J. Construct.

Steel Research 35 (1995) 49-69

0143-974X(94)00042-5 ELSEVIER

Elsevier Science Limited Printed in Malta. 0143-974X/95/$9.50

A Simple Model for the Fire Resistance of Axially-loaded Members According to Eurocode 3 Jean-Marc Franssen
Institut du G6nie Civil, Universit~ de Li6ge, Quai Banning 6, B-4000 Liege, Belgium

& Jean-Baptiste Schleich & Louis-Guy Cajot

ProIilARBED Recherches, Route de Luxembourg, Esch/Alzette, Luxembourg (Received 16 March 1994; revised version received 27 July 1994; accepted 10 October 1994)

ABSTRACT A general model, i.e. a non-linear computer code, has been extensively used to determine the buckling load of axially-loaded members according to the hypotheses of Eurocode 3, Part 10 or Part 1.2. Two yield strengths, two buckling axes, six ultimate temperatures and 10 different lengths have been considered for 339 different steel H-sections. The numerical results have been statistically sorted and compared to the simple models presented in Eurocode 3, Part I0 and Eurocode 3, Part 1.2. The simple models that have been proposed up to now can lead to different safety levels when compared to the general model, the safety level depending mainly on the buckling length. A new proposal has thus been made for a simple model that systematically ensures a conservative result when compared to the oeneral model.

1 INTRODUCTION
In Eurocode 3, Part 10,1 two different calculation methods are presented to determinate the fire resistance of axially-loaded members: The general calculation model is based on acknowledged principles and assumptions of the theory of structural mechanics, taking into account the effect of temperature. The application of this m e t h o d requires the utilisation of non-linear c o m p u t e r programs,
49

50

J.-M. Franssen, J.-B. Schleich, L.-G. Cajot

The simple calculation model is based on an analytical formula which can be applied with the use of tabulated data and/or a pocket calculator. This analytical formula has been recently amended in Eurocode 3, Part 1.2.2 It is considered, as stated in Ref. 1 that 'the general calculation models lead to a more realistic analysis of structures exposed to fire. Compared with simple calculation models, they give an improved approximation of the actual structural behaviour under fire conditions'. Simplified methods should therefore be allowed only if they prove to be safer than the more general methods. This paper presents some results of a research project. The aim of the research was to verify the validity of the simplified methods presented i Ref. 1 and in Ref. 2 and to make an alternative proposal if necessary. A very large number of examples was analysed by the general model and the results then compared to the results obtained using the analytical proposals.

2 THE GENERAL MODEL

2.1 The numerical code

The general model was applied by the aid of the computer programme SAFIR, 3 a non-linear finite element programme which analyses the behaviour of structures submitted to fire with a step-by-step simulation. The 2-D beam finite element that has been used in this study is based on the following formulations and hypotheses: Displacement type element in a total corrotational description. Prismatic element. The displacement of the node line is described by the displacements of three nodes, two nodes at each ends of the element supporting two translations and one rotation plus one node at mid-length supporting the non-linear part of the longitudinal displacement. The longitudinal displacement of the node line is a second-order power function of the longitudinal co-ordinate. The transversal displacement of the node line is a third-order power function of the longitudinal co-ordinate. The plane sections remain plane and perpendicular to the node line, i.e. the shear energy is not considered (Bernoulli hypothesis). The strains are small (Von Karman hypothesis), i.e.
1 8u ~ t~-~ 1

(1)

Fire resistance of axially-loaded members

51

The non-linear part of the strain is averaged on the length of the element to avoid locking, i.e.
+L/2

e:' = ~xx + ~

~u 1 f (~v~2dx \dx]
Lt2

(2)

The,,angles between the deformed node line and the undeformed but translated node line are small, i.e. sm ~p~ (p cos ~p~- 1 (3)

The', longitudinal integrations are numerically calculated using Gauss' method. The, cross-section is discretised by means of triangular or quadrilateral fibres. At every longitudinal point of integration, all variables, suclh as temperature, strain, stress, etc., are uniform in each fibre. The tangent stiffness matrix is evaluated at each iteration of the convergence process ( N e w t o n - R a p h s o n method). Residual stresses are considered by means of initial- and constantstrains. 4 The,' material behaviour in case of strain reversal is elastic, with the elastic modulus equal to the Young's modulus at the origin of the stress-strain curve. The plastic strain is presumed not to be affected by an increase in temperature, s In one cross-section, some points may exhibit a tangent modulus because they are still on the loading branch, whereas at the same time some other points behave elastically.

2.2 Hypotheses of the numerical simulations End conditions: simply supported, i.e. rotation not restrained. Load eccentricity: no eccentricity, i.e. axial load. ColLumn imperfection: sinusoidal, maximum value = H/IO00. Dead weight: neglected, i.e. only one-half of the length is simulated, by means of five beam elements. Longitudinal integration: two Gauss points. Cross-section: only one-half is simulated. The dimensions are according to Fig. 1. where only one-quarter of the section is represented.

In Fig. 1: d: dimension of the triangle having the same surface as the root fillet. (2r) 2 -- rcr2 = 4(dZ/2); d = 0-655r; (4)

52

J.-M. Franssen, J.-B. Schleich, L.-G. Cajot B/20 B/40

i' l'ldl

r i
Web

Flange

Fig. 1. Cross-section of one-quarter of an I-section.

ncou: number of layers on the thickness of the flange. Dimension = tf/ncou, with t f = flange thickness; ns: number of layers in one-half of the flange. D i m e n s i o n = ( B / 2 - 3 B / 4 0 - a / 2 - 2d)/ns with B = width of the section, a = web thickness. na: number of layers in one-half of the web. D i m e n s i o n = ( H I 2 - t f - d ) / n a with H = height of the section. The values for ncou, ns and na in the simulations have been set according to Table 1.
TABLE 1 Parameters of the Discretisation
Buckling around ncou na ns Major axis 3 5 5 Minor axis 1 4 10

Fire resistanceof axially-loaded members

53

--

Residual stresses: constant across the thickness of the web and of the flanges; -triangular distribution; - - maximum v a l u e = 0 . 3 0 x 2 3 5 M P a if H/B>1.20, = 0 . 5 0 x 235 M P a if H/B <~1.20; - - no residual stress in the triangle accounting for the root fillet.

3 THE NUMERICAL
3.1 Uniform temperature

SIMULATIONS

The situation of uniform temperature has been considered as the limit case of highly-insulated sections. An important consideration is that with thermal insulation the heating rate is limited and the high thermal diffusivity of steel has time to homogenise the field of temperature in the section. The otlher limit case leading to the most severe thermal gradient in the sections is when unprotected profiles are submitted to the ISO heating. This will be considered in the next section.
Strategy

The nominal yield strength of 235 M P a has been considered first. The nominal value has been adapted depending on the thickness of the flange according to EN 10025, 6 i.e.

fy = 23:5 M P a
fy=225 MPa fy=215MPa

fy=195 MPa

if if if if

16 40 100

tf<<.16 mm <tf~<40mm <tf~<100mm <tf

The sections which are classified under class 4 in pure compression, according to Eurocode 3, Part 1, 7 have not been considered because the beam element does not allow local buckling to be taken into account, whereas it is clearly stated in Ref. 1 that "If there are failure modes, not

covered by the general calculation model, they shall be excluded by appropriate detai,!ing.' The fire resistance of class 4 profiles should therefore be
assessed by means of experimental tests or by means of numerical simulations taking local buckling into account, which is beyond the scope of this project. The H-sections from the Arbed company's catalogue 8 have been considered. In this catalogue, some sections have different names but the same geometrical characteristics: HD400 x 216 = W360 x 410 x 216. In that case,

54

J.-M. Franssen, J.-B. Schleich, L.-G. Cajot

one of the two twin profiles has been eliminated, which left 400 different H-sections. F o r grade $235, 61 sections are classified as class 4 under axial loading, according to Ref. 7, 339 sections had thus to be considered: 34 128 6 28 41 28 32 42 HD HE HL HP IPE UB UC W sections sections sections sections sections sections sections sections

F o r each section, buckling around the major axis, as well as around the minor axis, have been taken into account separately. F o r each section and buckling axis, 10 different lengths have been considered, with the slenderness at ambient temperature 2 equal to 20, 40, 60 . . . . ,200. F o r a fixed length, the curve describing the decrease of the ultimate load as a function of the temperature can be established if a very large number of simulations is performed with different loads. It appears then that this curve is made of successive linear segments, one of them from 400 to 500C, the next one from 500 to 600C, and so on (Fig. 2). This is due to the fact that in the hypotheses of Eurocode 3, Part 10 the parameters describing the material law at elevated temperature are interpolated every

90000

..................... i. . . . . . . . . . . . . . .

i............................. !.............................. i

80000 ~ - . . ~ ........... i..............................i............................i..............................:, .............................!

,oooo t

. . . . . . . . . . .

60000 ~...................... i " ~

............i........................1

Int.erpolated

l!

2o000

~
.

"
. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . .

~
. . . . . . . . . . . . . . . . . . . . . . . . . .

....................f.............................i

,o ot ............................ ii ol
400 500

L
600 TEMPERATURE

ii

~
800

J
g00

700 [*C]

Fig. 2. Validation of the method of linear B-spline functions. Note: In order to obtain sufficient precision on the interpolated values for N(400C) and N(900C), at least one original point must exist for a temperature between 400 and 430C, and one point between 870 and 900C.

Fire resistance of axially-loaded members

55

100C. This multilinear shape is observed only when the temperature is uniform across the section. For each length, an average of 15 different loads has been applied in succession and the ultimate temperatures corresponding to the loss of stability have been recorded. Due to the fact that the load curve is multilinear, linear B-spline functions have been applied, in order to determine the loads leading to the ultimate temperatures of 400, 500, 600, 700, 800 and 900C. Example: for IPEA80 section, buckling around the major axis, H = 1-272 m, f y = 2 3 5 MPa, Table 2 gives the ultimate temperatures found for the 16 different loads that have been applied. Table 13 gives the loads corresponding to the ultimate temperatures of 400, 500, 600, 700, 800 and 900C. These loads have been found from the values of Table 2 by application of linear B-spline functions and the least square method, in order to cope with redundant points. Figure 2 shows how well the original results from Table 2 and the interpolated results from Table 3 compare when plotted on the same graph.
TABLE2 OriginalPoints:UltimateTemperatures

Load (N)
84696 81167 '77638 '70580 63522 49406 139525 125409

UIt. temp (cC)

409.84 431.14 451.75 492.84 519.45 567.12 601.25 668.89

Load (N)
18351 15528 12704 9881 9175 7764 7058 5646

UIt. temp (C)

704.37 736.86 767.50 798.56 815'03 852'48 871.75 920'66

TABLE 3 Interpolated Points: Ultimate Loads

Temp. (C)
400 500 600 700 800 900

Ultimate load (N)

86476 69288 39794 18822 9747 5988

TABLE 4 Interpolated Matrix. Ultimate Loads in kN for IPE80A, Major Axis, $235

Temp. 1.272 1.908 2.544 3.180

Lenyth (m) 0 0"636

3 " 8 1 6 4 " 4 5 2 5.088

5.724

6"360

.=

t~

20 400 500 600 700 800 900

149-93 149"93 116"95 70"47 34"48 16-49 9-00

146"64 120"20 93"83 56-31 27-74 13-25 7"23

141"16 86"47 69"29 39"79 18-82 9-75 5"99

131"61 71'60 58'81 32'14 14"66 8'24 5"53

1 1 6 " 4 5 95'23 74'41 62"15 5 3 " 3 0 44'29 51-86 4 4 - 8 3 3 7 " 4 1 27-53 2 3 - 5 1 1 9 - 5 2 12'17 10-20 8"38 7"30 6'36 5'37 5"08 4"49 3"84

57-97 45"86 3 6 " 0 0 29'25 30-54 24"88 1 5 " 8 8 12'90 6"77 5"48 4'44 3"64 3"21 2'65

36'98 24-00 20"45 10-58 4'49 3"00 2"21

30'38 19"95 17'02 8"80 3"72 2"51 1-85

.r-

Fire resistanceof axiaUy-loadedmembers

57

For each section, for each buckling axis and for each nominal yield strength, a matrix was built, such as the one given in Table 4. Each horizontal line of Table 4 is the buckling curve for the corresponding teml~,~rature. These curves have a very regular and continuous shape. Therefore a sinusoidal expression is used when the ultimate load has to be found for a temperature belonging to the matrix (T= 600C for example), but for intermediate length (e.g. H = 2 m)

H N(T, H)=AI(T)+ A2(T) - - +

Bronx

[ircU'~ ~ BI(T) s i n / - - /
9 .

(5)

i= 1

\nmax]

where N(T, H ) = axial load corresponding to temperature T (belonging to the matrix) and height H; N(T, 0)= plastic load at elevated temperature T; Hm,x maximum height (corresponding to 2 = 200) for which calculations have been completed (Hm,=6"360m in Table 4); AI(T),A2(T),Bi(T) coefficient to be calculated from the 11 interpolated points at temperature T. Note: A x(T) = N(T, 0), A2(T) = N(T, nmax)- N(T, 0). When the ultimate load has to be found for a temperature and a length that do riot belong to the matrix (T=660C and H = 3 m for example), eqn 5 is applied twice for the temperatures of the matrix immediately superior and inferior to required temperature (600 and 700C in this case) and linear interpolation is used between the two results (N(600, 3) and N(700, 3)).
Results

In order 1:o allow comparisons with the analytical formula of Eurocode 3, Part 10, ~ the results have firstly been presented in the form of a buckling coefficient which is a function of the relative slenderness ~ evaluated at 20C. The buckling coefficient is defined as

N(T, H) N(T, H) N(T, H) Np,(T) fy(T)f~ Kfy(T)fyf2

N(T, H)
Np,(T) =

for the general method (6) for the simple method

xlx

where Nt,I(T)=N(T,O), plastic load at temperature T; f y( T) = yield strength at temperature T; f y = y i e l d strength at ambient temperature; f~=cross-.sectional area; Kfy=fy(T)/fy according to Ref. 1; ;(=buckling

58

J . - M . Franssen, J.-B. Schleich, L.-G. C a j o t

coefficient calculated at ambient temperature from the c-stut curve; ~=correction factor, equal to 1.20 in Eurocode 3, Part 10.1 Ten different values of ~, from 0.20 to 2.00, have been considered. For each relative slenderness, the buckling coefficient has been evaluated for the 339 sections, for two buckling axes and for five ultimate temperatures (400, 500 ... 800C). The temperature of 900C has not been considered because it leads to higher buckling coefficients than the other temperatures and gives the false impression of a high scatter between the numerical results when presented in Fig. 3. This temperature of 900C in fact corresponds to very low load levels and is not very commonly reached in practical situations. As the results that are not considered give higher buckling coefficients than those that are used in the comparisons, the simplification is on the safe side anyway. A total of 3390 buckling coefficients has thus been calculated for each relative slenderness. In Fig. 3, SAFIR max is the curve linking the maximum values that have been calculated by the general model, SAFIR min links the minimum values and SAFIR 50% means that 50% of the numerical results are below this curve. The analytical proposal made by the project team of Eurocode 3, Part 101 is also drawn on Fig. 3 (it corresponds in fact to the buckling curve c of Eurocode 3, Part 16 divided by the correction factor of 1"20). It can be observed that the analytical curve is too conservative for short columns (~<0.20), whereas for intermediate length columns (0.30<~< 1.20) the simplified model systematically provides results that are not as safe as the general model. Also for very slender columns
1
"~. i i

! 0 . 9 . . . . . . ~,"............................... \ .: 0.7 ,., ,,

...........................

~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .......... . . . . . . . . . SAFIR. . .m.a .x. . ,I . ........ . . i

......................................... ~ ...................................

~x-x-:-.-~

"~,x "~ ~ ~ v.o - - - ~ - - - - - - - ' - , ~ - ' ~

o s

- - - - ~ ' ~ . - - - " ~ : ' . ~

! . . . . .................i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..........

S A F I R min E c 3 P rt l o ]

[ i i

0.4 ~ ~

~
, ~

............................... ~ ..............................................
........................................ l

0,3-~~

0.1 ~-............................................................................................ L i............................................. i...................... ~-I

.......

ol 0 0.5

~ RELATIVE 8LENDERNE88 A T 20C.

i 1.5

I 2

Fig. 3. $235, u n i f o r m t e m p e r a t u r e , 339 sections, t w o axes, five u l t i m a t e t e m p e r a t u r e s .

Fire resistance of axially-loaded members

59

(1.20 < ;[) the result provided by the simple model is unsafe when compared to the general model for more than 50% of the cases. The s]aape of the distribution of numerical buckling curves is too different from the shape of the analytical curve to allow any hope that a modification of the correction factor used in the simple model could lead to a better correlation. In ordLer to allow comparisons with the analytical formula of Eurocode 3, Part 1.2,2 the results have been presented in the form of a buckling coefficient as a function of the relative slenderness evaluated at the ultimate temperature:

,~

_;L/

/-E~_-~

;~(~)=,~(T)-/~4I-TC~)-~4

K-~

(7)

where KE=E(T)/E, according to Ref. 1. Figure 4 presents the curves SAFIR max, SAFIR min and SAFIR 50%, as well as the analytical proposal of Eurocode 3, Part 1.22 (it also corresponds to buckling curve c of Eurocode 3, Part 16 divided by the correction factor of 1.20, but the variable on the horizontal axis of the diagram is evaluated at the ultimate temperature). The analytical curve is also conservative for short columns (~<0"20), whereas lit systematically provides results that are not as safe as the general model for 0-40 < ~ < 1.00. For very slender columns the result provided by the simple model tends to be safer when compared with the general model. The buckling curve's distribution is thinner in Fig. 4 than it was in Fig. 3.
1 ~ - S ~ -. . . . .
09

i
SAFIR
max

J,-.~.~'--',,.__ ......................... i ............................................... , .....................................

~.~'x",,
0.8 ] : ' ~

i
~

i
SAFIR 50%

.................................................. i ..........................................

o., I" ........................ ~

0.6

..................................................................... - . . . . i
........ i .....................................

~,,F,R rn,o
New proposal

+ - . . ....................................................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

o . s 1- ........................................ ~

.................................... !

o.4

+ ........................................... ~
l

_
:.

......
~

0.3 I.

o.~. +
"
040

............................................ ~! .............................................. :i.............................................. r.....

~ 1 I 1.5

--

......

:.I
i 2

I ........................................................................................... i i ......................................................................................... i 1
0.5

RELATIVE SLENDERNESS AT FAILURE TEMPERATURE

Fig. 4. $235, u n i f o r m t e m p e r a t u r e , 339 sections, two axes, five u l t i m a t e t e m p e r a t u r e s .

60

J.-M. Franssen, J.-B. Schleich, L.-G. Cajot

This means that there is a better possibility of representing the results of the general model, without excessive safety, with an analytical expression when the relative slenderness is evaluated at elevated temperature. The nominal yield strength of 355 MPa has also been considered. The nominal value has been adapted depending on the thickness of the flange according to EN 10025, 6 i.e.

fy = 355 M P a
fy=345MPa

fy=335MPa fy=325MPa
fy=315MPa

fy=295MPa

if if if if if if

16 40 63 80 100

tf<<.16 mm <tf~<40mm <tf<<.63mm <tf<<.80mm <tf~<100mm <tf

The sections which are classified under class 4 according to Eurocode 3, Part 17 have not been considered, which leaves 258 sections from the catalogue s for consideration: 34 106 4 23 25 6 32 28 HD HE HL HP IPE UB UC W sections sections sections sections sections sections sections sections

Figure 5 presents the results when the relative slenderness is evaluated at the ultimate temperature. As two buckling axes and five ultimate temperatures are considered, 2580 buckling coefficients have been calculated for each relative slenderness. The numerical curves are higher on the vertical axis for $355 than they were for $235. This is due to the fact that the residual stresses do not depend on the yield strength. Their relative influence is therefore smaller when the yield strength is increased. This phenomena is not accounted for in the simplified model of Ref. 2, where the buckling coefficient does not vary with the yield strength.

4 THE NEW PROPOSAL The simplified model proposed in Eurocode 3, Part 10 or Part 1.2 can produce safety levels that are significantly lower than the general method.

Fire resistance of axially-loaded members

61

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0

............................................... i [ ............................................... i ! r ............................................... i ! ~............................................... i I I I

0.5

1.5

RELATIVE SLENDERNESS AT FAILURE TEMPERATURE

Fig. 5. $355, uniform temperature, 258 sections, 2 axes, 5 ultimate temperatures.

It appears that safety levels vary with different slenderness for the simple method. 'Therefore, a simple modification of the proposed method, such as another value for the correction factor, would require a higher value to ensure safety or intermediate length columns (Fig. 4) and would lead to an excessively uneconomic situation for short and slender columns. A new proposal for the simple model is given below. The equation is

N(T, H)= %(T)Kfy(T)fyfl

where x(T) =
1

(8)

p(T) + x/tp2(T)- ~2(T)

(9) (10)

q~(T) = (1 + ~ ( T ) + ~2(T)) with

= fie 1Lheimperfection factor; fl = the severity factor, to be chosen in order to ensure the appropriate safety level; e=x/~3-~f-}fy in MPa is the nominal yield strength, i.e. not reduced according to EN 10025, 7 depending on the flange thickness.

(11)

Equations 9 and 10 are in fact exactly the same as those defined at ambient

62

J.-M. Franssen, J.-B. Schleich, L.-G. Cajot

temperature in Eurocode 3, Part 1, 7 except that the threshold limit of 0.20 for ~[ does not appear in eqn 10. The fact that the threshold limit does not appear changes the shape of the buckling curve. It differs from that at ambient temperature. The new curve starts at ~ = 1.00 for ~ =0.00 but it decreases even for very low slenderness, instead of having a horizontal plateau up to 2 = 0.20 (Fig. 4 and Fig. 5, where fl = 1.20). The buckling curve varies with the yield strength due to the parameter e that appears in the imperfection factor. For the situations where the temperature is known, eqns 8-11 allow direct calculation of the ultimate load. If the ultimate temperature is the unknown, an iterative process has to be applied, owing to the fact that the relative slenderness is evaluated at the ultimate temperature. This process is facilitated with computers. It converges very quickly, especially if the initial value of the slenderness is chosen according to ~(T) = 1-20 x ~.(20C). The determination of the ultimate load is even faster with the help of tabulated data. Equation 8 can be rewritten in the form

N(T, H)= ~(T, H)Npt

where U?(T,H)=z(T,H)Kfy(T) is the N~,t= f y ~ is the plastic load at 20C. overall buckling

(12) coefficient;

Tables such as Tables 5 and 6 have been established (with fl= 1.20) and can be directly used. The effect of the reduction of the material properties, as well as the effect of buckling, are incorporated simultaneously. The ultimate load is directly related to the plastic load at 20C, according to eqn 12.
Notes

1 Low temperatures F o r m u l a 8 should probably be adapted for the temperature range, 20-400C.
F r o m 20 to 100C for example, the stress-strain relationship of steel is not modified (Kfy = Kfp = KE = 1.00) and there is no reason why the failure load should decrease. F r o m 100 to 400C, a linear interpolation could be proposed between the load calculated at 100C ( = t o the load at 20C,

Fire resistance of axially-loaded members

TABLE 5 Overall Buckling Coefficient ~b for $235 Steel

63

Relative slenderness at 20C 0.03 0-10 0.20 0.30 0.40 0.50 0-60 0"79 0"80 0"90 I'OE) 1'19 1'29 1"39 1"49 1"59 1"69 1"79 1"8'9 1"90 2"0C)

Ultimate temperature 400C 1.000 0.873 0.769 0.680 0.601 0.531 0.469 0"414 0"367 0'326 0'290 0"259 0"233 0"210 0"189 0"172 0"157 0"143 0'132 0'121 0'112 500C 0.780 0.685 0.607 0.539 0.479 0-426 0.378 0"336 0"299 0"267 0"238 0"214 0"193 0"174 0'158 0"143 0"131 0"120 0"110 0"102 0"094 600C 0.470 0.409 0.359 0.316 0.278 0.245 0.215 0-190 0"167 0"148 0"132 0"117 0"105 0'095 0'085 0"077 0"071 0"064 0"059 0"054 0"050 700C 0.230 0.198 0.172 0.150 o. 131 0.114 0.099 0"087 0"076 0"067 0"059 0"052 0"046 0'042 0'037 0"034 0"031 0"028 0"026 0"024 0"022 800C o. 110 0.097 0.086 0.077 0.069 0.061 0.055 0'049 0'043 0'039 0'035 0"031 0"028 0"026 0"023 0"021 0"019 0"018 0"016 0"015 0"014 900C 0.060 0.054 0.049 0.044 0.040 0.036 0-033 0'030 0"027 0'025 0'022 0'020 0'019 0'017 0'015 0"014 0"013 0"012 0"011 0"010 0"010

calculated a c c o r d i n g to E u r o c o d e 3, P a r t 17) and the l o a d calculated at 400C b y the simple m o d e l for elevated t e m p e r a t u r e . This is true w h a t e v e r the analytical p r o p o s a l at elevated t e m p e r a t u r e . 2 Linear interpolations Strictly speaking, the i n t e r p o l a t i o n o n the t e m p e r a t u r e in T a b l e 5 or 6 s h o u l d not be linear because of the t e m p e r a t u r e ' s d e p e n d e n c y o n the relative slenderness. In fact, K f y has an o v e r w h e l m i n g influence in the d e t e r m i n a t i o n of the overall buckling coefficient, and a linear i n t e r p o l a t i o n is a m p l y sufficient. Example: F o r ~ = 0 . 4 0 , $355, T = 6 5 0 C . L i n e a r interpolation. ~0(600, 0.40) = 0-300 ~0(700, 0.40) = 0.141 ~k(650, 0.40) = (0.300 + 0.141)/2 = 0.2205 (Table 6) (Table 6) (eqn 11)

64

J.-M. Franssen, J.-B. Schleich, L.-G. Cajot

TABLE 6 Overall Buckling Coefficient ~k for $355 Steel

Relative slenderness at 20C

0-00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80 1.90 2.00

Ultimate temperature
400C 500C 600C

700C
0.230 0-203 0.180 0.160 0.141 0.124 0.109 0-095 0-083 0-073 0-064 0-057 0.051 0.045 0.041 0-037 0-033 0-030 0.027 0.025 0.023

800C
0.110 0.099 0.090 0.081 0.073 0.066 0.059 0-053 0-048 0-043 0-038 0.034 0.031 0.028 0-025 0-023 0.021 0-019 0-018 0-016 0.015

900C
0.060 0-055 0-050 0.046 0.043 0.039 0.036 0.033 0.030 0.027 0.025 0-022 0-020 0.019 0.017 0.016 0-014 0.013 0.012 0-011 0-010

1-000 0.894 0.803 0.721 0.645 0.576 0.512 0.454 0.403 0.357 0.318 0.284 0.254 0.228 0-206 0-186 0-169 0.154 0-141 0.130 0.119

0.780 0.701 0.633 0.571 0.514 0.462 0.413 0.369 0.329 0.293 0-262 0.235 0.211 0.190 o. 172 0.156 0-142 0-129 0-119 0.109 0.101

0-470 0-419 0.375 0.336 0.300 0.266 0.236 0.208 0.184 0-163 0.145 0.129 0.115 0.103 0-093 0-084 0-076 0.069 0-063 0-058 0.054

Exact value.

= ( 2 3 5 / 3 5 5 ) o.5 = 0"814 = 1"20 x 0 " 8 1 4 = 0 " 9 7 6 K E ( 6 5 0 C ) = (0"31 + 0" 13)/2 = 0"220 g f y ( 6 5 0 C ) = (0"47 + 0"23)/2 = 0"3 50 ~(650C) = (0"35/0"22) s ~ = 1"261 x 0"40 = 0"505 ~o = (1 + 0 " 9 7 6 " 0"505 + 0"505 * 0"505)/2 = 0"874 1 Z = 0.874 + ~ / 0 - 8 7 4 2 - - 0.5052 =0"630 (Ref. [ 1 ] ) (Ref. [ 1 ] ) ( e q n 10) ( e q n 9) ( e q n 11)

~0(650, 0-40) = 0.630 0"350 = 0-2206

5 UNPROTECTED ISO

SECTIONS HEATING

SUBMITTED

TO

The case of unprotected profiles submitted to ISO heating leads to the m o s t s e v e r e t h e r m a l g r a d i e n t s in t h e steel s e c t i o n , e x c e p t if a m o r e s e v e r e

Fire resistance of axially-loaded members

65

heating curve is applied (hydrocarbon curve for example). All the other cases of thermally-protected profiles should have results between those calculated in Section 3 (totally-protected and uniform temperature) and those that could be calculated for unprotected profiles. The general model has been used to calculate the temperature distribution in the steel sections heated by the ISO curve. The boundary conditions have been taken according to Eurocode Actions, Chapter 209 and the t!hermal properties of steel according to Eurocode 3, Part 1.2.2 For $235 steel, for all the sections, for both buckling axes and for 10 different column i[engths, eight different loads have been applied and the fire resistance noted. The average temperature for the non-uniform distribution at the moment of failure, Tm~a,, has also been noted. For the 50 000 odd non-uniform results calculated by the general model, the load leading to the same,, ultimate temperature in the uniform situation has been calculated, as well as the ratio between both values Nis(Zmean,H) and Nunierm(Tmean, Values higher than 1-00 indicate that the ISO heating is H). not as severe as the uniform heating, whereas values lower than 1.00 indicate that the ISO heating leads to a lower ultimate temperature. Figure 6, representing the distribution of the NiS/Nunifrm ratio, shows that the difference between ISO and uniform temperature seldom exceeds 15%. Figure 7, representing the cumulative distribution, shows that in 95% of cases, the ISO load is equal to at least 90% of the U N I F O R M load. In fact, the question is whether the new analytical proposal made in Section 4 for totally protected sections and uniform temperatures can be

8236

0.14 ....T....1.....T.....[.....T....T.....[.....T....T~...T...~.[.....T....T.....r...T....1.....T.....V...T.....i 0.12 ....i.....i............ [.....i.....i ........... I.....L...J......L....i.....L....L...ii__Li._

I : i J i
.

~ ~ 0.1 ~i ~i ~ ~ z ~ ~ ~ :......... , , i~ ~ E ~ ~oi 0.~ .. ,.....;......~......t..........~......,.....=....~~ :~t~;-;---*~i-~*"~ ;z ;~ ~ ~ ~ ~

.... .....~..,-..t......~.-...~-.---~-...--r.--.-~-.,-.~ . . . . . r.-.-.~-.-.-~ ........... ~-----~------:-.-.-~-----~.----:

~00.08
. . . .

- - - - - ~ - - - - - - ~ - - - - - - h - - - - f - - - - - ~ - - - - - - h . -~- + . - , - . ~ i ~ i .

.......

r--.-.+ ...........

h----+-----~ ............

~----- .....

g
0.02

.... ,...-.;.....i......~.-.-.z...-.~.....~....,.,~ ~ ~ ~ ~ ~

~....-~....--;--..s....;i i ~

~ ; * ~ .....

o
0 0 0 0 0 ,rT,r~~N(iso)/N(uniform)

Fig. 6. Distribution of the ISO/UNIFORM ratio. N(iso) and N(uniform)calculatedby the general model.

66

J.-M. Franssen, J.-B. Schleich, L.-G. Cajot

S236

o.a

._a_ . . . .
~

-___-__~__l_a,_~_~___~_.
~ ~

0.8

07 -~
O.~

_'_ . - - ; - . . . .
;

....

h---'~i-;

- - , , - , - , - , - , , , - , - , - , - , - - , - - - , - i / ; ~ ; Jr ; / ; ] ;

I-~-.-/+

....

+--~-

-~

-~

o.~

T,

l--,q

,-1

, I,

1-

--;---~--

I- t -I-- ~- - F - P -,"- -I - - I - -I - ~ -/-I- - + - -t- - f --k---:-----I-o13 _ - _ t- _ -, - F _ L, . _ _ -I- _ , -I __1, I L L - , - - _ + _ . L, - t . _ L . . . . ,. . . . ,,o1" , '

0 I-~-f-~--I--k-I
0.5 0.6 0.7

"I-- -

-~---'--~ ....
0.8 0.9

*' . . . . L 1 1.1 1.2

--~----!---~,-' ' 1.3 1.4 1.5

N(iso)/N(unifonn)
Fig. 7. C u m u l a t i v e distribution of the ISO/UNIFORM by the general model. ratio. N(iso) and N(uniform)

calculated

s a f e l y a p p l i e d t o u n p r o t e c t e d p r o f i l e s s u b m i t t e d t o t h e I S O fire, a n d h e n c e t o n o r m a l l y p r o t e c t e d p r o f i l e s a s well. T h u s , t h e m e a n i n g f u l r a t i o is b e t w e e n calculated by the general model and Nanalyt'(Tmean, c a l c u l a t e d b y t h e n e w p r o p o s a l ( e q n 8). F i g u r e 8, r e p r e H) s e n t i n g t h e d i s t r i b u t i o n o f t h e NiS/N anazyt" r a t i o , s h o w s t h a t t h e I S O

NiS(Tm~a,,H)

b u c k l i n g l o a d is g e n e r a l l y h i g h e r t h a n t h a t c a l c u l a t e d b y e q n 8. F i g u r e 9, r e p r e s e n t i n g t h e c u m u l a t i v e d i s t r i b u t i o n , s h o w s t h a t in 9 5 %

$236
0.04 T-i ....... F ) ....... F J ....... i....... i ....... i....... i - i ....... F - I ....... i....... J i ....... i i ....... i i ....... i

ell
i~

te

0.035
0.03

..... ".' ....... ~ ' " ' " ' : " ' " ' - ' " ' " i ' " ' " + ' "

+ ' " " ' " - - ~ ' " - ~ ....... : . ' " ' " ' f ' " " - ' " ' " ' " ' " ' - ' " ' " ~ ' " ' " " ....... " ' " ' " ....... i

0.025

0.02
0.015

I|

..... ~.......~'"'""'"""""~'"'"* . . . . . i J i J i i ..... ~....... ~'-----~-'--'------+-----

~'"'"~ ~'""'*""+'""~'"'"+'""'~'"'"~ ""'"":....... ....... .......

i J i J i i i i i i i i '.-----~-....------':-----------------------+-----~ ....... i------i .......
. . . . . . . . . . . . . .

..... I....... i
..... ~..............~'"'"'"'"" .

. .

. .

. .

. . .

\~

"'"'-~"""-"""~"'"'~ """'~....... .......

'-----------.:--.--.~ ....... ~------~....... !

0.01

..... ~....... ~...-.-i.-..-.+....-.~ .....

0.005

.....~ ~"'"'~.......~"'~ . .......

~""~.......~'""'~ .......

N ( i s o ) I N ( n e w proposal)
F i g . 8. D i s t r i b u t i o n of the ISO/new proposal ratio.

Fire resistance of axially-loaded members

67

8236

o.o

_+ . . . . _+ . . . .

.... _~. . . .
1
I

_. i_.

__. _s_-

-4--~-I

4. . . . - t ~ , ~
';--,"5, ....
"r

-~I

i I

- - ~ -I -

~ 1
1

0.7
S . o0.5 .o
0.4 0.3 0.2 0.1
o

-i---r--!
-T .... -+ .... _~ . . . . _~ -i
-~ ....

'l
I

!
'

I -4--

-~__

-~

' ....

I
! __/___4

' ....
I

' I
I

T----T----J'-~ +-t --t~ .... 4-~-/ + ~ i i T T i I

~--~ - - ' ~ -'-"

.....
---..1.--/

--4
I

....
.L . . . . I

d ....
I -- -q.---

d-I
--1--I

1
- 4 - I
--.,-1---I

-d
-I I
--..J---I

I -- -t- -- -

I
-t. . . . . I

;;I~"
1.1

_..I I I

J. I I ~----t------

1 I I -'f . . . .

1 I I -t' .

I I I .

J I I .

--7t - - "

-. 7 .

_J--~--1 .

0.7

0.8

0.9

1.2

1.3

1.4

1.5

1.6

1.7

N(iso)lN(new proposal)

Fig. 9. Cumulative distribution of the ISO/new proposal ratio. of the cases, the ISO load is higher than the load calculated according to eqn 8. That means that the new proposal can be safely used for unprotected steel profiles, although it has been established from results based on the uniform temperature distribution.

6 THE SEVERITY FACTOR The value given to the severity factor directly influences the safety level: ultimate ]Loads decrease with an increase in severity factor. The value of 1.20 chosen for fl in Figs 4, 5, 8 and 9 and for Tables 5 and 6 ensures that the simple model is always safer than the general model. This is in fact rather conservative, especially at high slenderness, because the ultimate load calculated by the general model can in some cases be 50% higher than the ultimate load calculated by the simple model (compare SAFIRmax and new proposal on Fig. 4). It has been proposed in this paper to use one single buckling curve for every yield strength in case of fire. Could it be possible to differentiate between different buckling curves in order to reduce this safety margin that can sometimes be excessive? A more detailed analysis of the numerical results shows that the classification established at ambient temperature between different buckling curves a, b, c and d tends to vanish at elevated temperature. The fact of using a different severity factor, and therefore a different buckling curve, according to the same classification as at 20C, would therefore not reduce the signilficant safety level that is achieved in some cases. Two other

68

J.-M. Franssen, J.-B. Schleich, L.-G. Cajot

options have been considered without better success: either considering a different buckling curve for each ultimate temperature, or differentiating between both buckling axes. It should also be mentioned that the numerical calculations have been completed with structural imperfections, residual stresses and geometrical imperfections, having characteristic values, 0.30 or 0.50 x 235 M P a for the residual stresses and H/IO00 for the imperfection, and that the observed values in well instrumented tests are usually not as severe. An analysis is now under way of different experimental results. The database now comprises up to 53 tests, including 14 tests specially completed in Spain by LABEIN within this research project. This analysis should allow determination of the severity factor suitable for use in practical design. For the reasons explained in the above lines, the general impression is that the severity factor could be decreased and that higher buckling loads could be considered. If confirmed, this fact should lead to the conclusion that structural imperfections which are not as severe should also be considered when applying the general model. Indeed, if the general approach of safety in case of fire is to consider average, rather than characteristic, imperfections, in order to better represent the results of experimental tests, to remain consistent this approach should be applied in both calculation models.

7 CONCLUSIONS The physical fact that the Young's modulus decreases faster than the yield strength when the temperature increases, plus the fact that the stressstrain relationship at elevated temperature is not the same as at ambient temperature, produces a modification of the buckling curve at elevated temperature. The horizontal plateau valid at 20C up to a relative slenderness of 0-20 vanishes in the case of elevated temperatures. The simple models based on the buckling curve that is valid at ambient temperature lead to a safety level that depends on the slenderness of the column, the results being unsafe for intermediate length columns. It has been possible to make a new proposal of a buckling curve for hot-rolled H-sections submitted to fire, in the hypotheses of Eurocode 3, Part 1.2.z The buckling coefficient is based on the relative slenderness evaluated at the ultimate temperature and depends on the steel grade. An analytical formula is given for ultimate temperatures beyond 400C. This formula can be used with the aid of a pocket calculator. Simple tables can be established which give the overall coefficient, simultaneously taking into account the effect of buckling and the decrease of the steel strength. The severity factor of the proposed simple calculation model has been

Fire resistance of axially-loaded members

69

established, in order to give results that are never higher than those provided by the use of the general calculation model in the case of uniform temperature distribution. It has been verified that the analytical proposal is safer than the general model in 95% of the cases for unprotected sections submitted to the ISO heating. Further analysis of the numerical results and of experimental tests is under way to determine the severity factor giving a safety level that is similar tc, that which is obtained in experimental tests.

ACKNOWLEDGEMENTS This work was sponsored by the European Convention for Steel and Coal. 1 This research project involved Mr L. Twilt from T N O in The Netherlands and Dr J. Kruppa from CTICM in France, who are investigating the interaction between axial loads and bending moments, and Prof. Ramirez from LABEIN in Spain, who completed the original experimental tests. We acknowledge their contribution to the discussions on the subject reported in this paper.

REFERENCES 1. Kruppa, J., Law, M. & Twilt, L., Eurocode No. 3. Design of Steel Structures. Part i!O: Structural Fire Design. Commission of the European Communities, Draft, April 1990. 2. prENV 1993-1-2. Eurocode No. 3. Design of Steel Structures. Part 1.2. Structural Fire Design. CEN, in press. 3. Franssen, J. M. et al., A comparison between five structural fire codes applied to steel elements. Proc. of the 4th Int. Symp. of Fire Safety Science, Ottawa, Int. Ass. for Fire Safety Science, June 1994. 4. Franssen, J. M., Mod61isation et influence des contraintes r6siduelles dans les profils m6talliques soumis ~ l'incendie. Construction Mdtallique, 3, (1989) 35-42. 5. Franssen, J. M., The unloading of building materials submitted to fire. Fire Safety J., 16, (1990) 213-227. 6. EN 10025, Hot Rolled Products of Non-alloy Structural Steels: Technical Delivery Conditions. CEN, 1990. 7. ENV 1993-1-1. Eurocode 3: Design of Steel Structures. Part l.l: General Rules and Rules for Buildings. CEN, April 1992. 8. ARBED. Sale Programme. Structural Shapes. ARBED, Luxembourg, January 1993.
9. Eurocode on Actions on Structures. Chapter 20: Actions on Structures Exposed to Fire. Commission of the European Communities, Draft, April 1990.

10. C. E. C. Agreement 7210-SA/515/931/316/618, Buckling curves of hot-rolled H-steel sections submitted to fire.