FUNDAMENTAL PRINCIPLES OF STRUCTURAL BEHAVIOUR

UNDER THERMAL EFFECTS

A.S.Usmani and J.M. Rotter

School of Civil and Environmental Engineering, University of Edinburgh
Keywords: Composite structures, thermal expansion, thermal bowing, restraint to thermal actions,
non-linear geometrical responses.
Abstract
Behaviour of composite structures in re has long been understood to be dominated by
the effects of strength loss caused by thermal degradation and that large deections and runaway resulting from the action of imposed loading on a weakened structure. Thus strength
and loads are quite generally believed to be the key factors determining structural response
(fundamentally no different from ambient behaviour). The exceptional longevity of this view
derives in no small measure from observations in the standard re tests. These observations
have little relevance to realistic structural congurations present in large multi-storey composite frame structures. The investigations as part of the DETR, PIT project on composite structure
behaviour in re has clearly shown that this understanding is gravely in error. A new understanding has been produced from the PIT project, sponsored by DETR (UK) and executed by
a consortium led by Edinburgh University and including British Steel (now CORUS) and Imperial College. The key message from this new understanding is that, composite framed structures of the type tested at Cardington possess enormous reserves of strength through adopting
large displacement congurations, and that thermally induced forces and displacements, not
material degradation, govern the response in re. Degradation (mainly steel yielding and buckling) can even be helpful in developing the large displacement load carrying modes safely. This
paper attempts to lay down some of the most important and fundamental principles that govern
the behaviour of composite frame structures in re in a simple and comprehensible manner.
This is based upon the analysis of the response of single structural elements under a combination of thermal actions and end restraints representing the surrounding structure.

1 Introduction
The assessment of the adequacy of composite steel frame structures in re continues to be based
upon the performance of isolated elements in standard furnace tests. This is despite the widespread
acceptance amongst structural engineers that such an approach is over-conservative and even more
importantly, unscientic. Current codes such as BS 5950 Part 8 and EC3 (draft) allow designers
to take advantage of the most recent developments in the eld by treating re related loading as
another limit state. The advances in understanding of structural behaviour in re achieved in the
last few years have been considerable. In theory, these advances make it possible for designers
to treat the design for re in an integrated manner with the design of a structure for all other
types of loading by using the numerical modelling tools that have been instrumental in developing
this understanding. However the use of such tools, which are indispensable for research, is not
practical in the design ofce. Exploitation of the new knowledge can only become feasible if
this understanding is further developed into simpler analytical expressions, enabling consulting
engineers and designers to undertake performance-based design of steel frame structures without
having to resort to large scale computation.
This paper builds upon earlier work presented at the INTERFLAM [1] and SiF [2] conferences.
The most fundamental relationship that governs the behaviour of structures when subjected to

thermal

mechanical

with

mechanical
and

total

total

thermal effects is:

(1)

The total strains govern the deformed shape of the structure , through kinematic or compatibility
considerations. By contrast, the stress state in the structure (elastic or plastic) depends only on
the mechanical strains.
Where the thermal strains are free to develop in an unrestricted manner and there are no external
loads, axial expansion or thermal bowing results from
thermal

and

total

total

(2)

By contrast, where the thermal strains are fully restrained without external loads, thermal stresses
and plastication result from

thermal

mechanical

with

mechanical

(3)

The single most important factor that determines a real structures response to heating is the manner
in which it responds to the unavoidable thermal strains induced in its members through heating.
These strains take the form of thermal expansion to an increased length (under an average centroidal temperature rise) and curvature (induced by a temperature gradient through the section
depth). If the structure has insufcient end translational restraint to thermal expansion, the considerable strains are taken up in expansive displacements, producing a displacement-dominated
response. Thermal gradients induce curvature leading to bowing of a member whose ends are free
to rotate, again producing large displacements (deections).
Members whose ends are restrained against translation produce opposing mechanical strains to
thermal expansion strains and therefore large compressive stresses (see Equation 1). Curvature
strains induced by the thermal gradient in members whose ends are rotationally restrained can
lead to large hogging (negative) bending moments throughout the length of the member without
deection. The effect of induced curvature in members whose ends are rotationally unrestrained,
but translationally restrained, is to produce tension.
Therefore for the same deection in a structural member a large variety of stress states can exist; large compressions where restrained thermal expansion is dominant; very low stresses where
the expansion and bowing effects balance each other; in cases where thermal bowing dominates,
tension occurs in laterally restrained and rotationally unrestrained members, while large hogging
moments occur in rotationally restrained members. This variety of responses can indeed exist in
real structures if one imagines the many different types of re a structure may be subjected to. A
fast burning re that reaches ashover and high temperatures quickly and then dies off can produce
high thermal gradients (hot steel and relatively cold concrete) but lower mean temperatures. By
contrast, a slow re that reaches only modest temperatures but burns for a long time could produce
considerably higher mean temperature and lower thermal gradients.
Most situations in real structures under re have a complex mix of mechanical strains due to applied loading and mechanical strains due to restrained thermal expansion. These lead to combined
mechanical strains which often far exceed the yield values, resulting in extensive plastication.
The deections of the structure, by contrast, depend only on the total strains, so these may be quite
small where high restraint exists, but they are associated with extensive plastic straining. Alternatively, where less restraint exists, larger deections may develop, but with a lesser demand for
plastic straining and so less destruction of the stiffness properties of the materials. These relationships, which indicate that larger deections may reduce material damage and correspond to

higher stiffnesses, or that restraint may lead to smaller deections with lower stiffnesses, can produce structural situations which appear to be quite counter-intuitive if viewed from a conventional
(ambient) structural engineering perspective.
The ideas presented above will be more formally explored in the following sections in the context
of simple structural congurations and analytical expressions will be developed for many cases of
fundamental importance.

2 Standard re test and runaway failures

Runaway
0
laterally fixed
laterally free
-500

-1000

Unrestrained

Laterally restrained

Deflection (mm)

-1500

-2000

-2500

-3000

-3500

-4000

-4500
0

100

200

300

400
500
600
Temperature (C)

700

800

900

1000

Figure 1: Runaway in unrestrained and restrained beams

Figure 1 shows a simple comparison of two geometrically non-linear analyses. The rst case is
heated, simply supported (laterally unrestrained) steel beam with a uniformly distributed load and
the second is a laterally restrained (but rotationally unrestrained) beam with the same uniformly
distributed load. Initial deections are lower in the rst beam because the supports are able to
translate outwards upon expansion. However, runaway occurs at around 450 C (even though
considerable steel strength remains) mainly because of pulling in of the supports when the exural
stiffness of the beam reduces to a point where it cannot sustain the imposed load and there is
nothing to restrain the growing deections. In the second beam larger initial deections occur
because the beam buckles due to the restraining forces very early on (70 C ) and further increases
in length due to thermal expansion can only be accommodated in deection. But runaway does not
occur until much later (800 C) when the steel properties are completely lost. This illustrates that
the presence of restraints to end translation delays runaway to much higher temperatures because
of development of catenary action to replace the highly depleted exural stiffness.

The second beam is a much more appropriate model for beams in large redundant structures. In
real structures not only is this restraint available but the steel beam is in composite action with the

concrete slab which produces a much stronger and more robust structure. This strength and robustness is enhanced by the redistribution mechanisms present in redundant structures (for instance the
load may be carried by the transverse slab supported in tensile membrane action which retains its
strength for much longer than the steel beams). Large deections seen in real structures are often
misinterpreted as impending runaway failure. Figure 1 clearly shows that for temperatures below
300 C, the deections for the restrained beam are much larger than that for the simply supported
beam, however they have nothing to do with runaway. These deections are caused entirely by the
increased length of the beam through thermal expansion and are not a sign of loss of strength or
stiffness in the beam until much later. In fact approximately 90% of the deection at 500 C and
75% at 600 C is explained by thermal expansion alone. Most of the rest is explained by increased
strains due to reduced modulus of elasticity. However the behaviour remains stable until about
700 C when the rst signs of runaway begin to appear.

-500

Deflection (mm)

uniformly distributed load

-1500

Laterally restrained beam of length

L and maximum flexural capacity

-1000

Mp

w = 8Mp
L2

-2000

Load

-2500

3w
700 C

-3000

2w

3w
2w
w
0.5w
0.2w
0.02w

-3500

800 C
w
0
1

Defl.
(m)

-4000
0

200

400

600
Temperature (C)

800

1000

1200

Figure 2: Effect of increasing loads on runaway in a restrained beam

Figure 2 shows another interesting analysis, where the same laterally restrained beam (as in Figure
1) has been loaded with different magnitudes of uniformly distributed load (w) in proportion to
the udl that will cause a plastic hinge to occur at midspan. The Figure clearly shows that tensile
membrane (or catenary) action allows loads much larger that the ultimate to be supported to very
high temperatures. It also shows that deections (plotted at midspan) remain insensitive to the load
until just before runaway. This is true even for the case of high loads (2w and 3w) where the initial
deections before heating are already very large indicating that tensile membrane action is being
exploited. The difference between the midspan deection for the lowest load (0 02w) against the
highest load (3w) stays practically constant upto approximately 600 C, indicating that upto this
temperature the imposed loads have little effect on the response of the structure, which is almost
completely governed by the thermal actions. Just before runaway failure, the loads begin to exert a
much greater inuence on the weakened structure. The temperature at which runaway eventually
occurs, is sensitive to the magnitude of load. The inset graph shows the change in load against
deection for two temperatures, where reducing slope indicates runaway.

3 Thermal expansion
Heating induces thermal expansion strains (say T ) in most structural materials. These are given
by,
(4)
T T

If a uniform temperature rise, T , is applied to a simply supported beam without axial restraint, the
result will simply be an expansion or increase in length of lT as shown in Figure 3. Therefore
the total strain (say t ) is equal to the thermal strain and there is no mechanical strain (say m )
which means that no stresses develop in the beam.
Uniform temperature rise T

   

  

  
l

Figure 3: Uniform heating of a simply supported beam

3.1 Thermal expansion against rigid lateral restraints

Clearly beams in a real structures do not have the freedom to elongate in the manner described
above. Therefore a more realistic case is to consider an axially restrained beam subjected to a
uniform temperature rise, T (as shown in Figure 4). It is clear to see that in this case the total
strain t is zero (no displacements). This is because the thermal expansion is cancelled out by
equal and opposite contraction caused by the restraining force P (i.e. t T m 0 therefore
T
m ). There exists now a uniform axial stress in the beam equal to Em . The magnitude
of the restraining force P is,




EAT




Uniform temperature rise T

  
  




EAT

  
  

EAm

Figure 4: Axially restrained beam subjected to uniform heating

If the temperature is allowed to rise indenitely, there are two basic responses, depending upon the
slenderness of the beam:
1 If the beam is sufciently stocky the axial stress will sooner or later reach the yield stress y
of the material and if the material has an elastic-plastic stress-strain relationship, the beam
will continue to yield without any further increase in stress, but it will also store an increasing
magnitude of plastic strains. The yield temperature increment Ty is,

Ty

y
E

2 If the beam is slender then it will buckle before the material reaches its yield stress. The
Euler buckling load Pcr for a beam/column as in Figure 4 is,
2 EI
l2
equating this to the restraining force P, we have,

Pcr

2 EI
l2
which leads to a critical buckling temperature of,

or

2 r
l

(5)

Tcr

EAT

2
(6)
2
l
where r is the radius of gyration and is the slenderness ratio ( r ). This expression is valid
for other end-restraint conditions if l is interpreted as the effective length.

Tcr

In this case, if the temperature is allowed to rise further, the total restraining force will stay
constant (assuming elastic material and no thermal degradation of properties) and the thermal
expansion strains will continue to be accommodated by the outward deection of the beam
as shown in Figure 5.

" " "

Uniform temperature rise T

# # #
# # #

Pcr

Pcr

Figure 5: Buckling of an axially restrained beam subjected to uniform heating

The above cases represent the two fundamental responses in beams subjected to restrained thermal expansion. Either of the two (yielding or buckling) may occur on its own (based upon the
slenderness of the beam) or a more complex response consisting of a combination of yielding and
buckling may occur.
The pattern of development of deections, axial compression forces and moments with increase
in temperature in slender restrained elastic beams is as shown in Figures 6and 7. The deection
and axial force gures clearly show a pre-buckling and post-buckling type response. The sharp
bifurcation pattern is absent as a uniformly distributed load is imposed on the beam, imparting an
initial displacement to it. The midspan moment continues to rise even after buckling as it consists
mainly of the P moment generated by the axial restraint force times the midspan deection
(which continues to rise beyond buckling).

Ideal elastic properties were assumed when discussing the case of buckling above. If the properties
are ideal elasto-plastic the deections and axial compression variations will have a pattern as shown
in Figure 8. If the properties remain elastic albeit with a uniform degradation with temperature,
the pattern of deection and axial compression in the beam changes to the one shown in Figure
9. Clearly the response of real composite beams subject to restrained thermal expansion will
consist of a combination of the responses shown here. That this is indeed the case, can be seen
in report AM1, where the results of modelling the British Steel restrained beam Test are shown
(which comes closest to the ideal case of rigid lateral restraint). There are other factors in that
Test that govern the response of the heated composite beam, particularly the effect of deection
compatibility in the two directions, however the similarity of the development of axial forces in
the steel joist and the composite beam to the patterns shown here is clear to see.

'56 & 4 ) )&3 & 21 )0 & ( ' & % $

-2.00E+01

Deflection at Mid-span
0.00E+00

T S R A Q Q A H H A C P A B C AI H DAEFG B BAC A @9 8 7

Deflection at Mid-span
0.00E+00

-5.00E+01

-4.00E+01

-1.00E+02
Deflection (mm)

Deflection (mm)

-6.00E+01

Deflection at mid-span

-8.00E+01

-1.00E+02

-1.20E+02

D/L = 1.78%
D/L = 2.17%
D/L = 2.67%

-1.50E+02

-2.00E+02

-1.40E+02

-2.50E+02
-1.60E+02

-1.80E+02

-3.00E+02
0

50

100

150

200

250

300

350

50

100

150

Temperature (C)

200

250

300

350

Temperature (C)

(a)

(b)

Figure 6: Deection of axially restrained elastic beams subjected to heating: (a) Single beam, (b)
Three beams of varying slenderness

'56 & 4 ) )&3 & 21 )0 & ( ' & % $

0.00E+00

Moment at Mid-span
1.20E+09

Moment due to Distributed Load

1.00E+09
Moment due to Non-Linear geometrical
effect (P-Delta)

Normal Force at support

-1.00E+06

Xfg W e ` `Wd W cb `a W Y X W V U

Normal Force at Mid-span

1.00E+06

Moment Carried by the beam

Normal Force at mid-span

8.00E+08

Moment (N.mm)

Force (N)

-2.00E+06

-3.00E+06

6.00E+08

4.00E+08

-4.00E+06

2.00E+08
-5.00E+06

0.00E+00

-6.00E+06

-7.00E+06

-2.00E+08
0

50

100

150

200

250

300

350

50

100

Temperature (C)

150

200

250

300

350

Temperature (C)

(a)

(b)

Figure 7: Forces in an axially restrained elastic beam subjected to heating: (a) Axial Forces, (b)
Moments

u t s m d r q i pd g h o n m g h l jk i h g f ed

u q p u r x q y r x q y pq y p p r x s spw p vu st p r q p i h

Normal Force at Mid-span

Deflection at Mid-span
0.00E+00

-2.00E+01

5.00E+05

0.00E+00

-4.00E+01

-5.00E+05

-6.00E+01

-1.00E+06

-8.00E+01

-1.50E+06
Force (N)

Deflection (mm)

Normal Force at support

-1.00E+02

Normal Force at mid-span

-2.00E+06

-2.50E+06

-1.20E+02

-3.00E+06

-1.40E+02
Deflection at mid-span
-1.60E+02

-3.50E+06

-1.80E+02

-4.00E+06

-4.50E+06

-2.00E+02
0

50

100

150

200

Temperature (C)

(a)

250

300

350

50

100

150

200

250

300

Temperature (C)

(b)

Figure 8: Deections (a), & Axial forces (b), in an axially restrained elastic-plastic beam

350

 y x { { | z ~ { x | z { x x  {~ } {| z y x w v

Deflection at Mid-span

-2.0E+01

Normal Force at Mid-span

0.0E+00

1.0E+06

0.0E+00

Deflection at mid-span
-4.0E+01

Normal Force at support

Normal Force at mid-span

-1.0E+06

-8.0E+01

Force (N)

Deflection (mm)

-6.0E+01

-1.0E+02

-2.0E+06

-3.0E+06

-1.2E+02

-4.0E+06
-1.4E+02

-5.0E+06
-1.6E+02

-6.0E+06

-1.8E+02
0

50

100

150

200

250

300

350

50

100

150

200

250

300

350

Temperature (C)

Temperature (C)

(a)

(b)

Figure 9: Deections (a), & Axial forces (b), in a restrained beam with reducing elastic stiffness

3.2 Thermal expansion against nite lateral restraints

In the previous discussion we have assumed the axial restraints to be perfectly rigid. This is an
upper limit and practically impossible to achieve in real structures which offer only nite restraints.
Figure 10 shows such a beam restrained axially by a translational spring of stiffness kt . The
compressive axial stress developed by thermal expansion is,
ET

(8)

prebuckling state: expansion develops axial compression

postbuckling state: expansion produces deflections

EA
kt L

kt

Pcr

2
1
2

Tcr

and critical buckling temperature is now given by,

(7)

EA
kt L

Pcr

end restrained
with stiffness
kt against axial
translation

Figure 10: Heating of beam with nite axial restraint

From Equation 8 it can be seen that buckling and post-buckling phenomena should be observable
at moderate re temperatures in structures with translational restraint stiffnesses (kt ) which are
quite comparable with the axial stiffness of the member ( EA ). Figure 11 shows a plot derived
L
from Equation 8, where critical buckling temperatures are plotted against slenderness ratio for
different restraint stiffnesses. The results clearly show that the amount of restraint required is not
large for slender sections to reach buckling temperature. Bearing in mind that the axial stiffness

of the member ( EA ) itself is reduced by heating through the reduction in E, so these post-buckling
L
phenomena are very likely to be observed in beams in typical res.

kt
kt
kt

kt

5 EA
l
2 EA
l
EA
l

Figure 11: Buckling temperatures for thermal expansion against nite lateral restraint

4 Thermal bowing
In the previous section we discussed the effects of uniform temperature rise on axially restrained
beams. In real res the temperature distributions are anything but uniform. In a small to moderate
size compartment of a regular shape one may assume that the compartment temperature will be
roughly uniform at a given time. The temperature of the structural members in the compartment
depends upon the material they are made of and other details of geometry, construction and design
(such as insulation). Concrete beams and slabs on the ceiling of the compartment can be subjected
to very high temperature gradients due to the very slow rates of heat transfer to concrete. Therefore
the surfaces exposed to re will be at a much higher temperature than the surfaces on the outside of
the compartment. This causes the inner surfaces to expand much more than the outer surfaces inducing bending in the member. This effect is called thermal bowing and is one of the main reasons
of the deformations of concrete slabs and masonry walls in res. Another very important source
of thermal bowing in composite beams/slabs is the large difference between the temperatures of
the steel joist and the slab. This effect is much more important in the early stages of the re when
steel retains most of its strength.
Relationships can be derived for thermal bowing analogous to the one derived earlier for thermal
expansion. Figure 12 shows a beam subjected to a uniform temperature gradient through its depth
(d) along its whole length (l). Assuming the beam is simply supported (as shown in Figure 12) we
can derive the following relationships:

1 The thermal gradient (T y ) over the depth is,

T2

Ty

T1 1
d

2 A uniform curvature () is induced along the length as a result of the thermal gradient,
T y

3 Due to the curvature of the beam the horizontal distance between the ends of the beam will
reduce. If this reduction is interpreted as a contraction strain (not literally) (analogous
to the thermal expansion strain T earlier), the value of this strain can be calculated from
analysing Figure 12 as:
sin l
2
(9)
1
l
2

T2 > T1
d

NA

T1

T2

Figure 12: Simply supported beam subjected to a uniform thermal gradient

Now consider the laterally restrained beam of Figure 5. If a uniform thermal gradient T y is applied
to this beam (as shown in Figure 13), the result (in the absence of any average rise in temperature,
i.e. mean temperature remaining constant) is a thermally induced tension in the beam and corresponding reactions at the support (opposite the the pure thermal expansion case discussed earlier).
This is clearly caused by the restraint to end translation against the contraction strain ( ) induced
by the thermal gradient.
Figure 14 shows a xed ended beam (by adding end rotational restraints to the Beam of Figure 13)
subjected to a uniform temperature gradient through its depth. Recalling that a uniform curvature
T y exists in a simply supported beam subjected to gradient T y . If that beam is rotationally
restrained by support moments M (uniform along length) an equal and opposite curvature induced

Uniform temperature gradient T,y

Figure 13: Laterally restrained beam subjected to a uniform thermal gradient

by the support moments cancels out the thermal curvature and therefore the xed ended beam
remains straight with a constant moment M EI along its length.

Uniform temperature gradient T,y

Figure 14: Fixed end beam subjected to a uniform thermal gradient

From the above discussion it is clear that the effect of boundary restraints is crucial in determining the response of structural members to thermal actions. The key conclusion to be drawn from
the discussion so far is that, thermal strains will be manifested as displacements if they are unrestrained or as stresses if they are restrained through counteracting mechanical strains generated
by restraining forces.
As discussed earlier for lateral restraint, perfect rotational restraint is also not very easily achieved
in real structures (other than for symmetric loading on members over continuous supports, without
any hinges from strength degradation). Figure 15 shows a beam restrained rotationally at the ends
by rotational springs of stiffness kr . In this case the restraining moment in the springs as a result
of a uniform thermal gradient T y can be found to be,

EIT y
2EI
kr l

(10)

Mk

This equation implies that if the rotational restraint stiffness is equal to the rotational stiffness of
the beam itself ( EI ) then the moment it will attract will be about a third of a xed support moment.
l

kr
Uniform temperature gradient T,y

kr

Figure 15: Beam with nite rotational restraint with a uniform thermal gradient

5 Deections
In the previous sections we have looked at the overall behaviour of beams subjected to expansion
and bowing for various restraint conditions. One interesting aspect of structural response to re
is the large deections that are found in structural members such as beams and slabs. Large deections are normally associated with loss of strength in structures under ambient conditions. In

case of re such a simple interpretation can be highly misleading. Both the thermal mechanisms
discussed earlier (thermal expansion and thermal bowing) result in large deections, however the
state of stress associated with a member subjected to varying degrees of these two mechanisms
is not unique for a given deection and a large range of stress states exist (large compression or
tension or very low stresses) depending upon the temperature distribution in the member and its
material properties and restraint conditions.
The chief reason for large deections is that the structural member tries to accommodate the additional length generated by thermal expansion, given that it is not possible for it to expand longitudinally due to end restraints. Consider a slender beam (very low buckling temperature) subjected
to uniform heating against rigid lateral restraints (as in Figure 5). Buckling will occur very early on
(at very low elastic strains) after which any further expansion will make the beam deect outwards.
The resulting midspan deection can be approximated quite accurately by,
2l

2
T
2

(11)
T , where T is the

which is an approximation of the deection of a sin curve of length l 1

thermal expansion strain (T ).

If the same beam is subjected to a uniform thermal gradient producing no net expansion, only
bowing as in Figure 13, the response is then determined by the exure-tension interaction. The
tensile P moments restrain the curvature imposed by the thermal gradients and limit deections.
The deections result from the tensile strains produced in the beam, i.e.

2l

(12)

2
t
2

and the deections can then be determined by

P
EA

(13)

The tensile force Pt can be determined from substituting Equation 12 in Equation 13 and solving a
quadratic equation for Pt ,
1
2

Pt

1 EA

(14)

To determine the deection y x in the beam of Figure 13 for a given curvature (arising from a
thermal gradient), a differential equation solution can be written as follows:

For a simply supported beam subjected to a uniform curvature one can write,

d 2y
dx2

If the beam is laterally restrained as in Figure 13, a tensile force P will be generated causing a
moment Py over the length of the beam, therefore,
d 2y
dx2

k2 y

d2y
dx2

or

Py
EI

(15)

where,

coshkl 1
sinh kx
sinh kl

k2

yx

P
EI

cosh kx

The solution of this equation is,

(16)

It may be seen that Equations 14 and 16 form a set of nonlinear equations. These equations can be
solved using an appropriate iterative technique (bisection, Newton-Raphson) to obtain the tensile
forces and deections for thermal gradient dominated problems.

6 Combinations of thermal expansion and thermal bowing

In the previous sections the response of beams to either thermal expansion or thermal bowing have
been considered in isolation. To study the combined response let us rst consider the case of a xed
ended beam as shown in Figure 16 which is both rotationally and translationally restrained at both
ends. If this beam is subjected to a mean temperature rise and a through depth thermal gradient,
it will experience a uniform compressive stress because of restrained expansion and a uniform
moment because of the thermal gradient. The stresses on any typical cross-section because of the
combined effect of the two thermal actions are also shown in Figure 16. It is clear that the bottom of

=ET

P
M

=ET,y y
Tens.

Comp.

Uniform temperature gradient T,y

Uniform temperature increase T

=
Comp.

Figure 16: Combined thermal expansion and bowing in a xed ended beam
the beam will experience very high compressive stresses, while the top may be anywhere between
signicant compression to signicant tension.
The above scenario is a common one in composite frame structures such as Cardington. The
composite action of a steel joist, framing into an interior column, with a continuous slab over it,
produces conditions very close to a fully xed support (as in Figure 16). The high compressions
resulting from the combined effect of thermal actions as described above almost invariably produce
local buckling in the lower ange of the steel joist very early on in a re. This why local buckling
of the lower anges is such a common occurrence in res (seen in all Cardington tests and other
res).
Once local buckling has occurred the pattern of stresses at the ends of the composite beam changes.
The hogging moment is relieved by the hinge produced by local buckling and the end restraint
conditions change to the one shown in Figure 13. As this happens quite early in real res, the end
conditions described by Figure 13 are the ones that govern the behaviour of a composite beam for
most of the duration of the re.

6.1 Combined thermal expansion and bowing in laterally restrained beam

The fundamental pattern of behaviour of a beam whose ends are laterally restrained (but rotationally unrestrained, see Figure 13) subjected to thermal expansion and thermal bowing separately
was established in previous sections. Restrained expansion resulted in compression and bowing
resulted in tension. This helped to illustrate that two opposite stress regimes can occur depending upon the thermal regime applied, however the apparent response of the beam is the same
(i.e. downward deection).
To study the effects of the applying combinations of thermal expansion and thermal bowing we
dene and effective strain as follows,
T

eff

(17)

The variation of eff (for various thermal regimes) can produce a large variety of responses. Positive values of eff imply compression (or the effect of mean temperature rise is dominant) and
negative values imply tension (or the effect of thermal gradients is dominant). Figure 17 shows the
variation of eff for different values of thermal gradient when the temperature is increased from 0
to 400 C (eff is plotted for a linear increase in gradient against temperature).

0.003
0.002

eff

0.001
0
-0.001

100

150

200 250
T ( C)

50

-0.003

C/mm
C/mm
C/mm
C/mm
C/mm

-0.002

T y =1
T y =2
T y =3
T y =4
T y =5

300

350

400

Figure 17: Effective expansion strains

Case 1: Zero stress in the beam (eff

0)

Figure 18 shows an interesting theoretical case. If the implied combination of and T are applied:

There are no stresses in the beam. All thermal strains are converted into displacement as
seen in the gure.

The deection of the beam is entirely due to thermal bowing to accommodate the excess
length generated by expansion.

the deection response of this beam can be analytically expressed as the increase in length
of the elastic curve of the beam versus its deection. Figure 19 shows a number of length
increase vs midspan deection plots based on assumed curve shapes. The gure shows that
the shape of the curve chosen does not matter much, therefore the formula given earlier based

Step 1 : Impose a temperature rise T

Step 2 : Impose a curvature

T,y

to return support to the original position

Figure 18: Case 1: Zero Stress

upon the sin curve (Equation 11) can be used to get a good approximation of the midspan
deection ym , i.e.
2l

ym

2
T
2

0
half sin-wave
circular arc
parabolic
triangular

-0.02

ym
l

-0.04
-0.06
-0.08
-0.1
-0.12

0.005

0.01

0.015

0.02

0.025

0.03

0.035

Figure 19: Strain v non-dimensional midspan deection (per unit length)

Case 2: Thermal expansion dominant (eff

0)

If T
, thermal expansion dominates and a two stage response is produced consisting of Prebuckling and Post-buckling phases. The thermal expansion produced is partly used up in generating
mechanical strains and partly in generating deections. This is governed by the magnitude of eff
which is the component that generates stresses to progress the beam towards buckling. The
component annihilates part of the expansion and produces deections by imposing curvature with
the available excess length. The pre-buckling deections (ym ) will for a small part result from the
elastic bending of the beam and a larger part will generally come from the deection resulting from

ym

y0

T
Tcr

ym

the imposed curvature (ym ).

(18)

Here y0 can be interpreted as the initial elastic deection before the re because of the imposed
loads on the beam and ym is the extra deection due to thermal bowing given by,

ym

2l

(19)

The presence of the gradient clearly delays the buckling event and the critical buckling temperature
(Tcr ) is increased to,
1 2
Tcr

(20)
2

Figure 20 shows the typical variation in buckling temperature with the change in gradient (for a
l
beam of slenderness ratio r equal to 70).
1000
Tcr

900

Tcr ( C)

800
700
600
500

400
300
0

500 1000 1500 2000 2500 3000 3500 4000 4500 5000
T y ( C/m)

200

Figure 20: Critical Buckling temperatures vs Thermal gradient

The postbuckling deections will carry on increasing because of all further expansion strains T as
in Equation 11, i.e.
T
2

2l

ym

The thermal bowing deection added to the elastic deections (due to P moments and loading)
will again act as imperfections to straightness of the beam and produce a smooth variation
of beam midspan deection with temperature until the large displacement post-buckling mode
begins (identied by the change of curvature of the temperature deection curve). This also has
the effect of reducing the development of compression forces in the beam (as the beam displaces
more for lower compressions because of the additional bowing displacements increasing the P
moments).
0)

T , the deection response will be the sum of two components,

When

Case 3: Thermal bowing dominant (eff

1 Deection caused by bowing of the excess length generated through expansion, as before
i.e.
2
2l
T
ym 1
T

2 The tensile strain t produced by the tension (Peff ) caused by the excess contraction strain
(eff ) as in the case of pure thermal gradients.

2l

ym

which will produce further deections as,

Peff
EA

2
t
2

The tension Peff and the deections can then be determined according to the iterative procedure suggested in the previous section on deections.
Finally Figure 21 shows the main types of deection responses that may be observed if a laterally
restrained beam is exposed to combinations of thermal actions discussed above.

Tensile forces
in the beam

Bifurcation
Tcr

Prebuckling

Compressive
forces in the beam
Postbuckling

Tcr ()

( =0)

T >

T =

(or y0 > 0)

(zero stress)

T <

Figure 21: Temperature deection responses for combinations of T and

7 Criterion for the various types of responses

From the discussion above a simple criterion for all the various types of responses observed can be
2
developed. If eff is close to 2 then there will be no buckling as not enough compression is gen
erated. A dimensionless number may be dened as follows to categorise the various responses:

To summarise,

2
2

(21)

1
1
typically generates pre and postbuckling type deection responses with thermal expansion
and compression dominant. The compression force patterns are as discussed earlier in the
restrained thermal expansion section.

2 1
generates responses where most of the thermal expansion is converted into deection but
there are negligible stresses in the beam (close to the zero stress case discussed earlier).

1
3
generates thermal bowing dominated response with deection patterns similar to the zero
stress case and with considerable tensile forces in the beam which grow with the increase in
the gradient.

8 Numerical analysis of the simple beam model

The analytical approach developed above to fully understand the structural response to the heating
regime has been checked numerically by modelling the same simple beam examples in ABAQUS.
The data for beam analysed was as follows:

length (l)=9000 mm

Coefcient of thermal expansion ()=8

Second moment of area I =8 55

Modulus of Elasticity (E)=210000 N/mm2

Area A=5160 mm2

10

107 mm2

l
Therefore the slenderness ratio r of the beam is approximately equal to 70. This calculation is
limited to investigated the simple beam model restrained laterally but free to rotate at its ends as
in Figure 13. The results conrm the theoretical solutions derived for the response of the beam
to thermal bowing and thermal expansion. Figure 22 shows the results of the numerical analysis in terms of the deections and axial forces produced when the beam is subjected to a mean
temperature rise (uniform over the length) of 400 C and an effective thermal gradient through
the depth of the beam. The temperature increase, T and thermal gradient, T y were applied to
the simple numerical beam model at a constant rate from zero to their maximum values. The
deection as a result of pure restrained thermal expansion shows the double curvature shape of
the pre-buckling/post-buckling response (see Figure 22). When a gradient is also applied to the
model and the response of the beam is governed by the interaction between thermal bowing and
restrained thermal expansion the deected shape becomes smoother, indeed for a specic combination of mean temperature rise and temperature gradient the response will be very close to linear. At
large gradients (T y 10 C/mm) when the response is dominated by thermal bowing the deected
shape is very non-linear.

The corresponding axial forces are also plotted in Figure 22. When a mean temperature rise of
400 C alone is applied to the model and the response of the beam is governed purely by restrained
thermal expansion and the axial force is in high compression. When the model is subjected to a
combination of mean temperature rise and temperature gradient the axial force becomes smaller
in compression and at high gradients moves into tension. The axial force at the beginning of the
analysis is always in compression because the mean temperature and the gradient are applied to

Deflection at mid-span due to thermal expansion and thermal bowing

50.00
0.00
0

50

100

150

200

250

300

350

400

-50.00

400C
400C_1C/mm
400C_3C/mm
400C_5C/mm
400C_10C/mm

Deflection (mm)

-100.00
-150.00
-200.00
-250.00
-300.00
-350.00
-400.00
-450.00

Temperature (C)

(a)
Axial Force in the model
3.00E+06
2.50E+06
2.00E+06

Axial Force (N)

1.50E+06
1.00E+06
5.00E+05
0.00E+00
0

50

100

150

200

250

-5.00E+05

300

350

400

400C
400C_1C/mm
400C_3C/mm
400C_5C/mm
400C_10C/mm

-1.00E+06
-1.50E+06
-2.00E+06
-2.50E+06
Temperature (C)

(b)
Figure 22: Numerical model results for combined thermal expansion and thermal bowing: (a)
Deections (b) Axial forces

the model linearly from zero to their maximum values and are governed by the development of the
effective strains, eff , as shown in Figure 17.
The actual values of deections and forces in the numerical exercise above can be estimated using
the formulas given here. For instance for the case of a temperature rise of 400 C the compression
2
force is simply the Euler buckling load ( EA ) equal to 2170 kN (approx.). The deection for this
2

case can be obtained from subtracting the elastic compression strain ( 2 ) from the thermal strain

(T ) to obtain the strain that produces the deections, therefore,

2

2l

ym

2
2

which produces a value of approximately 200mm (within 10%) of the numerical calculation above.
The difference is because the numerical calculation is fully geometrically non-linear while the
above formulas are based on 1st order denitions of strain.
If all the thermal strains were to produce deection (by appropriate combination of T and ) then
the internal forces would be very low and the deection would be approximately 324 mm (from
Equation 11) which lies between the cases of T y 3 C/mm to T y 5 C/mm. It may be noted
from Figure 22 that the axial force in the beam moves from compression to tension between these
values (suggesting that for the deections in the region of 324 mm) the forces in the beam will be
insignicant.

The tensile forces and deections for large gradients can be calculated from iteratively solving the
set of nonlinear Equations 14 and 16.
The above analysis clearly highlights the large range of deected shapes and axial forces possible
as a result of the interaction between thermal expansion and thermal bowing.

9 Other important factors

The discussion above has focussed upon the effects of thermal expansion and thermal bowing and
illustrated the large variety of responses possible in real composite frame structures. Analytical
expressions have been presented which allow a good quantitative estimate of forces and deections
to be made for simple structures.
Effects of thermal degradation and imposed loading were found to be relatively less important in
the modelling of Cardington Tests (see reports SM1-2). The effect of strength degradation was
shown to change the development of compressive forces in a restrained beam in Figures 8 and
9. The loading on a beam in a large displacement conguration (through thermal effects) will be
carried very effectively in a catenary (or tensile membrane) behaviour. It is clear from the above
discussion that for the most likely combinations of thermal actions (T and ) the mechanical
strains in a member are likely to be very low (compression or tension). If this is true and that
thermal degradation has been contained in the surface layers, then the tensile strains induced by
membrane mechanisms should be carried quite reliably. This is however an area which needs
further extensive investigation using all standard research techniques, experimental, computational
and theoretical.
Restraint conditions can certainly have a major effect on the distribution of the internal forces
and the displacements that occur as has been illustrated by the simple theoretical and computational analyses in this paper. The degree of restraint available also changes during a re , for
instance the rotational restraint available to the composite beam at the beginning is lost quite early
on (around 200 C) [3] due to the local buckling of the steel joist and tensile capacity of the slab

being reached. Rotational restraints result in increasing hogging moments until a plastic hinge
is achieved. Lateral translation restraints produce compression forces if thermal expansion was
dominant and tension forces if thermal bowing is dominant. The amount of restraint required is
not large to produce buckling as oor structures usually very slender. The source of this restraint
is obvious for interior compartments - the colder and stiffer surrounding structure. For exterior
compartments it is not so clear if sufcient restraints are still available. It is likely that sufcient
restraint to lateral expansion is available at exterior boundaries through the actions of tension rings
[1]. At large deections lateral restraints provide an anchor to the tensile membrane mechanisms.
Again, it is likely that sufcient lateral restraint is available at exterior boundaries through the
action of compression rings [4]. This however is a matter of much greater importance than the
restraint to thermal expansion as the survival of the oor system ultimately depends upon the reliability of the tensile membrane mechanism. This again is a key question for further investigation.
Another very important factor that has not been investigated here is the effect of the compartment
geometry. This can have a large effect on the development of thermally induced forces and deections in the heated structural members. The principle that allows one to make a quantitative
assessment of the effect of compartment geometry, is compatibility. For instance for a rectangular
re compartment, the thermal expansion in the shorter direction will be smaller than the expansion
the longer direction. This can lead to an increase in compression in the longer direction (because compatibility does not allow it to deect as much as its thermal expansion demands). In the
shorter the reverse happens, compatibility forces the deections in this direction to be somewhat
larger than thermal expansion would allow resulting in lower compressions or even tensile forces.
This has been identied clearly in the modelling of the British Steel restrained beam test (3m 8m)
where the midspan ribs are in tension. This allows redistribution of the thermally induced forces.

10 Conclusions
The fundamental principles presented in this paper provide a means of estimating forces and displacements in real structures with appropriate idealisations. Such estimates can be of considerable
use in assessing the results from more rigorous numerical analyses or they can be used in design
calculations. Examples of such usage will be presented in a subsequent paper. There are however
a considerable number of very important issues that remain to be investigated as mentioned in the
previous section. Considerable effort is required to address these issues to satisfaction before a
complete set of principles can be developed.

References
[1] J.M.Rotter, A.M.Sanad, A.S.Usmani, and M.Gillie. Structural performance of redundant structures under local res. In Interam99, 8th International Fire Science and Engineering Conference, pages 10691080, Edinburgh, Scotland, 29 June - 1 July 1999.
[2] J.M.Rotter and A.S.Usmani. Fundamental principles of structural behaviour under thermal
effects. In Proceedings of the First International Workshop on Strutures in Fire, Copenhagen,
Denmark, June 2000.
[3] A.M.Sanad, J.M.Rotter, A.S.Usmani, and M.OConnor. Composite beam in buildings under
re. 2000. Resubmitted after revision.

[4] Y.C. Wang. Tensile Membrane Action in Slabs and its Application to the Cardington Fire
Tests. Technical report, Building Research Establishment, 1996. Paper presented to the second
Cardington Conference 12-14 March 1996.