Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
ABSTRACT: This paper presents a general numerical method to calculate the nonlinear
behavior of load-bearing members under elevated temperature conditions. The method,
although similar in concept to earlier approaches, differs with respect to both solution
Downloaded from ascelibrary.org by National Institute of Technology Karnataka on 01/03/19. Copyright ASCE. For personal use only; all rights reserved.
procedure and capability. The governing equations are formulated in a succinct format
and an innovative iteration procedure is used to obtain the solution. The method can
take into account the combined actions of axial force and biaxial bending, external
restraints, temperature variation over the cross section and along the member, material
nonlinearity, geometric nonlinearity, unloading and reloading, residual or initial stresses,
and initial "out-of-straightness" of the member. It applies to members of any cross-
sectional shape and with any elevated temperature stress-strain relationship for the ma-
terial, including one with zero or negative slope. The application of the method is illus-
trated using a practical design example in which the fire resistance of an unprotected
steel column of large cross-sectional area was evaluated.
INTRODUCTION
The analysis of the behavior of load-bearing members in buildings during a fire can be
complicated. Various factors that influence the behavior of the members need to be taken into
account, including:
Because of the nonlinear nature of the problem, closed-form solutions usualIy cannot be found
and an iterative approach is required.
The nonlinear behavior of a member under elevated temperature conditions can, in principle,
be modeled using the finite-element method (Anderberg 1976; Okabe et al. 1991; Schleich
1987). Alternatively, as in the case of the methods described by Hass (1986), Jeyarupalingam
and Virdi (1991), Lie and Chabot (1993), and O'Meagher et al. (1991), moment-thrust-curvature
relationships for the member cross section can be used in conjunction with member equilibrium
and compatibility equations. Although not as general as the finite-element method, this latter
approach can, because of its inherent simplicity, alIow the nonlinear behavior of members to
be more closely modeled.
This paper presents a numerical procedure to determine the behavior of members at elevated
temperatures. It is similar, in concept, to the aforementioned moment-thrust-curvature ap-
proaches but differs with respect to solution procedure and capability.
The method is applicable to members of any cross-sectional shape and constructed of any
material as long as the elevated temperature stress-strain relationships of the material are known.
All the aforementioned factors influencing the behavior of a member at elevated temperatures
can be taken into account. The governing equations are formulated in a succinct format suitable
for incorporation into a computer program.
'Res. Engr., BHP Res.-Melbourne Laboratories. 245-273 Wellington Road. Mulgrave, Victoria. 3170, Aus-
tralia.
ORes. Assoc., BHP Res.-Melbourne Laboratories, 245-273 Wellington Road, Mulgrave, Victoria, 3170,
Australia.
Note. Associate Editor: Louis Geschwindner. Discussion open until September I, 1995. Separate discussions
should be submitted for the individual papers in this symposium. To extend the closing date one month, a written
request must be filed with the ASCE Manager of Journals. The manuscript for this paper was submitted for
review and possible publication on April 20, 1994. This paper is part of the Journal of Structural Engineering,
Vol. 121, No.4. April. 1995. ©ASCE, ISSN 0733-9445/95/0004-0664-0675/$2.00 + $.25 per page. Paper No.
8288.
Geometric Discretization
Downloaded from ascelibrary.org by National Institute of Technology Karnataka on 01/03/19. Copyright ASCE. For personal use only; all rights reserved.
For the purpose of analysis, the member is divided into a series of segments of appropriate
lengths. Each segment is represented at its midlength by a cross section. The cross section, in
turn, is divided into a number of small subareas of appropriate shapes and sizes (Fig. 1).
Each subarea is referred to as a level 1 element, and the cross section and member are levels
2 and 3 elements, respectively.
Temperature variations over the cross section and along the member are incorporated through
the combination of the subarea temperatures, which are assumed uniform over each subarea.
Analytical Formulation
Having defined the actions and deformations, succinct equations are formulated to relate the
actions and deformations of the three element levels. These actions and deformations are related
by material behavior, boundary conditions, compatibility, and requirements for static equilib-
rium (Fig. 2). To determine the structural behavior of the member, the solution sought is a set
of actions and deformations for the three levels such that all the requirements are satisfied.
Solution Approach
A simple iterative approach is adopted to obtain the solution, in which the actions and
deformations of the three levels are calculated successively in loops till the solution converges.
The procedure comprises two iterative loops (Fig. 2): the behavior of the cross section is
determined in the first loop (cross-section analysis) and the behavior of the member is determined
in the second loop (member analysis). While loop I iterates on material nonlinearity, loop 2
iterates on geometric nonlinearity.
Material Nonlinearity
One feature of the method is the treatment of material nonlinearity in the cross-section
analysis. To facilitate obtaining the solution, the subarea action is artificially separated into a
linear component (ignoring material nonlinearity) and a correction component to account for
material nonlinearity. By doing so, the behavior of the cross section can be easily determined
unIIarm
F?!
I I
I! ! ••
I
..:::=,.~.
tram ,, from
-i<t>I I . I
,,_~' •
tIle","",bot
!I i ~~I~~
Iii
W CNW_
'"
--
'/
ILAgend:
A~
DK
ocIion
deformatlon
ANALYTICAL FORMULATION
Mathematical relationships for the analysis are formulated such that the actions and defor-
mations for the three element levels can be readily calculated in continuous succession, with
each relationship using the results of the previous relationship.
In formulating the relationships, a Cartesian coordinate system is adopted. The coordinates
are oriented such that the x-axis and y-axis lie parallel to the plane of the cross section and the
z-axis lies in the longitudinal (axial) direction of the member. For convenience, the origin of
Downloaded from ascelibrary.org by National Institute of Technology Karnataka on 01/03/19. Copyright ASCE. For personal use only; all rights reserved.
_n
u-+----+-=_+.....z..
mem.....
ActIon
----------1I-~-~>=--_+_--~SlIlIln
,,
,,,
,,
,"
,t>'
.. .'
'
LIDI..a
FIG. 3. Analytical Formulation FIG. 4. Material Stress-Strain Relationship
f, (2a)
where U i = axial stress.
The axial stress u" in turn, is calculated from its stress-related strain and the relevant stress-
strain relationship. The stress-strain relationship may be different for each subarea, depending
on the material, temperature, and load path. For the purpose of general formulation, the stress-
strain relationship (Fig. 4) is expressed using arbitrary functions (to be specified by users) as
where g( E i', T,) = stress-strain function during loading; f j' = stress-related strain; Ti = tem-
perature; En = initial slope of the stress-strain curve during loading; f;" = irrecoverable plastic
strain; and h( f i', T i ) = stress-strain function during reversed loading, taking into account the
Bauschinger effect when appropriate.
The stress-related strain f i' is calculated from f i , by excluding all strain components that are
not directly related to stress, but including the initial strain, i.e.
(2(')
where f i' = strain components of fi that are not directly related to stress (which may include
free thermal strain f i", free creep strain fiT. free shrinkage strain f i", and so on depending on
the model and material used); and fi' = initial strain (i.e., strain prior to the application of
external actions or heating, e.g., residual effects associated with welding or rolling).
The influence of elevated temperature metallurgical processes such as the normalization on
fj' is ignored. In addition, as illustrated in Fig. 5, f;" is assumed to be unaffected by changes
in temperature (Franssen 1990).
Any stress-strain function may be specified, including one with a zero or negative slope. Since
stress is determined from strain, the uncertainty and ambiguity that often rises when determining
strain from stress are eliminated.
For the purpose of later formulation of the cross-section action-deformation relationship, the
subarea action is artificially separated into a linear component and a correction component. The
correction term is calculated as
"
curvature ¥" shape
stress-straln curve
at time step 1,
temperature T1
L
p-. . ._ _ stress-strain curve
at time step 2,
temperature T2
FIG. 5. Assuming EPI Remains Constant with Temperature FIG. 6. Calculating Member Deformation from Cross-Section
Change Curvature
i= I i= I i= 1
The effects of initial strains (or stresses) have been included in (3). This does not affect the
equilibrium of the cross section since the initial stresses of the cross section are in the state of
static equilibrium.
L"
;= I
s, L
i= I
"
S'Xi L
i= I
"
SiYi L
"
i=' I
(E jJ - E;')5;
"
Lfi
i= I
[::J L"
i= 1
SiX, L"
i= 1
s,X~
"
2: SiXiY;
i= I [;:J<P,
+ L
"
i= 1
(E jJ - E j')s"x i +
"
Lf:x,
i= I
(4a)
or simply
{A} = [S]{D} + {B} + {C} (4b)
where {A} = cross-section action; [S] = elastic stiffness matrix for the cross section; {D} =
cross-section deformation; {B} = cross-section action due to £j' and £;'; and {C} = cross-section
action due to fe, correction terms (material nonlinearity).
When analyzing the behavior of a member, {A} is usually treated as known and {D} as
unknown. In this case, it is more convenient to invert [S] and express {D} in terms of {A}, i.e.
{D} = [S]-'({A} - {B} - {C}) (4c)
Both [S] and {B} can be calculated once the geometry, material properties, and temperature
of the subareas are known. However, {C} is dependent on {D}. Thus, when {D} is unknown,
an iterative approach must be adopted by which {D} and {C} are alternatively calculated in
successive iterations.
Using the preceding approach, the behavior of the cross section under combined actions of
axial force and biaxial bending can be readily determined. Further, this formulation allows the
determination of any combination of three unknowns (cross-section actions or deformations),
as illustrated in Appendix 1.
(5a,b)
(5c,d)
(5e)
Downloaded from ascelibrary.org by National Institute of Technology Karnataka on 01/03/19. Copyright ASCE. For personal use only; all rights reserved.
W:
where &':b &~k' k , 6~k = nodal displacements in the x, y directions and nodal rotations about
the x-axis an'd y-axis, 'respectively, due to the member's initial out-of-straightness; and C~i and
CZi = geometric terms
Z,l;; when j < k
q,= { -(L - z;)!;; when j 2:: k
(5f)
(6d,e)
where P zb P"k' P'k' M"b M Xk = constant applied actions; and Kl'z' KI"" KI'I' K M ,., K Mx =
arbitrary restraint stiffness functions.
Pzi = 2:
J.. -,--j + I
P'k (7a)
-, /-
3OOx300mm
•
J.u.. ;:~:
FIG. 7.
rnemt>er
IICIIonS
-- SERVICE CORE
- •
tions FIG. 8. Dry-Wall Service Core with Unprotected Steel Columns
(7c)
{= ~L. ~j)&tk.' .-
Downloaded from ascelibrary.org by National Institute of Technology Karnataka on 01/03/19. Copyright ASCE. For personal use only; all rights reserved.
and Bxj , Byj = deflection at the midlength of the segment in the x and y directions, respectively.
Both Cf~~y and Cf~8x change as the member deflection varies, but Cft and CM remain constant
throughout the calculation of member behavior.
Convergence Criteria
Solution convergence is checked separately for both cross section and member analyses.
For the cross-section analysis, cross-section actions calculated from subarea actions [(3)] are
compared with those calculated from member actions [(7)]. The solution is considered to have
converged when the differences between the actions calculated using the two relationships are
within an acceptable limit.
For the member analysis, the lateral displacements calculated in one iteration are checked
against those calculated in the previous iteration. The solution is considered to have converged
when the difference in the displacements calculated in both iterations are within an acceptable
limit.
APPLICATION OF METHOD
The method of analysis described in this paper has been used to analyze a series of restrained
and unrestrained columns tested in the laboratory (Poh and Bennetts 1995), and is being used
to develop practical design rules for the use of unprotected steel members in multistory office
construction. In this paper, the method is used to analyze the behavior of steel columns of large
cross-sectional area subject to a temperature gradient.
In the early 1990s, a steel-framed office tower was proposed for the city of Melbourne,
Australia. The proposed construction included the use of a braced service core surrounded by
dry wall construction (Fig. 8). The columns within the service core formed part of the bracing
system for lateral load resistance but, more particularly, were required to resist high levels of
gravity loading. The columns were to be fabricated from thick plates and varied in cross-sectional
size from 300 x 300 mm to 500 x 500 mm. Unfortunately, for a variety of reasons, steel
construction was not used for the building, but was replaced by a reinforced-concrete solution.
It is arguable that members located within a fire-resistant envelope should not be required
to have a fire rating, but this argument was not accepted by the regulatory authorities at this
time, and members within the core were required to have a fire resistance of two hours. Never-
theless, it was considered desirable that these large columns not be fire-protected. As illustrated
in Fig. 8, the columns are potentially exposed to fire on two or three adjacent sides because of
the dry wall construction.
The massiveness of the columns suggested that fire protection was probably not required in
order to achieve the required fire resistance. The columns could not be fire-tested under load,
because of their high capacity, and it was necessary to assess their fire resistance by calculation.
The behavior of one of these columns when exposed to standard fire test conditions is now
considered. The member chosen for analysis has a solid steel cross section of 500 x 500 mm
and is 4 m long. The member is located at the corner of the service shaft and is, therefore,
potentially exposed to fire on two adjacent sides while the other sides are assumed to be fully
insulated.
The member is assumed to be subjected to the ISO standard fire for a period of 2 hr. A two-
dimensional transient heat-flow analysis of the cross section was conducted using the program
"TASEF-2" (Wickstrom 1979) to determine temperature distribution throughout the cross sec-
tion. The temperature distribution is assumed to be identical at every cross section along the
fully
insulated + FIRE • ~
••
• FIRE fully
Insulated
• FIRE
Downloaded from ascelibrary.org by National Institute of Technology Karnataka on 01/03/19. Copyright ASCE. For personal use only; all rights reserved.
••
'"
•
FIRE
z
(a)
•
• "'or] •
FIRE
(b)
FIG. 9. Temperature Distribution over Cross Section after Two
Hours of Exposure to ISO Standard Fire FIG. 10. Discretization of: (a) Member; (b) Cross Section
200
--
"room
\/_----2OO'C
A-----3OO'C
TI'CriP ·
200 ,
l
,,..-----4OO'C
l!.'50 .r-----l5OO'C
I 100 'Ir - - - - - 8 O O ' C
1000
...
...
0.01
0-
0-
<GO 0-
,..-----7OO'C JllIO ...
50 IJ V •
~
8OO'C
O'O-l,---r-.......,--r-"'T"""~it-J
IJV lIOO'C
8
: ~-o.25
"
200
_ «XI
_rorcJ
eoo eoo '000 2
_ 3
FIG. 11. Steel Properties: (a) Yield Stress and Modulus of Elas- FIG. 12. (a) Temperatures; (b) Total Strains; (c) Stresses As-
ticity; (b) Stress-Strain Curves sociated with Subarea at Two-Hour Fire Exposure Period
member throughout the fire exposure period. The temperature distribution over the cross section
after 2 hr of fire exposure is shown in Fig. 9.
For the analysis, the length of the member is divided into 20 equal segments and the cross
section is divided into 100 subareas of equal size (Fig. 10). For simplicity, the member is assumed
to be initially straight and stress free. The member is assumed to be pin supported at the ends
through the center of the cross section, and it does not have applied restraints. The origins of
the coordinates are placed at the center of the cross section.
thermal gradient. This is illustrated in Fig. 12, which shows the temperatures, total strains, and
total stresses associated with the subareas within the cross section at the two-hour fire exposure
period, assuming the cross section to be subject to zero external actions. With the exception of
the coolest corner, the subareas in the corners of the cross section are yielded in compression.
The cross-section analysis can be used to determine any three combinations of the six un-
knowns (actions or deformations). This method (Appendix I) has been used to determine the
capacity of the cross section under combined actions, which is shown in Fig. 13 for the combined
actions of axial force with uniaxial bending and with equal biaxial bending at the two-hour fire
exposure period. In this figure, the maximum axial force and moments are normalized against
the room temperature values of squash load and moment capacity, respectively.
"""
t
-
•.0-r-------:;;r.----:::-:::::o-1 1.0..,-----=------,
. . .r --
........ atroom
.OhOtJrs
\
""'"
1000
O'O-t---......,--+--~...L.-_____1 iii I
-0.5 0.0 0.5 1.0 -'.0 -0.5 0.0 0.5 1.0 I 10 15 20
max. m... 01 m,.J (moment ~ mIX. m. or m,1 (moment capedty) &. o0oi4, <mm) I, o0oi4, (mm)
(b) (eI)
FIG. 13. Force-Moment Interaction Curves under: (a) Uniaxial FIG. 14. Deflection of Member under Different Levels of Axial
Bending; (b) Equal Biaxial Bending Compression
20..,------------, 25..,_----------,
p.o
P.o cooler-
•• cooler--
••
·
P .10000kN
.::.
e
20 p 20000kN
lower
P.10000kN lhennel
- p. 2CIOOO kN •
..·P.30000kN """enslon
I.5 ....
1,0
j ~.
(!-., •• •
••
dIrection of FIRE dlrectlonof FIRE
cIefIeclIon deIlK1lon •
T
15
O~__,.._--r-,-~-_r_---I
o
/
.......
'
......
"-<'-__,.._--.-,---,--,..--1
holler-
higher
lhennel
expansion •• ho~~;-
slllIness
•
•
•
0.0 o.a
_"'IIro
1.0 1.6 2.0 2.1 3.0
~
0.0 0.5
_
1.0 1.6 2.0
"' .... ",pOI'''" (hau~
2.6 3.0
(b) (b)
(I) (IJ
FIG. 15. Member Deformations: (a) Mldhelght Lateral Deflec- FIG. 16. Effects of: (8) Differential Thermal Expansion; (b) Dif-
tion; (b) Axial Deformation ferential Material Stiffness on Direction of Member Deflection
CONCLUSIONS
This paper described a numerical model that can analyze the nonlinear behavior of members
under fire conditions. The model can take into account any stress-strain relationship, any tem-
perature distribution, unloading and reloading, end and intermediate restraints, and the appli-
cation of end and transverse actions. The influence of creep can also be taken into account.
The model considers both geometric and material nonlinearities.
The model uses an innovative solution procedure that enables the solution of highly nonlinear
situations, and has provided an important tool for the study of the behavior of members under
elevated temperature conditions and where there is a substantial temperature gradient and
restraint.
The application of the model was demonstrated in relation to a practical example in which
the fire resistance of an unprotected steel column of large cross-sectional area was evaluated.
APPENDIX I.
The cross-section action-deformation equation is
"
2:" S,
;--,-1
2:"
i= I
SiX; 2:" SiYi
ie-I
2:
"
1-'--1
(£j1 - £;')s; 2:f;
, 1
or simply
{A} = [S]{D} + {B} + {C} (9b)
This equation presents a set of three simultaneous equations, from which any chosen com-
binations of three unknowns (cross-section actions A or deformations D) can be solved.
The methods to solve for different combinations of unknowns are presented in the following
sections.
(11)
Thus, if AI' Dz, and D3 are the three unknowns, partially inverting [5] gives
(12)
Downloaded from ascelibrary.org by National Institute of Technology Karnataka on 01/03/19. Copyright ASCE. For personal use only; all rights reserved.
where
(13)
Similarly, if Db Az, and A3 are the three unknowns, partially inverting [5] gives
(14)
where
(15)
5 _ 5:12 5 21)
( 31 S..,,,