply:
1. The floor slab is ignored, but the heated beams are
restrained axially to provide lateral restraint to the
columns.
2. No static load is applied to the beams.
3. The beams are heated such that the top flange is 80%
of the temperature of the web and lower flange, which
causes the beams to deflect vertically owing to ther-
mal curvature.
4. The column is subjected to a concentric static load,
with the bottom 4185 mm uniformly heated.
5. The column has no axial restraint to thermal expan-
sion (except that provided by the flexural stiffness of
the connecting beams).
6. The column has an imperfection of L/1000 about the
minor axis.
Some of the above assumptions require further expla- Fig. 8. Simple model used in parametric study.
760 C.G. Bailey / Engineering Structures 22 (2000) 755768

nation in relation to the observed structural behaviour

during the full-scale fire tests. By ignoring the floor slab
the lateral displacement of the column due to the thermal
expansion of the beams will be higher. However, test
results and previous modelling [10] have shown that the
overall displacement of the columns is caused by the
floors above and below the fire compartment being dis-
placed in opposite directions. This causes a higher over-
all lateral displacement compared to the lateral displace-
ment at one floor level only. In the absence of modelling
the whole frame (which must include the bracing) it is
assumed that by ignoring the floor slab in the simple
model conservative lateral displacements will be esti-
mated.
By applying no static load, the vertical displacement
of the beams remains very small, with only thermal cur-
vature being modelled. If the vertical displacement of
the beams becomes significantly large it will alleviate
the column lateral displacement. However, it was
decided not to include this behaviour, in the simple Fig. 10. Structural behaviour leading to either beam or column fail-
model, for two reasons. Firstly, it has been shown by ure.
tests [2] that beams parallel to the compartment per-
imeter have nominal vertical displacement, since they
The frame shown in Fig. 8 was analysed with the axial
are directly above a compartment wall. Secondly, in a
restraint at one end of the heated beams fully fixed. The
fire situation it cannot be guaranteed that all beams are
column was subjected to an axial load of 2944 kN. Fire
subjected to the assumed design-imposed loads.
design codes [68] represent the load at the fire limit
Considering the simple model it can be seen that the
state in terms of Load Ratio, Load Level or Degree of
following parameters require investigation:
Utilisation. These terms correspond to the capacity of
1. Beam to column heating rates the member at ambient temperature divided by the load
2. Beam size on the member at the fire limit state. The load of 2944
3. Column size kN on the column corresponds to a Load Ratio of 0.6.
4. Beam-to-column connection rigidity This assumes a column effective length of 0.85 times
5. Axial restraint at one end of the heated beams the system length. To find the worst heating scenario the
6. Span of beams differential heating rate between the column and beams
7. Base rigidity of the heated column was varied. These heating rates will cover the cases
8. Column axial load where the columns are protected with the connecting

Fig. 9. Failure of structure for varying column-to-beam temperature ratios.

 C.G. Bailey / Engineering Structures 22 (2000) 755768 761

Fig. 11. Lateral displacement of column about minor axis.

Fig. 12. Predicted minor axis column moments.

Fig. 13. Column minor axis moments at varying column temperatures.

762 C.G. Bailey / Engineering Structures 22 (2000) 755768
Fig. 14. Predicted lateral displacement of column about minor axis.
beams protected, partially protected or unprotected. Fig.
9 shows the failure temperatures for the beam and col-
umn for varying column-to-beam temperature ratios. It
can be seen that if the column is protected from the fire,
so that its maximum temperature is less than 60% of
the temperature of the beam, then failure is by beam
squashing. This is caused by axial restraint to thermal
expansion of the beam by the sway stiffness of the col-
umn. If the column reaches higher temperatures then
failure is by column instability. Fig. 10 shows a simpli-
fied representation of the type of structural behaviour
that causes either beam or column failure.
Considering the behaviour of the column, Fig. 9 indi-
cates that the worst scenario is when the column reaches
a temperature of 509C and the beam reaches a tempera-
ture of 848C. This failure temperature increases to
528C when the beams and column are heated at the
same rate. If the column was heated, and the beams
remained at 20C, it was shown that failure of the col-
umn would occur at 541C. This corresponds well with
the BS5950 Part 8 [6] Limiting Temperature (failure
temperature) value of 540C. Therefore considering the
results presented in Fig. 9, the Limiting Temperature of
the column should be reduced from 541C to 509C if
the beams are unprotected, so that they can reach tem-
peratures of 848C or higher. This will ensure stability
of the columns during any fire scenario. If the beams are
protected to the same degree as the columns, then the
Limiting Temperature should be reduced from 541C
to 528C.
It is important to investigate the results presented in
Fig. 9, in more detail, to ensure that column instability
is occurring and failure in modelling terms is not
caused by numerical instability. The column displace-
ments about the minor axis, for the analysis where the
column was heated at 70% of the beam temperature, are
shown in Fig. 11. As the failure temperature of 514C
is approached the column deflected shape suggests that
instability is occurring. The minor axis moments, for the
same analysis, are shown in Fig. 12. These indicate that
the moment at point C (floor level) reverses sign to a
restraining moment prior to instability. Both the pre-
dicted displacements and moments strongly indicate that
column instability occurs and not numerical instability
of the computer model.
The predicted moments along the length of the column
are shown in Fig. 13, for varying temperatures. From
this figure it can be seen that plastic hinges have formed
in the bottom and in the centre of the heated column,
with a third plastic hinge beginning to form at the top
of the heated column as the moment reverses direction
to a restraining moment. The formation of these hinges
is caused by column instability.
Considering the structural behaviour of the column,
the effect of the expanding beams causes the column to
deflect into double curvature with high localised stresses
occurring at the top and bottom of the column. This
should not cause instability, due to the restraint present
at floor level. Therefore, instability of the column must
be initiated by the P effect of the applied axial load
and forced displacement of the column. To investigate
this argument the same previous analysis, where the col-
umn was heated at 70% of the beam, was repeated
except that the axial load in the column was assumed to
be zero. Compressive failure of the beams was identified
when they reached 730C, with the column at 511C.
To investigate the behaviour of the column for higher
temperatures the beam was replaced by a 914 419
388UB. The column minor axis displacements are shown
in Fig. 14. Instability of the column did not occur. This
can be seen clearly from Fig. 15, which indicates that a
nominal moment occurs at the centre of the column. This
figure also indicates that a plastic hinge forms at the top
and bottom of the heated column, which reduces in
 C.G. Bailey / Engineering Structures 22 (2000) 755768 763

Fig. 15. Predicted minor axis column moments.

capacity as the column is heated and thus reduces in
strength.
Comparison of these analyses show that instability of
the column occurs due to a combination of the axial load
in the column and its deflected shape caused by the
expansion of the connecting beams.
3.1. Effect of beam size
The previous analyses (where the Load Ratio in the
column was 0.6) were repeated with the beams increased
to 914 419 388UBs. Fig. 16 shows the failure tem-
peratures of the simple structural model for different
beam-to-column temperature ratios. It can be seen that
increasing the beam size increased the beam-to-column
temperature range (i.e. from 0.4 to 1.0) over which col-
umn instability occurred. The analyses indicate that the
beam has an important effect and for a connecting beam
of size 914 419 388UB the column Limiting Tem-
perature should be reduced to 430C. However, it should
be noted that this beam-to-column size combination is
highly unrealistic and the only purpose of these analyses
is to investigate the effect of the beam size. From these
results it can be concluded that increasing the size of the
connecting beams has a detrimental effect on the stab-
ility of the columns.
3.2. Effect of column size
The above analyses were repeated with the column
size reduced to a 152 152 23UC. This had the
effect of increasing the slenderness of the column from

Fig. 16. Failure of structure for varying column-to-beam temperature ratios (beam 914 419 388UB; column 305 305 37UC).
764 C.G. Bailey / Engineering Structures 22 (2000) 755768

Fig. 17. Failure of structure for varying column-to-beam temperature ratios (beam 914 419 388UB; column 152 152 237UC).

the previous value of 45.4 to 96.1. The applied axial load
was 254 kN, which gave a Load Ratio of 0.6 (assuming
an effective length of 0.85 times the system length). For
the worst scenario (Fig. 17) the failure temperature of
the column is 430C, which is the same value predicted
for the above analyses where the column size was 305
305 37UC. When the column was heated and the
beams remained at ambient temperature column insta-
bility occurred at 590C, significantly higher than the
temperature of 541C predicted in the above analyses.
This phenomenon has been shown previously by various
researchers [12,13] and is due to the beneficial end
restraint for slender columns.
The effect of column slenderness does not seem to be
significant for the worst fire scenario, indicating that the
dominant behaviour is caused by the expansion of the
beam forcing the column into a deflected shape. This
causes the P effects to become significant and ulti-
mately leads to column instability.
3.3. Effect of beam-to-column connection rigidity
The original analyses (Fig. 9), where the beams were
356 171 51UB and the column was 305 305
37UC, were repeated with the beam-to-column connec-
tions assumed as fully rigid. The predicted failure tem-
peratures are shown in Fig. 18 and comparison with Fig.
9 indicates that the results are very similar. This implies
that the rotational stiffness and moment capacity of the
beam-to-column connection has nominal effect on the
behaviour of the column. This can be attributed to the
following:

Fig. 18. Failure of structure for varying column-to-beam temperature ratios (rigid beam-to-column connection).
 C.G. Bailey / Engineering Structures 22 (2000) 755768 765

1. The beams are heated, and at temperatures of 800C

they have nominal stiffness and strength. Therefore
they cannot provide significant restraining moments
when instability of the column occurs.
2. The amount of rotation stiffness required for a heated
column to tend towards the behaviour of a fully fixed-
ended strut is significantly less than the rotational
restraint provided by the continuation of the column
above the fire, which remains cold. This is clearly
indicated from the results of the analyses shown in
Fig. 19.

3.4. Effect of horizontal restraint to the heated beams

In all the previous analyses the beams were restrained

horizontally at the opposite end to the heated column.
To investigate the effect of varying the horizontal
restraint the original analyses (Fig. 9) were repeated with
a horizontal restraint of 33841 N mm for the 6.0 m span
beam and 22072 N mm for the 9.0 m span beam. This
represents the restraint from the unbraced structure
shown in Fig. 20.
Comparison between Figs. 9 and 21 shows that there
is nominal difference between the failure temperatures
of the column when infinite or finite horizontal restraint
is provided to the heated beams. However, only one
example is presented and further studies are required to
investigate a wider range of restraining stiffness.

3.5. Effect of beam span

The original analyses (Fig. 9) were repeated with the
9.0 m span beam increased to 15.0 m. Fig. 22 shows
that by increasing the span of the beam from 9.0 m to
15.0 m the column Limiting Temperature should be
reduced to 440C to maintain stability. This represents
a reduction of 100C, which will significantly increase
the amount of fire protection required for the column.
Fig. 20. Structural model to represent restraint to heated beams.
3.6. Effect of rotational restraint to the heated column
The original analyses (Fig. 9) were repeated with a
pinned connection assumed for the base of the heated
column. This effectively models a lower storey column

Fig. 19. Effect of rotational restraint on column failure temperatures.

766 C.G. Bailey / Engineering Structures 22 (2000) 755768

Fig. 21. Failure of structure with varying column-to-beam temperature ratios (spring stiffness used to represent horizontal restraint to heated
beams).

Fig. 22. Failure of structure for varying column-to-beam temperature ratios (beam spans 15.0 m and 6.0 m).

with a pinned base. The applied static load to the column
was reduced to 2728 kN, assuming an effective length
equal to the system length. Fig. 23 shows that the worst
scenario is when the column is heated at 40% of the
temperature of the beam. To ensure column stability the
Limiting Temperature of the column should be reduced
to 330C. This represents a reduction of 196C. How-
ever, it should be noted that a nominally pinned base
will provide some rotational restraint and it may be poss-
ible to increase the failure temperature of the column.
Further studies are required to investigate this possi-
bility.
3.7. Effect of column load ratio
The original analyses (Fig. 9) were repeated, with the
Load Ratio of the column varied between 0.7 and 0.2.
Fig. 24 shows that the worst case is when the column
is subjected to a Load Ratio of 0.7 (as expected since it
represents the highest axial load in the column). With
this level of applied static load the column Limiting
Temperature should be reduced from 510C to 425C.
As the Load Ratio becomes smaller the reduction in
Limiting Temperature also reduces. For the structural
layout used in this analyses, the required reduction in
 C.G. Bailey / Engineering Structures 22 (2000) 755768 767

Fig. 23. Failure of structure for varying column-to-beam temperature ratios (column base pinned).

Fig. 24. Failure of structure for varying column-to-beam temperature ratios and varying Load Ratios.

Limiting Temperature for a Load Ratio of 0.4 or less is
lower than 10C and can effectively be ignored.
4. Conclusions
Results from the Cardington full-scale fire tests indi-
cated that high moments occurred in the columns during
the test. At present these moments are not considered in
existing design methods. Investigation of the test results,
supplemented with computer models [10], concluded
that these column moments were caused by the thermal
expansion of the connecting beams.
To carry out a detailed investigation into the conse-
quence of these column moments, on the overall stability
of the column, a simple structural model was developed
(Fig. 8). This model represented a continuous column in
a multi-storey frame, with beams framing into the minor
and major axes of the column from one side only. It was
shown that the column was forced into double curvature
with high localised stresses occurring at the top and bot-
tom of the heated column. It was assumed that the col-
768 C.G. Bailey / Engineering Structures 22 (2000) 755768
umn was restrained laterally at floor level. In the simple
model this was achieved by restraining the heated
beams. In the Cardington frame the composite floor, due
to its good inherent fire resistance, will provide the
restraint.
The analyses presented showed that instability could
occur in the column, even though it was forced into dou-
ble curvature and restrained at floor level. This instability
was caused by the P effect in the column, which
was enhanced due to the enforced deflected shape of the
column caused by the expansion of the connecting
beams.
The analyses showed that column instability was sig-
nificantly affected by:
1. Beam to column heating rates
2. Beam cross-section size
3. Span of the beams
4. End rigidity of the heated column
5. Column axial load
In addition the analyses showed that the following
parameters had nominal effect on the behaviour of the
column:
6. Column cross-section size
7. Beam-to-column connection rigidity
8. Horizontal restraint to the heated beams (provided
realistic values are chosen)
At present, only a limited parametric study has been
produced. The results from this study have shown poss-
ible scenarios where column instability could occur. The
analyses indicate that in certain situations the column
Limiting Temperature (failure temperature) used in cur-
rent design methods should be reduced. The required
reduction will range from approximately 35C to 195C
depending on the parameters explained in (1)(5) above.
Acknowledgements
The author is grateful for the technical and general
support from Dr David Moore (Building Research
Establishment) and Dr Ian Burgess (University of
Sheffield).
References
[1] The behaviour of multi-storey steel framed buildings in fire: a
European Collaborative Research Programme. British Steel
(submitted for publication).
[2] Bailey CG, Lennon T, Moore DB. The behaviour of full-scale
steel-framed buildings subjected to compartment fires. The Struc-
tural Engineer 1999;77(8):1521.
[3] Martin DM, Moore DB. Introduction and background to the
research programme and major fire tests at BRE Cardington. In:
National Steel Construction Conference, London, May 1997.
[4] OConnor MA, Martin DM. Behaviour of multi-storey steel
framed building subjected to fire attack. Paper no. 169: Proceed-
ings of the Second World Conference on steel in Construction.
Journal of Constructional Steel Research 1998;46(1,3):31525.
[5] Association of Specialist Fire Protection. Fire protection for
structural steel in buildings, 2nd ed., revised. Ascot (UK): Steel
Construction Institute, 1992.
[6] British Standards Institution. BS5950: Structural use of steelwork
in building. Part 8: Code of practice for fire resistant design. Lon-
don: BSI, 1990.
[7] European Committee for Standardisation. Eurocode 3: Design of
steel structures. Part 1.2: General rules, structural fire design.
ENV 1993-1-2. ECS, Brussels, 1995.
[8] European Committee for Standardisation. Eurocode 4: Design of
composite steel and concrete structures. Part 1.2: General rules,
structural fire design. ENV 1994-1-2. ECS, Brussels, 1994.
[9] Armer GST, Moore DB. Full-scale testing on complete multisto-
rey structures. The Structural Engineer 1994;72(2):301.
[10] Bailey CG, Moore DB, Lennon T. The structural behaviour of
steel columns during a compartment fire in a multi-storey braced
steel frame. The Structural Engineer (submitted for publication).
[11] Bailey CG. Enhancement of fire resistance of beams by beam-
to-column connections. New Steel Construction 1998;6(2):44.
[12] Wang YC, Lennon T, Moore DB. The behaviour of steel frames
subject to fire. J Construct Steel Research 1995;35:291322.
[13] Bailey CG. The behaviour of steel column in fire. Document
RT524. Steel Construction Institute, Ascot, March 1996.