You are on page 1of 31

Chapter 4 Behaviour and Design of Steel Columns in Fire

CHAPTER 4

BEHAVIOUR AND DESIGN OF STEEL COLUMNS IN FIRE

4.1 Introduction

Columns are critical components of a structural system that undertake the major task of

transferring vertical loads down to the foundation. Failure of steel columns is a complicated

behaviour which may be axial squashing, flexural buckling, torsional buckling, local buckling

or a combination of them. Behaviour of steel columns in fire is more complicated due to

change of the material property with temperature, non-uniform temperature distribution,

possible effect from initial imperfection and residual stress, etc. In earlier times, a series of

research work has been done aimed at finding out the fire resistance of single columns with

idealized boundary conditions. Standard fire tests (Assen 1985; Wainman and Kirby, 1988)

were believed to provide the most reliable information. Due to the high cost of performing

fire tests, a number of numerical methods (Olawale and Plank 1988; Burgess and Najjar 1994;

Poh and Bennetts 1995a and 1995b) were developed to calculate the buckling load of steel

columns at elevated temperatures. However, for design purpose, analytical formulas would

always be the preferred choice. Culver et al. (1973) made the first attempt to derive an

expression for the allowable stress of columns at elevated temperatures by fitting into the

numerical study results. Janss and Minne (1981) derived an equation based on the ECCS

(1981) for column design at normal temperature. This equation is later adopted by ECCS

(1983) for fire safety design. The most recent contribution to development of analytical

formulas is from Franssen (1995a), whose proposal is partially adopted by EC3: Part 1.2

(CEN, 2001). For the capacities of beam-columns, EC3 proposed an interaction formula

similar to that for normal temperature design except that the reduced nominal bending and

compression capacities at elevated temperatures are used. Little information is available

regarding the background and validity of this equation.

Recently, research interests moved to the effect of structural continuity on the

behaviour of column. Effects of the surrounding members on the heated column can be

72
Chapter 4 Behaviour and Design of Steel Columns in Fire

classified into several aspects. Firstly, thermal expansion of the column will be restrained by

the rest building and additional axial force will be generated which accelerates the column

failure. Axial restraint also works as axial support, which, after the column failure, rapidly

transfers the axial load on the failed column to adjacent columns and maintains the stability of

the whole building. The second aspect is lateral thrust from thermal expansion of adjacent

beams that causes additional bending moments at the column ends. The final aspect is change

of stiffness of the heated column relative to the rest cooler members means that the heated

column is subjected to stronger rotational restraints at ends and its effective length is

shortened. In general, restraint from the rest of the structure will cause change to the boundary

condition and the internal force state of the beam-column.

The purpose of this chapter is to find a simple analysis method that can reasonably

reflect the real structural behaviour. In section 4.2, analytical formulas for checking the

capacity of columns and beam-columns at elevated temperatures will be firstly introduced.

Numerical simulations are then performed to get the accurate resistance of beam-columns

considering all necessary effects, such as non-uniform temperature distribution, residual stress,

initial imperfection, etc. A wide range of studies are performed to represent columns with

various slendernessES and subjected to different loadings. Finally, the analytical formulas

will be checked against the numerical study results to show the applicability of the analytical

formulas. Section 4.3 is aimed to find a solution for considering the effects of structural

continuity in design, where the following issues should be determined: 1. the failure criterion

for beam-columns at fire limit state; 2. the effective length of the columns; 3. the actions

working on the beam-column at fire limit state. In section 4.4, it will be shown that with the

knowledge from section 4.2 and section 4.3, a simple while accurate analysis and design

method for columns is achievable.

73
Chapter 4 Behaviour and Design of Steel Columns in Fire

4.2 Fire Resistance of Single Columns

4.2.1 Analytical formulas for checking capacities of columns

In general, all formulas for the resistance of compression members at elevated

temperatures take the form

Pcr , fi = χ , fi k y ,T Af y (4.1)

where Af y is the squash load of the section, k y ,T is reduction factor of the material yield

stress at temperature T and χ , fi is the reduction factor for flexural buckling in the fire design

situation. According to the proposals by Janns and Minne (1981), ECCS (1983) gives this

expression for χ , fi :

1
χ , fi = ≤ 1 .0 (4.2)
[
φ + φ − λ,T 2
2
]
0.5

where

[
φ = 0.5 1 + α (λ,T − 0.2 ) + λ,T 2 ] (4.3)

α is the member imperfection factor and is equal to 0.49, equivalent to buckling curve c at

normal temperature. λ,T is the slenderness ratio at temperature T and is calculated by

modifying the slenderness ratio at normal temperature λ with k y ,T k E ,T [ ]


0.5
. In the material

model of ECCS(1983), Young’s modulus and the yield strength degrade at similar rates and

λ can be used to replace λ,T in simplified calculations.

EC3 (1995) adopted this formula but the flexural buckling factor χ ,T is divided by 1.2

to allow for a number of effects. It should be noticed that ECCS (1983) and EC3 (1995) use

different material models and the same formula does not necessarily yield the same results.

Therefore, it cannot be concluded that EC3 is more conservative than ECCS simply because it

has introduced a factor of 1.2. For the material model in EC3, λ,T ≈ λ is no longer valid and

the non-dimensional slenderness for any temperature must be re-calculated.

74
Chapter 4 Behaviour and Design of Steel Columns in Fire

Franssen (1995) proposed some modification to EC3 (1995) based on the numerical

analysis performed using SAFIR. The flexural buckling coefficient is calculated as below:

For T ≤ 100o C , χ , fi is the same as for normal temperature design.

For T ≥ 400o C ,

1
χ , fi = (4.4)
[
φ + φ 2 − λ,T 2 ]
0.5

where

φ=
1
2
(
1 + αλ,T + λ,T
2
) (4.5)

and α = 0.65ε .

For 100 o C < θ < 400 o C , χ , fi is linearly interpolated between the values calculated at

100 and 400 o C .

This proposal is widely verified against test data (Franssen, 1996) and gives a very good

prediction to the resistance of compression members. It is accepted in the later version of EC3

(2001) with some simplification. Continuity with normal temperature design is discarded and

the expression for T ≥ 400o C is applied to the whole temperature range. This new code

formula, when compared to the original proposal of Franssen (1995a) significantly

underestimates the column resistance when the temperature is from 100 o C to 400 o C .

To check the capacities of beam-columns, EC3 (2001) proposes a formula similar to the

interaction formula for beam-columns at normal temperature. The equations for Class 1 and

Class 2 sections are reproduced below:

N fi ,Ed k y M y , fi ,Ed k z M z , fi ,Ed


+ + ≤1 (4.6)
fy fy fy
χ min, fi Ak y ,T W pl , y k y ,T W pl , z k y ,T
γ M , fi γ M , fi γ M , fi

75
Chapter 4 Behaviour and Design of Steel Columns in Fire

N fi ,Ed k LT M y , fi ,Ed k z M z , fi ,Ed


+ + ≤1 (4.7)
fy fy fy
χ z , fi Ak y ,T χ LT , fiW pl , y k y ,T W pl , z k y ,T
γ M , fi γ M , fi γ M , fi

where χ y, fi and χ z, fi should be determined according to the procedures for compression

members for the strong axis and the weak axis respectively. χ min, fi is taken as the smaller of

χ y, fi and χ z, fi .

χ LT , fi is the reduction factor for lateral torsional buckling. It is calculated from the non-

dimensional lateral torsional slenderness at normal temperature λ LT according to

1
χ LT , fi = (4.8)
φLT ,T com
+ φLT ,Tcom − λ LT ,Tcom
2 2

with

φLT ,T com
=
1
2
[
1 + αλLT ,Tcom + λLT ,Tcom ( )] 2
(4.9)

and

k y ,Tcom
λLT ,T = λLT (4.10)
com
k E ,Tcom

The factors k LT , k y and k z account for the effect of bending moment distribution and should

be determined from

µ LT N fi ,Ed
k LT = 1 − ≤1
fy
χ z , fi Ak y ,T
γ M , fi (4.11)

with µ LT = 0.15λ z ,T β M ,LT − 0.15 ≤ 0.9

µ y N fi ,Ed
ky = 1− ≤3
fy
χ y , fi Ak y ,T
γ M , fi (4.12)

( )
with µ y = 1.2 β M , y − 3 λ y ,T + 0.44 β M ,LT − 0.29 ≤ 0.8

76
Chapter 4 Behaviour and Design of Steel Columns in Fire

µ z N fi ,Ed
kz = 1 − ≤3
fy
χ z , fi Ak y ,T
γ M , fi (4.13)

( )
with µ z = 2 β M , z − 5 λ z ,T + 0.44 β M , z − 0.29 ≤ 0.8 and λ z ,T ≤ 1.1

In equation (4.11) to (4.13), β M , LT , β M , y and β M , z are factors of moment distribution that

can be directly read from design charts.

4.2.2 Numerical calculation for the capacities of beam-columns

a. Construction of the numerical model

The numerical analysis for both heat transfer analysis and structural analysis is

performed in ABAQUS. The numerical model for the steel member is shown in Figure 4.1.

Each plate that constitutes the H-shaped member is divided into 8 elements along the width

and 60 elements along the length. The element used is four-node shell elements with 4

integration points in the plane and three integration points along the thickness. In the global

coordinate system, the longitudinal axis of the column is along x axis. The major bending of

the section is about y axis and the minor bending of the section is about z axis. Thereafter, all

terms related to bending about the major axis is indicated by subscript “y” and all terms

related to bending about the minor axis is “z”. At both ends, the column is restrained against

lateral displacement and torsion, but free to rotate about either the major axis or minor axis.

Its effective length for both flexural buckling and lateral-torsional buckling is 4.0m according

to Eurocode 3: Part 1.1 (CEN, 1992). One stocky column and one slender column are studied.

Both columns are 4.0m long. Section size of the stocky column is C305×118, whose

slenderness about major axis and minor axis are λ y = 29.41 and λ z = 51.48 . The slender

column uses C150×30, whose slenderness about major axis and minor axis

are λ y = 59.17 and λ z = 104.44 .

To apply idealized boundary conditions and end loadings so that the results are

comparable to the code specification, the end section of the member is restrained to follow the

beam section law. This is done by connecting each node of the end section to the central node

77
Chapter 4 Behaviour and Design of Steel Columns in Fire

using rigid-beam elements as shown in Figure 4.1 so that the whole section could only deform

as a rigid plane. Axial force is and bending moments are applied to the central node. The

moment applied is the maximum at one end and zero at the other end about either major axis

or minor axis.

Initial imperfection and residual stress are very important in calculating the critical

strength of columns. According to the observations of Aasen (1985), the initial out-of-

straightness comprises components of the lateral displacements and twist. Lack of parallelism

between the flanges is also present but is of minor importance. Measurements from 20 test

specimens give an average initial curvature of w0 L = 0.00083 . In the current study, an

initial curvature of w0 L = 0.001 as well as α 0 = 0.006o is used and the initial deformed

shape is depicted in Figure 4.2. The pattern for residual stress distribution over the H section

is shown in Figure 4.3 with a peak stress value equal to 0.2fy following the suggestions by Poh

and Bennetts (1995b).

The stress-strain relationships for the structural steel at elevated temperature are

constructed for S275 steel according to EC3 Part 1.2. Geometrical nonlinearity is considered

by turning large deformation effect on in ABAQUS analysis so that the stiffness matrix is

updated for the current geometry after each load increment.

b. Calculation of the non-uniform temperature distribution

In reality, the temperature distribution on the columns could be non-uniform along the

length and over the section. Normally, the temperature at the ends of the members will be

lower than that in the middle span because the column is connected to cooler boundaries.

Columns could be subjected to one-side, two-side, three-side or four-side heating depending

on its position in the fire compartment. Non-uniform heating will generate temperature

gradient over the cross-section of the member. A summary of the typical mode of temperature

distribution for unprotected steel columns and the effect of non-uniform temperature

distribution on the compressive resistance of steel columns can be found from Wang (2002b).

Principally, thermal gradient will generate thermal curvature, whose effect is equivalent to

78
Chapter 4 Behaviour and Design of Steel Columns in Fire

enlarged out-of-straightness. The thermal gradient will, at the same time, increase the section

capacity and thus may enhance the buckling resistance.

To get the real temperature distribution, heat-transfer analysis should be performed.

ABAQUS provides the function of sequentially coupled thermal and stress analysis where

heat transfer analysis is conducted first and the temperature distribution is then read into to

the stress analysis from node to node. The shell element used for heat transfer analysis is DS4

which allows heat transfer through the thickness of the shell by specifying a number of

integration points along the thickness of the element and allocating the corresponding

numbers of degree of freedom for temperatures to each node. Considering that columns are

normally protected by fire protections, 5mm thick gypsum board is used to wrap the column.

Temperatures at both ends of the column tend to be lower because of the heat sink effect from

the floor slab and other constructional members. This effect is symbolically considered by

attaching two concrete plates to each end of the column. Thus, the whole model for heat

transfer calculation includes three parts as shown in Figure 4.4. The gypsum board is divided

into four layers by 5 integration points along the thickness. The outer surface receives heat

from the fire source and heat is transferred to the inner surface of the gypsum board. The heat

is then transferred from the inner surface of the gypsum board to the steel member by gap

conductance and cavity radiation as shown in Figure 4.5. The material thermal properties and

the heat transfer parameters are also shown in the figure. At both ends, the column loses some

heat to the concrete plate by conduction and the temperatures at the end of the column are

assumed to be the same as the inner surface of the concrete plates. Three representative

column fire exposure conditions are considered. In the first case, the column is embedded in

the compartment wall with only one flange exposed to fire. In the second case, the column is

at the corner of the compartment with two sides attached to the wall surface. In the third case,

the column is inside the compartment with only one flange attached to the wall surface. They

are respectively designated as one-side exposure, two-side exposure and three-side exposure.

Standard ISO834 curve is used as environmental temperature and the analysis is continued

79
Chapter 4 Behaviour and Design of Steel Columns in Fire

until the maximum temperature reaches 700 o C or more. Three levels of heating that

representing the low, medium and high temperature are taken from each fire exposure for

each member. The temperature distribution along the height of the member for various section

points are shown in Figure 4.6 and Figure 4.7. The reference temperatures shown in the

figures are the maximum temperatures of the member. About the temperature distribution, it

can be observed that

1. Along the length of the member, the temperature distribution is almost uniform over the

middle 3.0m. At both ends, the temperatures are decreased by 200 to 400oC within the

range of 0.5 meter from each end.

2. Over the section, one-side exposure and three-side exposure imposes thermal gradient

along the weak axis only. For the fire protected columns, it is found that the thermal

gradient is very small. Generally, the temperature difference over the section is less than

100oC and with the increase of the average temperature, this difference will tend to

diminish.

3. Two-side heating generates much larger temperature difference over the section. The

temperature difference between the hottest tip and the coolest tip could be over 400oC.

However, with the increase of the average temperature, the temperature will be more

uniform. By end of the analysis, the temperature difference is around 100oC only.

In the subsequent structural analysis, the temperature of the steel column is transferred

from the heat transfer analysis result file to the structural analysis model from node to node.

The non-structural parts of the model, i.e. the fire protection and the concrete plate are

removed from the model.

4.2.3 Comparison of the analytical formulas and the numerical calculation

a. Results for Uniform Temperature Distribution

The resistance of the column is firstly analyzed when the temperature is assumed to be

uniformly distributed over the whole member. For each member, the interaction curve is

established for N-My and N-Mz at three temperatures 300oC, 550oC and 700oC. The results are

80
Chapter 4 Behaviour and Design of Steel Columns in Fire

shown in Figure 4.8 by the dotted points. The curves are interaction relationships according to

equations (4.6) and (4.7). In Figure 4.8,

My Mz N
my = ; mz = ; n= (4.14)
f yW y f yWz fyA

Comparison of the numerical analysis results show that for the axial compression capacity,

the EC3 on average gives very good prediction of the ultimate resistance. The minor axis

bending capacity is slightly underestimated and the major axis bending or lateral-torsional

buckling capacity is significantly underestimated. In general, the interaction curve generated

by the numerical analysis is larger than the EC3 analytical formula, especially at lower

temperatures.

Except for the slender column C152x30 at 300oC, the calculated compressive

resistances for all the other studied cases are smaller than the code specification. This

difference is small enough to be compensated by safety factors introduced in design. However,

it should be borne in mind that the limiting state for the compressive members calculated

from the EC3 is very close to the real failure state.

When the member is subjected to bending moment about the weak axis alone, its

maximum resistance should be the plastic section capacity. In Figure 4.8, the maximum

resistance given by the numerical analysis is 10-20% larger than the plastic section capacity.

The deformation of the beam near the ends is shown in Figure 4.9. Obviously, the

deformation of the end does not follow the Timoshenko beam theory anymore. Both the

flange and web near the ends are subjected to plane stress. The material yielding is controlled

by the two-dimensional Von-Mises yield surface. Assuming that the shear stress σ xy ≈ 0 , the

Von-Mises yield surface can be plotted as shown in Figure 4.9 where σ 0 is the one-

dimensional yield stress according to the stress-strain relationship. The maximum stress the

material can take in the longitudinal direction is 1.15 times of the one-dimensional yield stress.

This explains why plate theory gives higher moment resistance than the beam theory, which is

based on one-dimensional stress state.

81
Chapter 4 Behaviour and Design of Steel Columns in Fire

The lateral-torsional buckling resistance calculated from the numerical model is much

larger than the EC3 prediction. Enhanced bending resistance due to the two-dimensional

stress state as explained in the previous paragraph could play only a very small role to this

difference. Considering that in the study of Vila Real et al. (2004), the Eurocode interaction

formula gives very good correspondence to the numerical prediction from SAFIR, the

difference between these two studies could be only explained to be due to the different

moment distributions adopted. Moment distribution also imposes significant effect on the

lateral-torsional buckling moment. A study of the lateral-torsional buckling moment of the

two members under different kinds of moment distribution is shown in Figure 4.10. It can be

seen that when ψ=1 (uniform moment distribution), the numerical analysis results is almost

exactly the same as the Eurocode 3 specification. However, when ψ is from -1 to 0.25, the

numerical analysis results are 20% larger when T=300oC and 27% larger when T=600oC.

Study to the slender member C152x30 yields similar results. The pure lateral-torsional

buckling capacity of the beam-column studied in Figure 4.8 corresponds to the value when

ψ=0 in Figure 4.10. It is then not surprising that in the current study, the calculated major axis

resistance is much larger than the code specification.

b. Resistance of Beam-columns with Real Temperature Distribution

For the three fire exposures obtained in section 4.2.2.b, the interaction curves for

combined effect of axial force and major axis moment or combined axial force and minor axis

moment are constructed for the three temperature distributions of Figures 4.6 and 4.7. The

numerical analysis results versus the EC3 analytical formula are shown in Figure 4.11 to

Figure 4.14. The analysis results generally show similar trend to columns that are uniformly

heated. The two typical capacities- compression and bending will be discussed separately and

the interaction curve as a whole will be commented:

Axial compression: Lower temperature at the ends of the member should help to

increase the compressive resistance. This effect can be observed by compare the compressive

resistances of one-side heating and three-side heating at lower temperature of around 300oC.

The temperature change along the length of the one-side heated member is slower than the

82
Chapter 4 Behaviour and Design of Steel Columns in Fire

three-side heated member and the compressive resistance is higher. The columns studied are

pin-supported and both ends are subjected to relatively less critical stress state. The effect of

cooler temperature at the ends may be more significant if other boundary conditions are

studied. Thermal gradient will decrease the compressive resistance if it is small and increase

the resistance if it is large. Observing all compressive resistances for one-side and three-side

heating, the compressive resistance is more often lower than the code specification at lower

temperatures than at higher temperatures because thermal gradient is generally larger at lower

temperatures. The two-side heating is found to generate significantly larger temperature

variance over the section and in all cases subjected to two-side heating, the calculated

compressive resistance is larger than the code specification.

Bending: For both major axis and minor axis bending, it was found that the temperature

gradient- either along the length or over the section, will enhance the bending capacity. The

larger is the temperature variance, the more prominent will be the enhancement.

In general, the interaction curve proposed by the Eurocode 3 is on the safe side. The

more predominant is bending moment on the failure of a beam column, the more conservative

will the design be. The effect of non-uniform temperature distribution is enhance the bending

or lateral-torsional buckling resistance. However, for the axial compressive capacity, the

current database is not enough to draw any conclusive results about the effect of non-uniform

temperature distribution. For a tentative design recommendation, it may be suggested that

1. when temperature gradient exists about both axes, the non-uniform temperature

distribution tends to yield larger fire resistance than uniform temperature distribution.

2. when temperature gradient exists about only flexural buckling axis, for slender members

( λθ ≥ 1 ), the fire resistance tends to be reduced by the temperature gradient. For stocky

members ( λθ ≤ 1 ), the fire resistance tends to be increased by the temperature gradient. As

the range of temperature difference studied is generally less than 100oC, the effect of

temperature gradient is very small and may be ignored in the approximate calculation.

83
Chapter 4 Behaviour and Design of Steel Columns in Fire

3. for beam-columns subjected to both axial force and bending moment, if F Pc is smaller

than 0.8, it may be supposed that the EC3 analytical formula is conservative. Otherwise, EC3

gives on average the failure state.

4.3 Behaviour of Restrained Columns

Possible effects of structural interaction on the behaviour of a column in fire can be

summarized into three categories:

1. change in the column axial load

2. change in the column buckling length

3. change in the column bending moment

The change in the column axial load occurs because the axial expansion of the column

with the increase of temperature is restrained by the rest of the structure. The change in the

column buckling length is due to the variance of the relative stiffness of the objective column

and the other adjacent members at heated condition. Change in the column bending moment

could be the result of the relative stiffness change at the beam-to-column connection or the

lateral displacement due to the thermal expansion of adjacent beams. All three effects have

been widely investigated. This chapter will not repeat the previous work. Instead, the research

conclusions by previous researchers will be summarized to form useful design guidelines.

4.3.1 Effect of axial restraint on behaviour of columns

The complete load-temperature relationship of a restrained column is described in

Figure 4.15. In range AB, the axial load increases with the increase of temperature. The top

end of the column moves upward. At point B, the column reaches its maximum load

resistance Pmax and buckles with a sudden lateral movement. The axial load drops to point C

suddenly. After buckling, the column still retains some load bearing capacity. However, the

post-buckling strength decreases with the increase of temperature and comes to zero gradually.

The mathematical representation of the phases AB, BC, and CD can be found from Wang

(2002b). In reality, the restrained column will behave as shown by the dotted line due to the

84
Chapter 4 Behaviour and Design of Steel Columns in Fire

existence of initial deficiency. The increasing and decay rate of the axial load depends on the

stiffness of axial restraint and the extent of initial deficiency (Neves, 1995).

The limiting temperature of restrained columns has been investigated based on two

criterions. The first one is the buckling temperature in Figure 4.15 (Wang, 1997). Obviously,

the limiting temperature of a restrained column would be lower than that of a free column if

this criterion is adopted. By equating the maximum axial load to the load resistance

calculation method recommended by EC3, Wang’s study showed that the limiting

temperature is decreased with the increase of the restraint stiffness. At a rational scale of

restraint stiffness and load ratio, the decreasing extent varies from 10% at a lower slenderness

of λ = 60 to 30% at a higher slenderness of λ = 120 . The other one is the failure temperature

of Figure 4.15 at which the load bearing capacity equals to the initial load of the column. This

criterion is more reasonable considering that the column should maintain its design load

bearing capacity. Based on this criterion, Franssen (2000) proposed a simple method to obtain

the limiting temperature of steel columns considering the effect of axial restraint only. His

method is based on the observation that when the stiffness of the axial restraint is strong

compared to that of the column, a column with various axial restraint stiffness follows the

same post buckling curve as shown in Figure 4.16. Franssen’s study is on columns in ideal

compression. Study on eccentrically loaded columns shows that when the load eccentricities

are small, the axial restraint will decrease the limiting temperature; when the eccentricity of

the compression load is great enough, the limiting temperature has no dependence on the

restraint stiffness (Neves, 1995).

4.3.2 Buckling length of columns in fire

At normal temperature, the effective length of a column is calculated from its stiffness

relative to those members connected to it at both ends. In fire condition, the column is heated

and its stiffness is significantly reduced. It is normally acknowledged that the surrounding

structure will be stiffer compared to the column in fire and the effective length of the column

should be reduced. A systematic study was performed by Wang (1997). In his study, a sway

sub-frame and a non-sway sub-frame were established. The effective length of the column

85
Chapter 4 Behaviour and Design of Steel Columns in Fire

was obtained by equating the fire resistance of a single column to the fire resistance of the

sub-assembly. His conclusion was that the effective length factor of the column is 1.1 for

sway frame and 0.55 for non-sway frame in fire condition independent of its effective length

factor at normal temperature.

4.3.3 Change of column moment in fire

The column moment changes in fire due to two reasons. Firstly, a change in the

stiffness of the heated column relative to the adjacent members will cause re-distribution of

the bending moment at the connection. Secondly, thermal expansion of the heated beam will

impose lateral movement to the column and thus introduce bending moment in the column.

Another effect of the lateral movement is the moment generated from P-δ effect of the column

axial load.

The change of bending moment due to the change of relative stiffness was

experimentally studied by Wang (2003). It was observed that if pin-connection was used to

connect the beam and column, little bending moment existed at the column end during the

whole burning range. In case that end-plate connection was used, the initial bending moment

in the column was decreased rapidly with the increase in temperature. By the time of column

failure, the column moment was very small. Therefore, the column failure temperatures were

mainly dependent on the total applied load with little influence of the connection type and

beam load. However, the tests were performed with the column and only a very short part of

the beam at both sides of column heated in the furnace. Its application to the normally

assumed compartment fire exposure needs further verification.

The change in column bending moment due to the lateral movement of the beam was

studied by Bailey (2000). In his structural model, the column was subjected to the restraint of

one beam along major axis and minor axis respectively. The beams were unloaded and the

redistribution of the beam moment at beam-to-column connection was removed. The only

factors studied were the moment generated due to the lateral expansion of the beam and P-δ

effect of the column axial load. The study result showed that when the axial load was zero,

the bending moment increased from zero because of the lateral movement of the beam. After

86
Chapter 4 Behaviour and Design of Steel Columns in Fire

certain time, it began to decrease. By the time of column failure, the bending moment in the

heated column tended to become zero. This can be easily explained by the simple model in

Figure 4.17. At normal temperature, the moment distribution of the member when subjected

to a displacement δ at middle span is shown in Figure 4.17b. With an increase in temperature,

δ will increase because of the expansion of the beam and the Young’s modulus E is decreased.

Initially, the increase in δ controls the response of the member. However, when approaching

failure, the reduction of E will control the member response. If only the left column CL is

heated as shown in the figure, near failure, the middle point can be taken as a fully restraint

for CL and pin connection for CR; the moment distribution should be as shown in Figure 4.17c.

In reality, the columns generally would not be allowed to be heated to so high that its end

moment becomes negligible because with existence of axial force, it would have failed long

before it get excessively softened. In Bailey’s study, the steel beam is not loaded. If the steel

beam is loaded, the catenary action of the beam will be developed at high temperature and the

additional moment due to the lateral movement of the beam could be generated and its

magnitude could be very large (Yu and Liew, 2005).

The bending moment due to P- δ effect can be statically determined by the amount of

the axial load and the deformed shape of the column. This effect will never diminish with the

increase of the temperature and may become detrimental at very large deflection. The

parametric studies performed by Bailey (2000) showed that with the existence of an axial load,

the bending moment at failure was much larger than without axial load and the moment

diagram along the column was very hard to predict because of the multi-possibilities for the

buckled shape of the column. However, for framed structures with its beam in catenary action

phase, the additional moment due to P- δ could also be easily calculated as shown in Figure

3.5.

To sum up, it may be concluded that no matter the column is in simple or continuous

connection, when heated, its initially applied moment will get diminished. By failure, the

bending moments on the column would be mainly due to catenary action of beams and P- δ

effect of the axial load. These two parts can be determined according to the relevant theories

87
Chapter 4 Behaviour and Design of Steel Columns in Fire

regarding the catenary action of beams and calculated as shown in Figure 3.5. It should be

mentioned that in EC3, the reduced column slenderness in fire is used with the axial force and

bending moment at normal temperature. While it is reasonable to assume the axial force

equals to the initial value because by failure, the excessive axial force will be released; the

bending moment may be larger than that at normal temperature. The EC3 recommendation

tends to be unsafe.

4.4 Design of Columns

When designing columns in a building, it would be the best if the actions on the column

can be accurately calculated as well as the member capacity. It has been discussed in Section

4.2 that EC3 does provide an analytical expression for the capacities of beam-columns at

elevated temperature. Taking into account the complexity of the problem, we have to say that

the performance of the current expression is satisfactory, although it can be further improved.

The actions on the member change with temperature increase and it would be difficult to

determine the development of actions in the temperature history. However, if appropriate fire

limit state can be defined, the axial force and bending moment can be determined specifically

for the fire limit state. A fire limit state has been established for beams based on catenary

action in Chapter 3. If the principle of the framed system design is that the columns should

provide sufficient fire resistance so that the system fails by the failure of beams in the

catenary action phase, then a design system can be established for columns.

Approximately, the development of actions on a beam-column will be like the arrowed

curve shown in Figure 4.18. The axial force keeps increasing due to the axial restraint. The

bending moment, although probably variable initially, will eventually get increasing as the

development of the catenary actions in the beams. This situation continues until the member

hits the boundary of the N-M interaction curve at temperature T1. After that, the member will

not fail immediately due to the post-buckling strength. In the post-buckling range, the bending

moment keeps increasing, but the axial force will be released rapidly and the column top

moves downward. According to the failure criterion for columns defined by Franssen (2000),

88
Chapter 4 Behaviour and Design of Steel Columns in Fire

at fire limit states, the column is required to support load that equals to its initial axial force.

Therefore, the failure point of the column is when its axial force is reduced to be equal to its

initial value. By that time, the temperature may be increased to T2 and the residual moment

capacity is M n ,T2 . To make sure that the beam failure happens before the column failure, the

residual moment capacity M n ,T2 must be able to support the bending moment generated by the

catenary action of beams. Or by using the analytical formulas presented in section 4.2.2, it

should be satisfied that

Γ(N 0 , M T , y , M T , z ) ≤ 1.0 (4.15)

where Γ represents the left side of equation (4.6) and (4.7); M T , y and M T , z are moments

generated at the column ends due to the catenary action calculated from Figure 3.5.

89
Chapter 4 Behaviour and Design of Steel Columns in Fire

FEM Model

Central
node
Rigid beams

y
4 3
n
Element Formulation
z
x
1 2 Integration
along the
Four-node shell element thickness
with full integration

Figure 4.1 Construction of the FEM model for H-shaped steel member using shell
elements

w0 = 4mm

z
α = 0.006 o

y
Figure 4.2 Initial imperfections for the column

90
Chapter 4 Behaviour and Design of Steel Columns in Fire

0.2 f y + tension
+
- - 0.2 f y - compression
+

0.2 f y -

0.2 f y
- + -
+
0.2 f y
Figure 4.3 Residual stress pattern assumed in analysis

Figure 4.4 Meshed 3D finite element model for heat transfer analysis

91
Chapter 4 Behaviour and Design of Steel Columns in Fire

Gap
Cavity
conductance
radiation

Cavity Five integration


radiation points to simulate the
heat transfer through
the thickness of the
gypsum board

a. an illustration of the finite element model.

b. 2D view of the finite element model of


the column cross-section with fire
protection

Gypsum board concrete steel


Conductance (W/mK) 0.2 2 (average) 27 (average)
Mass (kg/m3) 1700 2000 7850
Specific heat (J/kgK 800 1200 (average) 600 (average)

Figure 4.5 Illustration to the heat transfer analysis for steel columns with fire protection

92
Chapter 4 Behaviour and Design of Steel Columns in Fire

4
T1 T2
3.5
T3
3
Length (m)

2.5
t=63 t=121 t=180
2 T3
1.5
1 T2
0.5 T1

0
0 100 200 300 400 500 600 700 800 900 1000
o
Temperature ( C)

4
T1 T2
3.5
T3 T4
3 T5
Length (m)

2.5
2 t=22 t=50 t=102
T3 T5
1.5

1 T2
T1 T4
0.5

0
0 200 400 600 800 1000
Temperature (oC)
4
T1 T2
3.5
T3
3
Length (m)

2.5
t=24 t=50 t=10
2 T3
1.5
T2
1
T1
0.5

0
0 200 400 600 800 1000
o
Temperature ( C)

Figure 4.6 Temperature distributions of column C305x118 under one-side, two-side and
three-side heating

93
Chapter 4 Behaviour and Design of Steel Columns in Fire

4
T1 T2
3.5
T3
3

2.5
Length (m)

t=40 t=83 t=129


2
T3
1.5

1 T2
T1
0.5

0
0 200 400 600 800 1000
o
Temperature ( C)

4
T1 T2
3.5 T3 T4
3 T5
Length (m)

2.5

2 t=18 t=43 t=69 T3 T5


1.5

1 T2
T1 T4
0.5

0
0 200 400 600 800 1000
o
Temperature ( C)

4
T1 T2
3.5
T3
3
Length (m)

2.5

2 T3
t=17 t=32 t=51
1.5
T2
1 T1
0.5
0
0 200 400 600 800 1000

Temperature (oC)
Figure 4.7 Temperature distributions of the column C152x30 under one-side, two-side and
three-side heating

94
Chapter 4 Behaviour and Design of Steel Columns in Fire

1.2
T=300 (EC3)
1
T=550 (EC3)
a. 0.8
T=700 (EC3)
my 0.6
C305x118 T=300 (analysis)
0.4 T=550 (analysis)
my-n 0.2 T=700 (analysis)
0
0 0.2 0.4 0.6 0.8
n

1.2
1 T=300 (EC3)
T=550 (EC3)
0.8
b. mz 0.6 T=700 (EC3)
T=300 (analysis)
C305x118 0.4
T=550 (analysis)
0.2
T=700 (analysis)
mz-n
0
0 0.2 0.4 0.6 0.8
n

1
c. T=300 (EC3)
0.8
T=550 (EC3)
C152x30 0.6
T=700 (EC3)
my 0.4 T=300 (analysis)
my-n
0.2 T=550 (analysis)
T=700 (analysis)
0
0 0.1 0.2 0.3 0.4
n

1.4
1.2 T=300 (EC3)
d. mz
1 T=550 (EC3)
C152x30 0.8 T=700 (EC3)
0.6 T=300 (analysis)
mz-n 0.4
T=550 (analysis)
0.2
T=700 (analysis)
0
0 0.1 0.2 0.3 0.4
n

Figure 4.8 Resistances of the two columns subjected to uniform heating

95
Chapter 4 Behaviour and Design of Steel Columns in Fire

1.5

0.5
σ y σ0
1.15
0
-1.5 -1 -0.5 0 0.5 1 1.5
-0.5

a. Deformed shape of the beam -1


end (amplified by 10)
-1.5
σx σ0
b. Von-mises yield surface

Figure 4.9 Deformation and stress state of the beam near the ends

0.8
Numerical analysis
0.6 T=300oC
My EC3
M0 0.4

0.2
T=600oC
My ψMy

0
-1 -0.5 0 0.5 1
ψ
a. C305x118

0.8

0.6
My Numerical analysis
M0 EC3 T=300oC
0.4

0.2
My ψMy T=600oC
0
-1 -0.5 0 0.5 1
ψ
b. C152x30

Figure 4.10 Effect of moment distribution on the Lateral-torsional buckling moment of beams

96
Chapter 4 Behaviour and Design of Steel Columns in Fire

1.2
T=291 (EC3)
a. 1
T=535 (EC3)
0.8
C305x118 T=683 (EC3)
my 0.6
T=291 (analysis)
One-side 0.4
T=535 (analysis)
0.2
heating T=683 (analysis)
0
0 0.2 0.4 0.6 0.8
n

1.2
T=307 (EC3)
1
b. T=550 (EC3)
0.8
C305x118 T=740 (EC3)
my 0.6
T=307 (analysis)
Two-side 0.4
T=550 (analysis)
0.2
heating T=740 (analysis)
0
0 0.2 0.4 0.6 0.8
n

1.2
T=335 (EC3)
c. 1
T=574 (EC3)
0.8
C305x118 T=760 (EC3)
0.6
my T=335 (analysis)
Three-side 0.4
T=574 (analysis)
0.2
heating T=760 (analysis)
0
0 0.2 0.4 0.6 0.8
n

Figure 4.11 Resistance of C305x118 subjected to My and N

97
Chapter 4 Behaviour and Design of Steel Columns in Fire

a. 1.2

C305x118 1 T=291 (EC3)

0.8 T=535 (EC3)


One-side T=683 (EC3)
0.6
heating
mz 0.4 T=291 (analysis)
T=535 (analysis)
0.2
T=683 (analysis)
0
0 0.2 0.4 0.6 0.8
n
b.
1.2
C305x118 1 T=307 (EC3)
T=550 (EC3)
Two-side mz 0.8
0.6 T=740 (EC3)
T=307 (analysis)
heating 0.4
T=550 (analysis)
0.2
T=740 (analysis)
0
0 0.2 0.4 0.6 0.8
n
c. 1.2
T=335 (EC3)
C305x118 1
T=574 (EC3)
0.8
Three-side T=760 (EC3)
mz 0.6
T=335 (analysis)
heating 0.4
T=574 (analysis)
0.2
T=760 (analysis)
0
0 0.2 0.4 0.6 0.8
n

Figure 4.12 Resistance of C305x118 subjected to Mz and N

98
Chapter 4 Behaviour and Design of Steel Columns in Fire

1.0
T=300 (EC3)
a. my 0.8
T=560 (EC3)
0.6
C152x30 T=700 (EC3)
0.4 T=300 (analysis)
One-side
0.2 T=560 (analysis)

heating 0.0
T=700 (analysis)

0.0 0.1 0.2 0.3 0.4


n

1
T=300 (EC3)
my
b. 0.8
T=560 (EC3)
0.6 T=700 (EC3)
C152x30
0.4 T=300 (analysis)
Two-side
0.2 T=560 (analysis)

heating 0
T=700 (analysis)

0 0.1 0.2 0.3 0.4


n
c.
1
T=325 (EC3)
C152x30 0.8
T=550 (EC3)
my
Three-side 0.6 T=700 (EC3)
0.4 T=325 (analysis)
heating
T=550 (analysis)
0.2
T=700 (analysis)
0
0 0.1 0.2 0.3 0.4
n

Figure 4.13 Resistance of C152x30 subjected to My and N

99
Chapter 4 Behaviour and Design of Steel Columns in Fire

a. 1.4
1.2 T=300 (EC3)
C305x118 mz 1.0
T=560 (EC3)
One-side 0.8
T=700 (EC3)
0.6
heating 0.4
T=300 (analysis)
T=560 (analysis)
0.2
0.0 T=700 (analysis)
0.0 0.1 0.2 0.3 0.4
n
b. 1.4
1.2 T=300 (EC3)
C305x118 mz 1 T=560 (EC3)
Two-side 0.8 T=700 (EC3)
0.6
T=300 (analysis)
heating 0.4
T=560 (analysis)
0.2
0 T=700 (analysis)

0 0.1 0.2 0.3 0.4


n
c.
1.4
T=325 (EC3)
C305x118 1.2
T=550 (EC3)
1
Three-side mz 0.8 T=700 (EC3)
0.6 T=325 (analysis)
heating 0.4 T=550 (analysis)
0.2
T=700 (analysis)
0
0 0.1 0.2 0.3 0.4
n

Figure 4.14 Resistance of C152x30 subjected to Mz and N

100
Chapter 4 Behaviour and Design of Steel Columns in Fire

P
P Pre-buckling Post-buckling
Pmax
B
Buckling
C

P0
A
D
0
Temperature
Buckling Failure
temperature temperature

Figure 4.15 Temperature-axial load relationship of an axially restrained column

Figure 4.16 Evolvement of axial force at different axial restraint stiffness


(Franssen, 2000)

101
Chapter 4 Behaviour and Design of Steel Columns in Fire

(a) CL CR
δ
6 E0 I 6 E0 I
δ0 δ0
(b) L2 L2

6 E0 I
δ0
L2 3E0 I
6 ET I δ
δ L2
L2
(c)
6 ET I
δ
L2

Figure 4.17 Bending moment in the column due to lateral movement of the beam

T1

Fire limit
state
N0
Initial T2
state

M n ,T2 M

Figure 4.18 Development of column actions in fire

102

You might also like