A New Approach to Multi-Storey Steel Framed Buildings Fire and Steel Construction.
A New Approach to Multi-Storey Steel Framed Buildings
Fire and Steel Construction.
The Steel Construction Institute.
3 THE CARDINGTON FIRE TESTS
3.1 Research programme
In September 1996, a programme of fire tests was completed in the UK at the
Building Research Establishment's Cardington Laboratory. The tests were
carried out on an eight-storey composite steel-framed building that had been
designed and constructed as a typical multi-storey office building. The purpose
of the tests was to investigate the behaviour of a real structure under real fire
conditions and to collect data that would allow computer programs for the
analysis of structures in fire to be verified.
The test building (see Figure A.1.1) was designed to be a typical example of
both the type of braced structure and the load levels that are commonly found in
the UK. In plan, the building covered an area of 21 m x 45 m and had an
overall height of 33 m. The beams were designed as simply supported acting
compositely with a 130 mm floor slab. Normally a building of this type would
be required to have 90 minutes fire resistance. Fin-plates were used for the
beam-to-beam connections and flexible end plates for the beam-to-column
connections. The structure was loaded using sandbags distributed over each
floor to simulate typical office loading.
There were two projects in the research programme. One project was funded
by Corus (formerly British Steel) and the European Coal and Steel Community
(ECSC), and the other was funded by the UK Government via the Building
Research Establishment (BRE). Other organisations involved in the research
programme included Sheffield University, TNO (The Netherlands), CTICM
(France) and The Steel Construction Institute. Fire tests took place between
January 1995 and July 1996. The tests were carried out on various floors; the
location of each test is shown on the floor plan in Figure B.3.1.
Figure B.3.1 Test locations
More details of the test building, including member sizes, are given in Appendix 3.
Test 1 involved a single secondary beam and the surrounding floor slab which
was heated by a purpose-built gas-fired furnace. Test 2 was also heated using
gas, and was conducted on a plane frame spanning across the building on one
floor; the test included primary beams and associated columns. Tests 3, 4 and
5 involved compartments of various sizes subjected, in each case, to a natural
fire fuelled by timber cribs. The columns in these tests were protected up to the
underside of the floor slab and the beams and floor slab were left unprotected.
The last, and most severe, test was a demonstration using furniture and contents
typically found in modern offices.
A detailed description of the tests has been published . The complete test data,
in electronic form with accompanying instrument location maps, is available for
Tests 1, 2, 3 and 6 from Corus RD&T (Swinden Technology Centre) and for
Tests 4 and 5 from BRE [14,15].
3.2 Test 1: Restrained beam
The test was carried out on the seventh floor of the building. A purpose-built
gas fired furnace, 8.0 m long and 3.0 m wide, was designed and constructed to
heat a secondary beam (D2/E2) spanning between two columns and part of the
surrounding structure. The beam was heated over the middle 8.0 m of its
9.0 m length, thus keeping the connections relatively cool. The purpose of the
test was to investigate the behaviour of a heated beam surrounded by an
unheated floor slab and study the restraining effect of the unheated parts of the
The beam was heated at between 3 and 10°C per minute until temperatures
approaching 900°C were recorded. At the peak temperature, 875ºC in the
lower flange, the mid span deflection was 232 mm (span/39) (see Figure B.3.2).
On cooling, the mid-span deflection recovered to 113 mm.
Figure B.3.2 Central displacement and maximum temperature in restrained beam test
The contrast between the behaviour of this beam and a similar unprotected beam
tested in the standard fire test under a similar load  is shown in Figure B.3.3.
The `runaway' displacement typical of simply supported beams in the standard
test did not occur to the beam in the building frame even though, at a
temperature of about 900ºC, structural steel retains only about 6% of its yield
strength at ambient temperature. This gives an indication that standard fire test
results are not a reliable measure of real building performance in fire.
Figure B.3.3 Central displacement and maximum temperature in standard fire test and restrained beam test
During the test, local buckling occurred at both ends of the test beam, just
inside the furnace wall (see Figure B.3.4).
Figure B.3.4 Flange buckling in restrained beam
Visual inspection of the beam after the test showed that the end-plate connection
at both ends of the beam had fractured near, but outside, the heat-affected zone
of the weld on one side of the beam. This was caused by thermal contraction of
the beam during cooling, which generated very high tensile forces. Although
the plate sheared down one side, this mechanism relieved the induced tensile
strains, with the plate on the other side of the beam retaining its integrity and
thus providing shear capacity to the beam. The fracture of the plate can be
identified from the strain gauge readings, which indicate that during cooling the
crack progressed over a period of time rather than by a sudden fracture.
3.3 Test 2: Plane frame
This test was carried out on a plane frame consisting of four columns and three
primary beams spanning across the width of the building (B1/B4).
A gas-fired furnace 21 m long x 2.5 m wide x 4.0 m high was constructed
using blockwork across the full width of the building.
The primary and secondary beams, together with the underside of the composite
floor, were left unprotected. The columns were fire protected to a height at
which a suspended ceiling might be installed (although no such ceiling was
present). This resulted in the top 800 mm of the columns, which incorporated
the connections, being unprotected.
The rate of vertical displacement at midspan of the 9 m span steel beam
increased rapidly between approximately 110 and 125 minutes (see Figure
B.3.5). This was caused by vertical displacements of its supporting columns.
The exposed areas of the internal columns squashed by approximately 180 mm
(see Figure B.3.6). The temperature of the exposed part of the column was
approximately 670°C when squashing commenced.
Figure B.3.5 Maximum vertical displacement of central 9 m beam and temperature of exposed top section of internal column
The squashing of the column resulted in all the floors above the fire
compartment being displaced downwards by approximately 180 mm. To avoid
this behaviour, columns in later tests were protected over their full height.
Figure B.3.6 Squashed column head following the test
On both sides of the primary beams, the secondary beams were each heated
over a length of approximately 1.0 m. After the test, investigation showed that
many of the bolts in the fin-plate connections had sheared (see Figure B.3.7).
The bolts had only sheared on one side of the primary beam. In a similar
manner to the fracturing of the plate in Test 1, the bolts sheared due to thermal
contraction of the beam during cooling. The thermal contraction generated very
high tensile forces, which were relieved once the bolts sheared in the fin-plate
on one side of the primary beam.
Figure B.3.7 Fin-plate connection following test
3.4 Test 3: Corner
The objective of this test was to investigate the behaviour of a complete floor
system and, in particular, the role of `bridging' or membrane action of the floor
in providing alternative load paths as the supporting beams lose strength. Using
concrete blockwork, a compartment 10 m wide x 7.6 m deep was constructed in
one corner of the first floor of the building (E2/F1).
To ensure that the compartment walls did not contribute to supporting the
applied loads, all the restraints and ties in the gable wall and the top layer of
blockwork were removed. The mineral fibre board in the expansion joints was
replaced with a ceramic blanket.
The wind posts were detached from the edge beam above the opening.
All columns, beam-to-column connections and edge beams were fire protected.
The fire load was 45 kg/m2, in the form of timber cribs. This fire load is quite
high and is equivalent to the 95% fractile for office buildings. Fire safety
engineering calculations are normally based on the 80% fractile. Ventilation
was provided by a single 6.6 m wide x 1.8 m high opening. The peak
atmosphere temperature recorded in the compartment was 1071ºC.
The maximum steel temperature was 1014ºC, in the inner beam E2/F2. The
maximum vertical displacement of 428 mm (just less than span/20) occurred at
the centre of the secondary beam, which had a peak temperature of 954ºC. On
cooling this recovered to a permanent displacement of 296 mm. The variations
of deflection and temperature with time are shown in Figure B.3.8.
Figure B.3.8 Maximum vertical displacement and temperature of secondary beam
All the combustible material within the compartment was consumed by the fire.
The structure behaved extremely well with no signs of collapse (see
Buckling occurred in the proximity of some of the beam-to-column connections
but unlike Test 2, bolts in the connections did not suffer shear failure. This
might indicate either that the high tensile forces did not develop or that the
connection had adequate ductility to cope with the tensile displacements.
Figure B.3.9 View of structure following test
3.5 Test 4: Corner
This test was carried out on the second floor, in a corner bay (E4/F3) with an
area of 54 m2. The internal boundaries of the compartment on gridlines E and
3 were constructed using steel stud partitions with fire resistant board. The stud
partition was specified to have 120 minutes fire resistance, with a deflection
head of 15 mm. An existing full-height blockwork wall formed the boundary
on the gable wall on gridline F; the outer wall, gridline 4, was glazed above
1 metre of blockwork. The compartment was totally enclosed with all windows
and doors closed. The columns were fire protected up to the underside of the
floor slab, including the connections but, unlike Test 3, the lintel beam (E4/F4)
was unprotected and the wind posts above it remained connected. Twelve
timber cribs were used to give a fire load of 40 kg/m2.
The development of the fire was largely influenced by the lack of oxygen within
the compartment. After an initial rise in temperature, the fire died down and
continued to smoulder until, after 55 minutes, the fire brigade intervened to vent
the compartment by removal of a single pane of glazing. This resulted in a
small increase in temperature followed by a decrease. A second pane,
immediately above the first, was broken at 64 minutes and temperatures began
to rise steadily; between 94 and 100 minutes the remaining panes shattered.
This initiated a sharp increase in temperature that continued as the fire
developed. The maximum recorded atmosphere temperature in the centre of the
compartment was 1051°C after 102 minutes (see Figure B.3.10). The
maximum steel temperature of 903°C was recorded after 114 minutes in the
bottom flange of the central secondary beam.
The maximum slab displacement was 269 mm and occurred in the centre of the
compartment after 130 minutes. This recovered to 160 mm after the fire.
The unprotected lintel beam, E4/F4, was observed during the test to be
completely engulfed in fire. However, the maximum temperature of this beam
was only 680°C, with a corresponding maximum displacement of 52 mm,
recorded after 114 minutes. This small displacement was attributed to the
additional support provided by wind posts above the compartment, which acted
in tension during the test. The tops of these wind posts were fixed by bolts
through slotted holes 80 mm long. Thus, a large proportion of the 52 mm
displacement could be attributed to the bolt slippage. Unfortunately, the
position of the bolts was not recorded before the test.
The internal compartment walls were constructed directly under unprotected
beams and performed well. Their integrity was maintained for the duration of
the test. On removal of the wall, it could be seen that one of the beams had
distortionally buckled over most of its length. This was caused by the high
thermal gradient through the cross section of the beam (caused by the
positioning of the compartment wall), together with high restraint to thermal
No local buckling occurred in any of the beams, and the connections showed
none of the characteristic signs of high tensile forces that were seen on cooling
in the other tests.
Figure B.3.10 Furnace gas temperatures recorded in Test 4
Figure B.3.11 Maximum flange temperature of internal beam and edge beam
3.6 Test 5: Large compartment
This test was carried out between the second and third floor, with the fire
compartment extending over the full width of the building, covering an area of
The fire load of 40 kg/m2 was provided by timber cribs arranged uniformly
over the floor area. The compartment was constructed by erecting a fire
resistant stud and plasterboard wall across the full width of the building and by
constructing additional protection to the lift shaft. Double glazing was installed
on two sides of the building, but the middle third of the glazing on both sides of
the building was left open. All the steel beams, including the edge beams, were
left unprotected. The internal and external columns were protected up to and
including the connections.
The ventilation condition governed the severity of the fire. There was an initial
rapid rise in temperature as the glazing was destroyed, creating large openings
on both sides of the building. The large ventilation area in two opposite sides
of the compartment gave rise to a fire of long duration but lower than expected
temperatures. The maximum recorded atmosphere temperature was 746°C,
with a maximum steel temperature of 691°C, recorded at the centre of the
compartment. The structure towards the end of the fire is shown in Figure
The maximum slab displacement reached a value of 557 mm. This recovered to
481 mm when the structure cooled.
Extensive local buckling occurred in the proximity of the beam-to-beam
connections. On cooling, a number of the end-plate connections fractured down
one side. In one instance the web detached itself from the end-plate so that the
steel-to-steel connection had no shear capacity. This caused large cracks within
the composite floor above this connection, but no collapse occurred, with the
beam shear being carried by the composite floor slab.
Figure B.3.12 Deformed structure during fire
Figure B.3.13 Maximum and average recorded atmosphere
3.7 Test 6: The office demonstration test
The aim of this test was to demonstrate structural behaviour in a realistic fire
A compartment 18 m wide and up to 10 m deep with a floor area of 135 m2,
was constructed using concrete blockwork. The compartment represented an
open plan office and contained a series of work-stations consisting of modern
day furnishings, computers and filing systems (see Figure B.3.14). The test
conditions were set to create a very severe fire by incorporating additional
wood/plastic cribs to create a total fire load of 46 kg/m2 (less than 5% of offices
would exceed this level) and by restricting the window area to the minimum
allowed by regulations for office buildings. The fire load was made up of 69%
wood, 20% plastic and 11% paper. The total area of windows was 25.6 m2
(19% of the floor area) and the centre portion of each window, totalling
11.3 m2, was left unglazed to create the most pessimistic ventilation conditions
at the start of the test.
Figure B.3.14 Office before test
Within the compartment, the columns and the beam-to-column connections were
fire protected. Both the primary and secondary beams, including all the
beam-to-beam connections, remained totally exposed.
The wind posts were left connected to the edge beams, and thus gave some
support during the fire.
The maximum atmosphere temperature was 1213°C and the maximum average
temperature was approximately 900°C (see Figure B.3.15). The maximum
temperature of the unprotected steel was 1150°C. The maximum vertical
displacement was 640 mm, which recovered to 540 mm on cooling (see Figure
B.3.16). The peak temperature of the lintel beams, above the windows, was
813ºC. All the combustible material in the compartment was completely burnt,
including the contents of the filing cabinets. Towards the back of the
compartment, the floor slab deflected and rested on the blockwork wall. The
structure showed no signs of failure.
An external view of the fire near its peak is shown in Figure B.3.17. The
structure following the fire is shown in Figure B.3.18 and Figure B.3.19.
Figure B.3.18 shows a general view of the burned out compartment and Figure
B.3.19 shows the head of one of the columns. During the test, the floor slab
cracked around one of the column heads (see Figure B.3.20). This behaviour is
discussed in Section 5.1.
Figure B.3.15 Measured atmosphere temperature
Figure B.3.16 Maximum steel temperature and vertical displacement
Figure B.3.17 External view of fire
Figure B.3.18 Compartment following fire
Figure B.3.19 Column head showing buckled beams
Figure B.3.20 Cracked floor slab in region of non-overlapped mesh
3.8 General comments on observed behaviour in
In all tests, the structure performed very well and overall structural stability was
The performance of the whole building in fire is manifestly very different from
the behaviour of single unrestrained members in the standard fire test. It is
clear that there are interactions and changes in load-carrying mechanisms in real
structures that dominate the way they behave; it is entirely beyond the scope of
the simple standard fire test to reproduce or assess such effects.
The Cardington tests demonstrated that modern steel frames acting compositely
with steel deck floor slabs have a coherence that provides a resistance to fire far
greater than that normally assumed. This confirms evidence from other
4 EVIDENCE FROM OTHER FIRES AND
In recent years, two building fires in England (Broadgate and Churchill Plaza)
have provided the opportunity to observe how modern steel-framed buildings
perform in fire. The experience from these fires was influential in stimulating
thought about how buildings might be designed to resist fire and in bringing
about the Cardington experiments.
Evidence of building behaviour is also available from large-scale fire tests in
Australia and Germany. In both Australia and New Zealand, design approaches
that allow the use of unprotected steel in multi-storey, steel-framed buildings
have been developed.
In 1990, a fire occurred in a partly completed 14-storey office block on the
Broadgate development in London . The fire began inside a large site hut on
the first level of the building. Fire temperatures were estimated to have reached
The floor was constructed using composite long-span lattice trusses and
composite beams supporting a composite floor slab. The floor slab was
designed to have 90 minutes fire resistance. At the time of the fire, the
building was under construction and the passive fire protection to the steelwork
was incomplete. The sprinkler system and other active measures were not yet
After the fire, a metallurgical investigation concluded that the temperature of the
unprotected steelwork was unlikely to have exceeded 600°C. A similar
investigation on the bolts used in the steel-to-steel connections concluded that
the maximum temperature reached in the bolts, either during manufacture or as
a consequence of the fire, was 540°C.
The distorted steel beams had permanent deflections of between 270 mm and
82 mm. Beams with permanent displacements at the higher end of that range
showed evidence of local buckling of the bottom flange and web near their
supports. From this evidence, it was concluded that the behaviour of the beams
was influenced strongly by restraint to thermal expansion. This restraint was
provided by the surrounding structure, which was at a substantially lower
temperature than the fire-affected steel. Axial forces were induced into the
heated beams resulting in an increase in vertical displacement due to the P-delta
effect. The buckling of the lower flange and web of the beam near its supports
was due to a combination of the induced axial force and the negative moment
caused by the fixity of the connection.
Although the investigation showed the visually unfavourable effects of restraint
on steel beams, the possible beneficial effects were not evident because only
relatively low steel temperatures were reached during the fire. The beneficial
effects that could have developed were catenary action of the beams and
bridging or membrane action of the composite slab.
The fabricated steel trusses spanned 13.5 m and had a maximum permanent
vertical displacement of 552 mm; some truss elements showed signs of
buckling. It was concluded that the restraint to thermal expansion provided by
other elements of the truss, combined with non-uniform heating, caused
additional compressive axial forces, which resulted in buckling.
At the time of the fire, not all the steel columns were fire protected. In cases
where they were unprotected, the column had deformed and shortened by
approximately 100 mm (see Figure B.4.1). These columns were adjacent to
much heavier columns that showed no signs of permanent deformation. It was
thought that this shortening was a result of restrained thermal expansion. The
restraint to thermal expansion was provided by a rigid transfer beam at an upper
level of the building, together with the columns outside the fire affected area.
Figure B.4.1 Buckled column and deformed beams at Broadgate
Although some of the columns deformed, the structure showed no signs of
collapse. It was thought that the less-affected parts of the structure were able to
carry the additional loads that were redistributed away from the weakened areas.
Following the fire, the composite floor suffered gross deformations with a
maximum permanent vertical displacement of 600 mm (see Figure B.4.2). Some
failure of the reinforcement was observed. In some areas, the steel profiled
decking had debonded from the concrete. This was considered to be caused
mainly by steam release from the concrete, together with the effects of thermal
restraint and differential expansion.
A mixture of cleat and end-plate connections was used. Following the fire,
none of the connections was observed to have failed although deformation was
evident. In cleated connections, there was some deformation of bolt holes. In
one end-plate connection, two of the bolts had fractured; in another, the plate
had fractured down one side of the beam but the connection was still able to
transfer shear. The main cause of deformation was thought to be due to the
tensile forces induced during cooling.
Following the fire, structural elements covering an area of approximately
40 m x 20 m were replaced, but it is important to note that no structural failure
had occurred and the integrity of the floor slab was maintained during the fire.
The direct fire loss was in excess of £25M, of which less than £2M was
attributed to the repair of the structural frame and floor damage; the other costs
resulted from smoke damage. Structural repairs were completed in 30 days.
Figure B.4.2 View of deformed floor above the fire (the maximum deflection was about 600 mm)
4.2 Churchill Plaza Building, Basingstoke
In 1991, a fire took place in the Mercantile Credit Insurance Building, Churchill
Plaza, Basingstoke. The 12-storey building was constructed in 1988. The
columns had board fire protection and the composite floor beams had spray-
applied protection. The underside of the composite floor was not fire protected.
The structure was designed to have 90 minutes fire resistance.
The fire started on the eighth floor and spread rapidly to the ninth and then the
tenth floor as the glazing failed. During the fire, the fire protection performed
well and there was no permanent deformation of the steel frame. The fire was
believed to be comparatively `cool' because the failed glazing allowed a cross
wind to increase the ventilation. The protected connections showed no
In places, the dovetail steel decking showed some signs of debonding from the
concrete floor slab. This had also been observed in the Broadgate fire. A load
test was conducted on the most badly affected area. A load of 1.5 times the
total design load was applied. The test showed that the slab had adequate load-
carrying capacity and could be reused without repair.
The protected steelwork suffered no damage. The total cost of repair was in
excess of £15M, most of which was due to smoke contamination, as in the
Broadgate fire. Sprinklers were installed in the refurbished building.
Figure B.4.3 Churchill Plaza, Basingstoke following the fire
4.3 Australian fire tests
BHP, Australia's biggest steel maker, has been researching and reporting [18,19]
fire-engineered solutions for steel-framed buildings for many years. A number
of large-scale natural fire tests has been carried out in specially constructed
facilities at Melbourne Laboratory, representing sports stadia, car parks and
offices. The office test programme focussed on refurbishment projects that
were to be carried out on major buildings in the commercial centre of
4.3.1 William Street fire tests and design approach
A 41-storey building in William Street in the centre of Melbourne was the
tallest building in Australia when it was built in 1971. The building was square
on plan, with a central square inner core. A light hazard sprinkler system was
provided. The steelwork around the inner core and the perimeter steel columns
were protected by concrete encasement. The beams and the soffit of the
composite steel deck floors were protected with asbestos-based material.
During a refurbishment programme in 1990, a decision was made to remove the
The floor structure was designed to serviceability rather than strength
requirements. This meant that there was a reserve of strength that would be
very beneficial to the survival of the frame in fire, as higher temperatures could
be sustained before the frame reached its limiting condition.
At the time of the refurbishment, the required fire resistance was 120 minutes.
Normally this would have entailed the application of fire protection to the steel
beams and to the soffit of the very lightly reinforced slab (Australian regulations
have been revised and now allow the soffit of the slab to be left unprotected for
120 minutes fire resistance). In addition, the existing light hazard sprinkler
system required upgrading to meet the prevailing regulations.
During 1990, the fire resistance of buildings was subject to national debate; the
opportunity was therefore taken to conduct a risk assessment to assess whether
fire protecting the steelwork and upgrading the sprinkler system was necessary
for this building. Two assessments were made. The first was made on the
basis that the building conformed to current regulations with no additional safety
measures; the second was made assuming no protection to the beams and soffit
of the slab, together with the retention of the existing sprinkler system. The
effect of detection systems and building management systems were also included
in the second assessment. The authorities agreed that if the results from the
second risk assessment were at least as favourable as those from the first
assessment, the use of the existing sprinkler system and unprotected steel beams
and composite slabs would be considered acceptable.
A series of four fire tests was carried out to obtain data for the second risk
assessment. The tests were to study matters such as the probable nature of the
fire, the performance of the existing sprinkler system, the behaviour of the
unprotected composite slab and castellated beams subjected to real fires, and the
probable generation of smoke and toxic products.
The tests were conducted on a purpose-built test building at the Melbourne
Laboratories of BHP Research (see Figure B.4.4). This simulated a typical
storey height 12 m x 12 m corner bay of the building. The test building was
furnished to resemble an office environment with a small, 4 m x 4 m, office
constructed adjacent to the perimeter of the building. This office was enclosed
by plasterboard, windows, a door, and the facade of the test building. Imposed
loading was applied by water tanks.
Figure B.4.4 BHP test building and fire test
Four fire tests were conducted. The first two were concerned with testing the
performance of the light hazard sprinkler system. In Test 1, a fire was started
in the small office and the sprinklers were activated automatically. This office
had a fire load of 52 kg/m2. The atmosphere temperatures reached 60°C before
the sprinklers controlled and extinguished the fire. In Test 2, a fire was started
in the open-plan area midway between four sprinklers. This area had a fire
load of 53.5 kg/m2. The atmosphere temperature reached 118°C before the
sprinklers controlled and extinguished the fire. These two tests showed that the
existing light hazard sprinkler system was adequate.
The structural and thermal performance of the composite slab was assessed in
Test 3. The supporting beams were partially protected. The fire was started in
the open plan area and allowed to develop with the sprinklers switched off. The
maximum atmosphere temperature reached 1254°C. The fire was extinguished
once it was considered that the atmosphere temperatures had peaked. The slab
supported the imposed load. The maximum temperature recorded on the top
surface of the floor slab was 72°C. The underside of the slab had been
partially protected by the ceiling system, which remained substantially in place
during the fire.
In Test 4, the steel beams were left unprotected and the fire was started in the
small office. The fire did not spread to the open-plan area despite manual
breaking of windows to increase the ventilation. Therefore fires were ignited
from an external source in the open-plan area. The maximum recorded
atmosphere temperature was 1228°C, with a maximum steel beam temperature
of 632°C above the suspended ceiling. The fire was extinguished when it was
considered that the atmosphere temperatures had peaked. Again, the steel
beams and floor were partially shielded by the ceiling. The central
displacement of the castellated beam was 120 mm and most of this deflection
was recovered when the structure cooled to ambient temperature.
Three unloaded columns were placed in the fire compartment to test the effect
of simple radiation shields. One column was shielded with galvanized steel
sheet, one with aluminised steel sheet and one was an unprotected reference
column. The maximum recorded column temperatures were 580ºC, 427ºC and
1064ºC respectively, suggesting that simple radiation shields might provide
sufficient protection to steel members in low fire load conditions.
It was concluded from the four fire tests that the existing light hazard sprinkler
system was adequate and that no fire protection was required to the steel beams
or soffit of the composite slab. Any fire in the William Street building should
not deform the slab or steel beams excessively, provided that the steel
temperatures do not exceed those recorded in the tests.
The temperature rise in the steel beams was affected by the suspended ceiling
system, which remained largely intact during the tests.
The major city centre office building that was the subject of the technical
investigation was owned by Australia's largest insurance company, which had
initiated and funded the test programme. It was approved by the local authority
without passive fire protection to the beams but with a light hazard sprinkler
system of improved reliability and the suspended ceiling system that had proved
to be successful during the test programme.
4.3.2 Collins Street fire tests
This test rig was constructed to simulate a section of a proposed steel-framed
multi-storey building in Collins Street, Melbourne. The purpose of the test was
to record temperature data in fire resulting from combustion of furniture in a
typical office compartment.
The compartment was 8.4 m x 3.6 m and filled with typical office furniture,
which gave a fire load between 44 and 49 kg/m2. A non-fire-rated suspended
ceiling system was installed, with tiles consisting of plaster with a fibreglass
backing blanket. An unloaded concrete slab formed the top of the
compartment. During the test, temperatures were recorded in the steel beams
between the concrete slab and the suspended ceiling. The temperatures of three
internal free-standing columns were also recorded. Two of these columns were
protected with aluminium foil and steel sheeting, acting simply as a radiation
shield; the third remained unprotected. Three unloaded external columns were
also constructed and placed 300 mm from the windows around the perimeter of
The non-fire-rated ceiling system provided an effective fire barrier, causing the
temperature of the steel beams to remain low. During the test the majority of
the suspended ceiling remained in place. Atmosphere temperatures below the
ceiling ranged from 831°C to 1163°C, with the lower value occurring near the
broken windows. Above the ceiling, the air temperatures ranged from 344°C to
724°C, with higher temperatures occurring where the ceiling was breached.
The maximum steel beam temperature was 470°C.
The unloaded indicative internal columns reached a peak temperature of 740°C
for the unprotected case and below 403°C for the shielded cases. The bare
external columns recorded a peak temperature of 490°C.
This fire test showed that the temperatures of the beams and external columns
were sufficiently low to justify the use of unprotected steel and, as in the
William Street tests, the protection afforded by a non-fire-rated suspended
ceiling was beneficial. The role of ceilings is discussed further in
4.3.3 Conclusions from Australian research
The Australian tests and associated risk assessments concluded that, provided
that high-rise office buildings incorporate a sprinkler system with a sufficient
level of reliability, the use of unprotected beams would offer a higher level of
life safety than similar buildings that satisfied the requirements of the Building
Code of Australia by passive protection. Up to the beginning of 1999, six such
buildings between 12 and 41 storeys were approved in Australia.
4.4 German fire test
In 1985, a fire test was conducted on a four storey steel-framed demonstration
building constructed at the Stuttgart-Vaihingen University in Germany .
Following the fire test, the building was used as an office and laboratory.
The building was constructed using many different forms of steel and concrete
composite elements. These included water filled columns, partially encased
columns, concrete filled columns, composite beams and various types of
The main fire test was conducted on the third floor, in a compartment covering
approximately one-third of the building. Wooden cribs provided the fire load
and oil drums filled with water provided the gravity load. During the test, the
atmosphere temperature exceeded 1000°C, with the floor beams reaching
temperatures up to 650°C. Following the test, investigation of the beams
showed that the concrete-infilled webs had spalled in some areas exposing the
reinforcement. However, the beams behaved extremely well during the test
with no significant permanent deformations following the fire. The external
columns and those around the central core showed no signs of permanent
deformation. The composite floor reached a maximum displacement of 60 mm
during the fire and retained its overall integrity.
Following the fire, the building was reinstated. The refurbishment work
involved the complete replacement of the fire damaged external wall panels, the
damaged portions of steel decking to the concrete floor slab, and the concrete
infill to the beams. Overall, it was shown that refurbishment to the structure
was economically possible.
4.5 New Zealand design approach
The New Zealand Heavy Engineering Research Association (HERA) has
published a draft guide on the design of multi-storey buildings . The guide is
more ambitious in its scope than the guidance presented in Part A of this
publication but is more conservative in some areas. In particular, it requires the
ends of primary beams to be fire protected. (This advice is not included in the
Part A recommendations because, although the bottom flange of some beams
buckled at Cardington, no collapse occurred, and because finite element
modelling by many researchers has shown that preserving the bending continuity
of beams is not vital.)
The HERA guide details a method of design that allows the capacity of an
unprotected composite floor system, such as that used in the Cardington tests, to
be checked. The design method allows each individual element to be checked
to ensure that it is capable of transferring the loads to which it is likely to be
subjected at elevated temperatures. Useful guidance on the detailing of elements
in order to allow the composite action to be fully developed is also given.
4.5.1 Summary of HERA design method
The design method is based on the prediction of atmospheric temperatures based
on a realistic pattern of fire growth and spread. The method uses the
parametric temperature-time curves that are modified versions of those given in
the Eurocode, ENV 1991-2-2 . The time-temperature relationship for a fully
developed fire can be calculated, given the dimensions of the compartment, the
thermal conductivity of the linings, the area of openings and the fire load energy
density. These relationships are applied to design areas within which the fire
conditions are considered to exist. Defined rules allow the spread of fire
around the compartment to be modelled.
Methods of accounting for the beneficial effects of ceilings in shielding the
unprotected steelwork from fire are presented, based on the performance of the
BHP William Street tests. The design guidance is based on indicative times for
which the suspended ceiling would remain in place under fully developed fire
The temperature-time relationship of the steel is determined from the net heat
flux from the fire to the surface of the steel member, including thermal
convection and radiation based on ENV 1991-2-2. The temperature of the
bottom flange of the steel beams is calculated first; temperatures of other steel
components are calculated relative to these temperatures.
The load-carrying capacity of the slab and secondary beam system at elevated
temperature is calculated using a combination of yield line analysis and tensile
membrane action. The capacity given by the yield line analysis will usually not
be sufficient to resist the design load in fire conditions, so the strength
enhancement due to the development of tensile membrane action is utilised.
The strength enhancement is a function of the deflection in the slab. As the
required strength enhancement is known, this is used to determine the slab
deflection at elevated temperature. Limits are applied to the level of strength
enhancement obtained from membrane action and the maximum allowable
deflection, in order to prevent undesirable failure modes such as fracture of
An important feature of the method is that the floor slab should be reinforced
with ductile reinforcement. It is assumed that all reinforcement, including
shrinkage and temperature mesh, fulfils an integral role in maintaining load-
carrying capacity. The slab reinforcement must therefore be capable of a
minimum of 10% elongation. A mesh of 6 mm diameter bars at 150 mm
centres, giving an area of 188 mm2/m, is recommended. The mesh used is
normally known in New Zealand as `seismic mesh'. Although its minimum
elongation is 10% it is said to have an average elongation at failure of 20%.
Reinforcement must be properly lapped.
Columns are required to be fully fire protected, as are the ends and connections
of primary beams.
In fire conditions, the column is considered to be subject to normal gravity
loading, plus some out-of-balance moments generated by different magnitudes of
fire-induced rotation at the two primary connections together with some thermal
induced axial force because of restraint. As the columns are fully fire-
protected, their strength will remain close to their ambient temperature capacity
and the thermal expansion will be low; the induced force due to restraint will
therefore be small. No specific fire limit state design rules are given for
columns. It is considered that columns that resist the ambient temperature
conditions will be adequate under the reduced load of the fire limit state.
The main conclusion that can be drawn from the New Zealand approach is that
a complete methodology has been developed that includes both fire and
structural behaviour. The total approach has not been checked by the authors
and cannot therefore be endorsed. Some elements of it are, as yet, unproven
and parts of the New Zealand method (protecting the ends of beams) are
considered unnecessary. There is no doubt, however, that the existence of a
complete, verified, procedure of this type would be a great benefit and will be
one of the objectives of the next stage of development (Level 2) in the UK.
5 OBSERVATIONS AND COMMENTS
From the observed structural behaviour at Cardington, Broadgate and other
large-scale natural fires, some general observations can be made. These
observations and evidence from mathematical modelling have led to the
recommendations in Part A.
Modelling the behaviour of the structure in the Cardington tests has been
undertaken by several universities. This work is described in papers published
by The University of Sheffield , The University of Edinburgh with Corus,
Swinden Technology Centre and Imperial College , and BRE .
5.1 Floor slabs
Observations from Cardington and building fires such as Broadgate have shown
that the composite slab plays a crucial role in fire. It fulfils its separating
functions by preventing the spread of fire. It is often the main load-carrying
element, and it assists greatly in the maintenance of structural integrity through
in-plane diaphragm action. This was seen clearly in the Broadgate fire where,
despite vertical displacements of about 600 mm, horizontal displacements were
The composite floors in the Cardington frame suffered extensive cracking as a
result of the fire tests, however their separating function was generally
maintained during the tests. In Test 6, large cracks occurred around the
internal column (see Figure B.3.20), which created gaps in the floor slab. The
evidence suggests that these cracks occurred during cooling, possibly when the
steel connection below the slab partially failed. Investigation of the slab after
the test showed that the reinforcement had not been lapped correctly, and was
just `butted' together.
The slab resists vertical loads in two ways. For normal design, and also in the
established methods of fire design, e.g. BS5950-8 , the slab is assumed to act
as a one-way spanning beam, resisting loads through bending and shear. The
slab normally spans between the beams. Evidence from fire resistance tests and
from some actual fires indicates that the bending tension is carried more by the
reinforcement than by the profiled steel deck. At Cardington and Broadgate,
the beams were not fire protected. This had the effect of greatly increasing the
distance the floor slab was spanning, but no collapse occurred. The reason for
this is complex, and is still the subject of research, but it is clear that in some
of the Cardington tests the slab acted as a membrane and was supported by the
colder perimeter beams and protected columns.
Although composite slabs are normally considered to span in one direction
between the steel beams, their behaviour in fire conditions appears to be very
different. At ambient temperature, the strength of reinforced concrete slabs is
often greater than just its flexural resistance and is enhanced by in-plane load-
carrying capacity, which can be utilised after initial flexural yielding has
occurred. The first type of in-plane action is compressive membrane action,
which depends on the slab having in-plane restraint at its edges. Compressive
membrane action occurs immediately after yielding, when deflections are still
relatively small; it is very sensitive to the amount of in-plane restraint to the
slab. The second form of in-plane action is tensile membrane action, which can
occur in slabs with fixed or simply supported edge conditions.
In the case of fixed edge conditions, compressive membrane action occurs first,
followed by tensile membrane action when deflections become greater. The
tensile forces that develop in the reinforcement need to be anchored at the slab
supports. In the case of simply supported edges, the supports will not anchor
the tensile forces and a compressive ring beam will form around the edge of the
slab. When tensile membrane action develops, failure occurs at large
displacements, with fracture of the reinforcement.
Observations of the behaviour of the floor slab in the Cardington tests showed
that initially, as the steel beams lost their load-carrying capacity, the composite
slab utilised its full bending capacity in spanning between the adjacent cooler
members. As displacements increased, the slab acted as a tensile membrane,
carrying loads in the reinforcement.
All the structural mechanisms working in the slab rely on the mesh retaining its
structural integrity. At some locations in the Cardington building, it was
observed that the mesh fractured and there was evidence that it had not been
placed properly during construction. Observations following Test 6 showed that,
in at least one location, adjacent mats appeared to have been butted up to each
other rather than being lapped. In such places, the slab sheared locally but no
In order to mobilise the maximum load resistance of the slab, all the structural
mechanisms require the mesh to have a degree of ductility. This allows
redistribution of internal forces, resulting in a plastic mode of behaviour. The
evidence from Cardington, fire resistance tests and Broadgate is that there is
normally sufficient ductility, but the poor mesh placement in one area at
Cardington could have initiated a failure. Better site control is required.
In the future, a new type of mesh might be available with much greater
ductility. Such a mesh is already used in New Zealand, where it is required for
earthquake resistance. Although the Cardington tests and the Broadgate fire
have shown that the mesh used in UK appears to be adequate, it is likely that
improved ductility would give greater reliability.
5.1.3 Modelling of floor slabs in fire
The Building Research Establishment has developed a simple structural
model [7,8] that combines the residual strength of the steel composite beams with
the slab strength calculated using a combined yield line and membrane action
model. The model has been calibrated against the Cardington fire tests and
other test results on slabs. In carrying out this process, some assumptions have
been made that are thought to be conservative. The Design Tables presented in
Part A are based on the BRE model. It is expected that, in time, the model will
be revised to take into account other work being carried out [23,24]. In the future,
less conservative design information might be available.
The Design Tables include 150 x 150 mm and 100 x 100 mm square meshes.
These meshes have some benefits over the standard 200 mm square mesh when
used in composite floors. Cracks at the serviceability limit state are reduced
and there is increased health and safety for site personnel casting the concrete
(i.e. easier to walk on).
Design points - floor slabs
Floor slabs should be checked as part of floor design zones and their strength
should be verified using the design tables in Part A. Alternatively, they can be
checked using the BRE method [7,8]. Mesh reinforcement should be lapped
5.2 Steel beams
At both Cardington and Broadgate, the steel beams suffered gross deformation
without collapse. Unprotected non-composite and composite beams may reach
temperatures in excess of 1000°C in fires. At these temperatures, the bending
resistance of the cross section could be reduced by 95% and thus, unless the
applied loading is extremely low, the beam effectively ceases to act as a bending
element. If collapse is to be avoided, an alternative load transfer mechanism is
Steel beams that are protected to achieve the fire resistance required by building
regulations would not generally be expected to reach temperatures at which they
would fail in bending. Most beams would be protected so that, in a fire
resistance test, their temperature would not exceed about 620°C at the end of
the required time period. At this temperature, the beam could be expected to
have 60% of its original strength; in all but the most exceptional conditions, it
would carry the applied loads. Evidence from actual building fires shows that
the beam deformation would probably be small; reuse may sometimes be
In the majority of cases, beams are designed as simple beams and the effects of
rotational continuity are ignored. In fire, the behaviour of a beam is affected by
its end conditions. Some rotational restraint may be generated but, more
importantly, the beam may also be restrained from expanding by the other parts
of the structure.
Normally, rotational restraint is beneficial as it causes redistribution of
moments, leading to a reduction in the maximum mid-span bending moment
occurring in the beam, increasing the fire survival time. However, it was
evident from the Cardington fire tests that, due to local buckling near the beam
ends, which does not occur in unrestrained standard fire tests, any beneficial
effects of rotational restraint cannot be relied upon.
The effects of thermal expansion are unavoidable. A fully restrained piece of
steel will yield when heated to temperatures between 100°C and 150°C,
depending on the grade. A beam will be subject to both axial and rotational
restraint (even if designed to be simply supported). A beam of 9 m span heated
to 900°C will try to expand 100 mm. If the temperature distribution across the
depth of the beam is non-uniform, it will also bend towards the hotter, lower
side. The effect of any axial restraint is to induce compression into the beam.
The effect of non-uniform heating, coupled with the rotational restraint, is to
induce hogging (negative moments) which also tends to add to the compressive
stresses in the lower part of the beam at its supports. These compressive
stresses are in addition to those caused by rotational restraint at ambient
conditions. The lower part of the web and/or the bottom flange of the beam
may yield plastically in compression and then buckle, as observed on some of
the beams in the Cardington tests.
Until methods that can predict the effects of buckling and the consequent effect
on negative bending resistance are developed, unprotected beams should be
treated as simply supported members in any design assessment.
From examination of protected beams at Cardington and from observations of
beams following actual building fires, protected beams do not appear to suffer
from local buckling and continuity may be assumed.
In the Cardington tests, unprotected beams acting compositely with the floor
slab have not been observed to separate from the slab or to slip excessively
relative to the slab. From this, it can be deduced that the shear connectors
generally perform reasonably well. There was evidence in one of the
Cardington tests of the failure of one shear connector on one beam, but it did
not lead to a progressive (unzipping) failure of connectors. Generally, the
degree of shear connection provided in the composite beams at Cardington was
low, leading to the conclusion that degree of shear connection is not a factor
that may affect performance in fire. Evidence from other fires confirms this.
The effect of applied load on the deformation of the floors at Cardington has
been investigated using finite element techniques. Some studies have shown that
the deformation of the structure may be independent of the applied loading .
This is because the dominant loads and deformations are caused by thermal
expansion. The modelling has shown that the comparatively low loads applied
in the Cardington tests may not have contributed significantly to the good
performance in the fire tests and that the conclusions can be applied safely to
the slightly higher, more usually, assumed levels of design loading.
5.2.1 Edge beams
Edge beams have the additional role of supporting external cladding. Cladding
normally falls into one of two types: masonry/brick, which has an appreciable
self weight; and curtain walling, which is less heavy. The consequences of
damage are different in each case. Another factor that must be considered in
assessing consequences of any damage is the number of storeys.
In some tests, the edge beams were unprotected. In these cases, the wind posts
acting in tension above the fire compartment were beneficial, resulting in very
small vertical displacements of the beams. The design recommendations
therefore recognise the role of vertical ties (wind posts) as a means of
supporting edge beams.
Masonry cladding is usually supported at each storey from the edge of the floor
slab, with the weight being taken by the edge beam. Curtain walling is
normally designed to span between floors and is supported from the floor slab at
each floor level. With some types of curtain walling, wind posts are built into
the mullions to support the wind pressure on the face of the wall. The posts are
then connected to the floor slab or the edge beam. This was the case in the
Cardington building. It was also the case at Cardington that the wind posts
acted as vertical ties, supporting the edge beam and preventing significant
vertical deflection, although some twisting of the beams did occur. In the tests,
wind posts were shown to be as effective as applied protection in reducing
During some of the fire tests, the wind posts at Cardington carried a tensile
stress of about 50 N/mm2. This stress was low because the posts were designed
to resist bending.
Mathematical modelling in which either vertical ties or fire protection to the
edge beams has been omitted has demonstrated that the effect on time to
structural failure of the floor system is small. The effect of edge beam
deformation on the cladding system is thought to be the most important
All the Cardington natural fire tests (Tests 3, 4, 5 and 6) showed that the
temperatures of unprotected steel beams close to openings were lower than those
in the body of the compartment.
Visual evidence of a cool `window zone' was provided by Test 5. Although
atmosphere temperatures were not measured within 6 m of the windows, it was
clear from the appearance of the galvanized steel decking at ceiling level that
lower temperatures had been experienced within one metre of the windows. In
this zone, on both sides of the compartment, the galvanized decking was hardly
discoloured whilst beyond one metre the surface was heavily oxidised.
Temperatures of the upper surface of the profiled steel decking taken through
the slab at 5 m and 10.5 m from the windows reached levels above the melting
point of zinc (420ºC) whilst close to the window, at 0.3 m, only 298ºC was
recorded. This can be seen in Figure B.5.4. In the figure, the lighter colour of
the steel decking (towards the left of the figure) shows areas where the
galvanizing is unaffected by heat. The effect of the cool window zone on
the galvanized steel decking
Lintel beams above windows are subject to asymmetric heating. Heat radiating
from surfaces and flames inside the compartment only reaches the interior face
of the beam. The exterior face is subject to a far lower level of radiation from
the flames that issue from the windows and these flames may also be cooled by
cold air being drawn into the compartment. The test data indicate that the
temperature rise of lintel beams is about 25% less than that of fully exposed
internal beams (see Figure B.5.5).
Figure B.5.5 Lintel beam temperature rises in Cardington test fires
were about 25% cooler than those for internal beams
It is not easy to quantify the effect of window openings in reducing
temperatures because it depends on the severity of the fire. Conservatively, it
may be assumed that lintel beams above window openings are not heated to the
same extent as similar internal beams, and a 50°C reduction in temperature may
safely be assumed in assessing the thickness of any fire protection or the
strength of the beam. However, when using tabulated data for fire protection
materials, the beam temperature must be increased by 50°C to obtain a
reduction in thickness. This is because the thickness of fire protection given in
tables is sufficient to keep the steel temperature below a given temperature for a
given time. A reduction in thickness can only be obtained if the table
appropriate to a higher temperature is used.
5.2.2 Masonry cladding
Falling masonry is a hazard to firefighters and others in the vicinity of tall
buildings. As masonry will be supported from the edge beams or the edge of
the floor slab, there is less chance of dislodgement if the deformation at the
edge of the building is reduced. Deformation will be restricted if the edge
beams are either protected or supported with vertical ties: this will normally
occur as a consequence of the design recommendations in Part A.
5.2.3 Curtain walling
Glazed curtain walling does not perform well in fire and will frequently break
up (because of heat rather than movement of its support), allowing glass and
other debris to fall to the ground. Controlling the deformation of the edge of
the slab and edge beam is likely to have little effect on the degradation of most
curtain walling systems.
In the rare cases where fire resisting glass is used, its performance should not
be compromised by distortion of its support. It is sensible to control
deformation of the edge beam by protection or by the use of vertical ties, as for
Design points - beams
Beams contained within floor design zones may be unprotected.
When designing unprotected beams (in fire), no effect of continuity should be
taken into account.
In assessing strength or fire protection thickness, edge beams above windows
may be assumed to be 50°C cooler than a similar beam within the compartment.
Unprotected columns, like unprotected beams, may reach temperatures in excess
of 900°C in fires. At these temperatures, the buckling resistance will be
reduced to virtually zero, so any support to the floors above will be lost.
Although local squashing of columns was observed at Cardington and in the
Broadgate fire, no collapse occurred because the structure had the ability to
carry the column loads using alternative load paths. It is very likely that the
same would be the case in most composite steel-framed buildings. The
exception is when the column is a `key' element and essential to the stability of
Although none of the protected columns in the Cardington tests failed, analysis
of the strain gauge measurements showed that there were large moments in
some of the columns after the fires. These moments were caused by two
additive effects, expansion of the heated floor relative to other floors and
induced P-delta effects.
As a fire-affected floor structure heats up, it expands. Normally the structure is
anchored horizontally at the core areas, so the movement tends to be greatest
away from cores and at the edges of the floor. Horizontal displacements of the
floors above and below the fire-affected floor are much smaller, so columns
between these floors are displaced horizontally in the middle of a two-storey
length thus inducing bending moments. In addition to the moments induced by
floor movement, moments will be induced by the P-delta effect as the vertical
load carried by the column is displaced. Analysis carried out by BRE  has
shown that, in some circumstances, the failure temperature of an affected
column might be reduced and additional protection would be required.
Columns in multi-storey buildings are generally not heavily loaded. The load
ratio, as defined in BS5950-8, rarely exceeds 0.5. The effect identified by BRE
increases the applied loading and load ratio and thus reduces the limiting
temperature. To allow for this effect, it is recommended that the fire protection
should be based on a limiting steel temperature of 500°C rather than 550°C or
more. A reduction to 500°C is equivalent to increasing the steel stiffness and
strength by at least 25%.
Most column fire protection is based on limiting the steel temperature to 550°C.
However, in many cases, the minimum practical thickness of protection that
may be applied to columns is sufficient to achieve 60 minutes on columns with
a section factor not greater than 200 m-1. As a result of this, for 60 minutes,
the thickness will normally be sufficient to keep the column temperature well
below 500°C. For 30 minutes fire resistance, temperatures will often be well
In one of the Cardington tests, the vulnerability of partially protected columns
was demonstrated when the unprotected top part of two columns squashed. No
collapse occurred but the floors above the fire compartment were all affected.
It was concluded that, until more was understood about how much of a column
could be left exposed, in the subsequent tests and in any design
recommendations, columns should be protected for their full height. Protecting
columns over their full height is important if damage is to be confined to the
Design points - columns
For 30 minutes fire resistance, columns should be protected up to th e underside of
the deepest supported beam.
For 60 minutes fire resistance, columns should be protected up to the underside of
the floor slab.
The fire protection to a column should be based on a reduced limiting steel
temperature. The amount of protection should be based on a temperature of
either 500°C or 80°C less than that derived from BS5950-8, whichever is the
Although connections in braced multi-storey frames are normally designed to
transmit shear forces only, they do in fact have a significant moment capacity
that is not utilised in ambient temperature design. The rotational stiffness of
these `simple' connections can be significant. In fire, the connections have the
potential to develop moments that could enhance the load-carrying capability of
a nominally simply supported beam. However, in the case of the Cardington
tests, it was found that local buckling of the unprotected beams occurred
adjacent to the connections in many instances, which negated the ability to
develop negative moments. Further investigation of the effect of local buckling
on the moment capacity of connections will be required before redistribution of
moments in fire-affected steel beams could be relied upon.
During the heating phase of the tests, connections performed well, however
some connections suffered partial failure during the cooling phase. This was
because the beams are subject to large compressive stresses during the heating
phase; these arise from restrained thermal expansion and thermally induced
curvature. The restraints had the effect of causing plastic compressive
deformation and, in some cases, local compressive instability. Both of these
result in shortening of the beams. As the beam cools, the plastic strains cannot
be recovered and the beams are therefore shorter than before the fire. The
overall effect is similar to the `lack-of-fit' problems with which engineers are
familiar. The connections were thus subject to tensile forces, which in some
instances led to some form of failure.
Three types of simple connection, end-plate, fin-plate and cleated are considered
in detail below.
5.4.1 End plate connections
Observations of the performance of end-plate connections in the Cardington
tests have shown that these connections perform well in fire. There was no
complete failure in any connection involving end plates. However, during the
cooling phase following the fire, a number of effects were observed. These
included rupture of one side of the end plate close to the heat-affected zone of
the end plate to beam web weld and, in one case, failure of the web as the end
plate was torn off the end of the beam. Both types of failure were due to high
tensile forces generated by contraction of the beams during cooling.
In the case where the plate was torn off the end of the beam, no collapse
occurred because the floor slab transferred the shear to other load paths. This
highlights the important role of the composite floor slab with proper lapping of
In the case of the plate rupturing, only one side of the plate became detached
leaving half the original number of bolts in place to resist shear (see Figure
B.5.6). The floor slab may also have distributed some shear to other load paths
in this case.
Figure B.5.6 Schematic of end plate rupture
5.4.2 Fin-plate connections
Fin-plate connections were used in the Cardington tests to connect the secondary
beams to the primary beams. These connections appeared to perform well in
fire but some problems were encountered on cooling. It was found that the
tensile forces generated in the beam due to plastic strain at high temperatures
sheared the bolts in the fin-plate connections at one end of the beam. No
collapse occurred, and the shear was carried by a combination of the secondary
beam resting on the bottom flange of the primary beam and the shear resistance
of the composite slab. Local buckling in the bottom flange of the beam was
also experienced in this type of connection, as shown in Figure B.5.6.
Tensile tests at ambient temperature show that fin-plate connections have a large
amount of ductility. A major contribution is the elongation of the bolt holes.
At Cardington, the bolts fractured, with no apparent elongation of the bolt holes
(see Figure B.3.7), indicating that the bolt material had lost strength during the
fire. Although bolt temperatures were not measured, an unprotected fin-plate in
Test 6 reached 862ºC.
Fin-plate connections may be detailed in two ways. At Cardington, they were
used for beam-to-beam connections and in all cases the top flange of the
incoming beam was notched back. Fin-plate connections are also commonly
used for beam-to-column connections, especially when the column is a structural
hollow section. In this case, it will often be unnecessary to notch the top flange
of the beam and, should the bolts fail, the vertical displacement of the beam will
be limited because the underside of the top flange will bear on the top edge of
the fin plate. In combination with a composite floor slab, this is a safe mode of
failure that will permit longitudinal movements of the beam while maintaining
Although connections formed from notched beams will not have the same fail-
safe mode of behaviour as those utilising unnotched beams, they are
nevertheless considered adequately safe because there is the possibility of an
incoming beam resting on the bottom flange of a main beam, and the adoption
of the slab design model (see 5.1.3) will often result in larger reinforcing mesh
than was used at Cardington.
5.4.3 Cleated connections
Although there were no cleated connections used in the Cardington frame, SCI
has conducted a number of tests on composite and non-composite cleated
connections in fire . These connections consisted of two steel angles bolted to
either side of the beam web using two bolts in each angle leg, then attached to
the flange of the column also using two bolts. The connections were found to
be rotationally ductile under fire conditions and large rotations occurred. This
ductility was due to plastic hinges that formed in the leg of the angle adjacent to
the column face. No failure of bolts occurred during the fire test. The
composite cleated connection had a better performance in fire than the non-
It is likely that the ability of the angle cleats to deform will provide the
necessary ductility on cooling.
5.4.4 Bolts and welds
Observations of the connection behaviour at Cardington indicate that bolts or
welds do not fail prematurely in fire situations and it is not generally considered
necessary to design welds or bolts specifically for fire scenarios. However, the
material properties of the bolts and welds are known to vary in a similar way as
those of structural steel . Current guidance on the strength reduction for bolts
and welds given in BS 5950-8 allows 80% of the strength reduction factor used
for structural steel (at 0.5% strain). No guidance is given in the Eurocodes. In
the Cardington tests no failures of fasteners occurred during the heating phase
of the fire. However, during the cooling phase, a number of bolts acting in
single shear in fin-plate connections failed due to the tensile force generated by
thermal contraction as described above.
Design points connections
At the ends of unprotected beams, connections with the ability to deform as the
beam shortens on cooling are preferable to less deformable types.
5.5 Overall building stability and bracing
The complete collapse of a multi-storey building as a result of a fire is never
acceptable, so consideration must be given to overall building stability. The
design recommendations are concerned principally with the performance of
non-sway frames, which would usually be braced by a steel bracing system or
by some form of brick or reinforced concrete shear walls. The latter two forms
of shear wall can be expected to have a high degree of inherent fire resistance.
Bracing systems formed from steel have been shown to perform well in fires but
their design and location needs some consideration.
The Cardington building had three braced cores. All bracing members were
positioned in protected shafts and were thus shielded from any of the test fires.
It is normal to place bracing systems inside shafts to isolate them from potential
fires, in which case it is not necessary to protect the individual members.
The cost of fire protecting bracing members is often high because the members
are comparatively light and therefore have high values of section factor (A/V).
Bracing within a structure has two roles. It resists lateral wind forces but,
especially for tall buildings, it contributes to the overall stability of the
structure. In fire, it is important to recognise these two roles. In the case of
wind, some guidance is offered by the structural design codes [3,10]. BS 5950-8
recognises that it is highly unlikely that a fire will occur at the same time as the
building is subject to the maximum design wind load; it consequently
recommends that, for buildings over 8 m tall, only one-third of the design wind
load need be considered and, for buildings up 8 m tall, wind loading may be
ignored. None of the codes gives guidance on overall frame stability, but it is
self evident that the designer must consider it for multi-storey buildings.
The design recommendations in this guide are based on the assumption that
sway instability will not occur. Bracing will normally be fire protected but
there can be justification for reducing the amount of fire protection on bracing
by taking into consideration the following:
* Bracing may be shielded from fire by installing it in shafts or within walls.
The shielding should provide all the necessary fire protection.
* Masonry walls, although often designed as non-load-bearing, may provide
appreciable shear resistance in fire.
Bracing systems are often duplicated and loss of one system may be acceptable.
Design points bracing
Whenever possible, bracing should be positioned in protected shafts or built into
walls, to avoid having to protect individual members.
5.6.1 Structural considerations
One of the key strategies of fire safety is to contain the fire in its compartment
of origin. In most multi-storey buildings, each floor forms a single
compartment; within a floor, the walls are not generally compartment walls.
However, in some buildings, the design does assume that certain walls that
subdivide a floor are compartment walls.
From the observations at Cardington, it is clear that large displacements of the
floor above a wall can occur and this could affect the performance of some
walls. It is common practice to use lightweight walls that are constructed as
stud walls using fire-resisting boards fixed to cold formed channels. These
walls are not designed to be load-bearing and have little resistance to the
potential loads that could be imposed from a deflecting structure during a fire.
Blockwork walls are less commonly used but they are much more robust than
stud walls and could be expected to resist quite large loads, even if designed as
non-load-bearing. However, the presence of thermal gradients through masonry
walls may cause out-of-plane bowing towards the heat source where the wall is
simply supported top and bottom, and away from the heat source where the wall
cantilevers from the base. Out-of-plane thermal bowing may also occur. The
material properties of the masonry will also deteriorate at high temperatures,
which may cause reverse bowing as a result of eccentricity of the load.
Because of the precautions taken in the Cardington tests, no failures or near
failures occurred in the masonry compartment walls.
For Tests 4 and 5 on the Cardington frame, internal compartment walls were
constructed using steel-framed stud partitions with fire resistant board. In
Test 4, the walls were placed on the column grid under unprotected beams.
Due to the shielding effect of the wall, the vertical deflection of these beams
was very small and the integrity of the compartment wall was maintained. In
Test 5, the compartment wall was placed `off-grid' and under the soffit of the
floor slab. The deflection of the slab caused integrity failure of the
It is advisable to place compartment walls under beams whenever possible.
However, to comply with the insulation criterion for separating elements as
specified in BS 476, these beams would normally be protected, thus
incorporating an additional level of safety in maintaining the integrity of the
Walls positioned off column grid lines require careful detailing to accommodate
large vertical displacements. The deflection of a beam may be of the order of
span/30. In the design recommendations, it is presumed that this deflection
exists over the central 50% of the span, with a linear reduction from this value
to zero at the supports. This is a much more onerous requirement than current
The measurements in Test 4 indicate that the temperature of one side of an
exposed beam above a wall surrounding the fire compartment is significantly
lower than that of a fully exposed beam in the centre of the compartment. The
data indicates that a beam above a wall on the compartment perimeter shows a
similar temperature reduction to that observed in a lintel beam above a window
(Section 5.2.1). In addition, the fact that the beam is only exposed on one side
and perhaps subject to less convection in corner `dead zones' reduces the
temperature rise still further. Measured temperatures in Test 4 (see Figure
B.5.5) showed that this additional reduction was about 20%, between 100 and
200ºC in this case.
Beams above compartment walls (not partitions) require protection to meet the
insulation criteria, as the surface temperature on the unexposed side is likely to
exceed the allowable limits for insulation specified in BS 476.
Design Points Compartmentation
Whenever possible, compartment walls should be positioned on column gridlines.
Compartment walls that are crossed by an unprotected beam should be designed
to accommodate the possible vertical displacement of the structure during a fire.
5.7 Other influencing factors
In this Section, three factors that influence the behaviour of steel-framed
buildings in fire are discussed: robustness, sprinklers and suspended ceilings.
Steel-framed multi-storey buildings are designed to meet certain robustness
requirements. Although this design guidance does not make reference to it, the
excellent robustness of steel-framed multi-storey buildings underpins the design
The good performance of the structure in the Cardington tests may be directly
related to the inherent robustness of composite steel-framed buildings. In the
UK, buildings over five storeys are required to be able to resist disproportionate
collapse (for example, from gas explosions). Smaller composite buildings,
although not required to have the same degree of robustness, will also be quite
robust. Robustness exists through the continuity of the floor slab and
The expected performance (or robustness) of structures in fires can be seen as
more onerous than the robustness requirements to resist explosions. In a multi-
storey building fire, a structure is expected to resist the effects of fire without
collapse and to contain the fire within the compartment of origin. The area of
structure that might be engulfed in the fire may often be larger than that
required to be considered for an explosion. A fire is considered to `fill', the
compartment, which could be a complete floor. The area allowed to be affected
by local collapse and debris loading in an explosion is 210 m2, provided that
progressive further collapse does not occur. Each floor at Cardington was
945 m2; most buildings will have fire compartment areas larger than 210 m2.
The performance expected of buildings in fire is therefore onerous compared
with some other accidental design conditions; nevertheless, buildings designed
using the proposed recommendations will meet the higher safety standards
expected in fire.
`Active fire protection systems' is a general term used to describe systems
designed to control or extinguish fires or to alert people to the presence of fire.
The system that makes the greatest contribution to structural fire safety and
property protection is an automatic water sprinkler system.
The design guidance does not rely on the presence of sprinklers but the
installation of sprinklers will have a beneficial effect on performance. The
value of sprinklers is well known and many buildings will have sprinklers
Traditionally, passive and active fire protection measures have been considered
as independent components of an approach to fire safety. The use of active
protection systems has been promoted by insurers, who recognise the value of
such systems in controlling fire growth and therefore reducing the extent of fire
damage in buildings. Statistical information from insurance sources confirms
that sprinkler systems can play an important role in reducing financial loss.
5.7.3 Suspended ceilings
It is not well known that certain types of non-fire-rated ceilings can shield the
floor and beams from the extremes of the fire, and will therefore contribute to
the performance of the structure.
The natural fire tests undertaken by BHP Australia (see Section 4.3) showed
clearly that even non-fire-rated suspended ceilings can significantly reduce the
temperature to which beams are exposed (see Figure B.5.7). In both the
William Street and Collins Street tests, cast plaster tiles were used, and in the
latter case the tiles were described as having a fibreglass backing blanket. In
both tests, the tiles were supported on steel runners rather than aluminium
extrusions (which would melt at 660ºC). Most significantly, it demonstrated
that even though the ceiling was breached, its presence was sufficient to reduce
the atmosphere temperature in the region of the beams by around 400ºC. The
unprotected beams reached temperatures between 280ºC and 470ºC, measured
at the lower flange position; in similar fire conditions at Cardington, but without
the ceiling, temperatures over 1000ºC were recorded.
The reason for this difference becomes apparent from the time-temperature plots
(see Figure B.5.8,Figure B.5.9 and Figure B.5.10), which are reproduced from
the BHP report. They show that the ceiling began to lose its integrity at around
30 minutes, just before the fire temperature peaked at 33 minutes, and that
beam temperatures did not begin to rise until after the fire temperature had
passed its peak. The beams were thus only fully exposed to a declining fire,
not the growing fire that the standard BS 476 fire simulates.
Although it is difficult to quantify the effect of non-fire-rated ceilings with steel
runners, it is clear that they could be expected to reduce beam temperatures and
thus enhance stability significantly. The ceilings, in combination with other fire
safety provisions, could justify relaxation of structural fire resistance
Figure B.5.7 Suspended ceiling before and after the Collins Street test (BHP, Australia)
Figure B.5.8 Maximum atmosphere temperatures in Collins Street test (BHP, Australia)
Figure B.5.9 The effect of a suspended ceiling on atmosphere temperatures (BHP)
Figure B.5.10 The effect of a suspended ceiling on steel beam temperatures (BHP, Australia)
6 THE WAY FORWARD
In this publication, design recommendations are presented that will allow many
secondary beams in composite steel-framed buildings to be designed without
applied fire protection. The recommendations can be seen as Level 1
recommendations; they are simple to apply and are conservative.
The design recommendations described are based on the accepted concept of
fire resistance given in building regulations. As knowledge of real fire
behaviour develops and research leads to a better understanding and
quantification of the risks and consequences of our design decisions, it should
be possible to design buildings to resist realistic fires and not for notional
periods of fire resistance.
It is hoped that, in time, it will be possible to offer all consulting engineers and
architects more refined fire engineering design aids (Level 2) to enable them to
take full advantage of the real behaviour of composite steel-framed buildings in