Sei sulla pagina 1di 22

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/226137116

Compartment Masonry Walls in Fire Situations

Article  in  Fire Technology · January 2001


DOI: 10.1007/s10694-006-7509-6

CITATIONS READS

15 204

4 authors, including:

Ali Nadjai F. Ali


Ulster University Ulster University
135 PUBLICATIONS   708 CITATIONS    80 PUBLICATIONS   686 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

LOCAFI View project

Fire Performance of ultra-high performance fibre reinforced concrete beams View project

All content following this page was uploaded by Ali Nadjai on 05 February 2015.

The user has requested enhancement of the downloaded file.


Fire Technology, 42, 211–231, 2006

C 2006 Springer Science + Business Media, LLC. Manufactured in The United States
DOI: 10.1007/s10694-006-7509-6

Compartment Masonry Walls in Fire


Situations
A. Nadjai∗ , University of Ulster, FireSERT, Block 27, Jordanstown, Belfast,
BT37 0QB, Northern Ireland, UK
M. O’Gara, Civil and Structural Computer Services Ltd,
F. Ali and R. Jurgen, University of Ulster, FireSERT, Jordanstown, Belfast,
BT37 0QB, Northern Ireland, United Kingdom E-mail:a.nadjai@ulst.ac.uk

Published online: 24 April 2006

Abstract. A series of analytical studies were conducted to investigate the behaviour of


masonry walls supporting a reinforced concrete slab under a number of fire scenarios. The
main purpose is to use various types of sub-assembly in predicting the structural behaviour in
fire. The analysis of sub-frames rather than complete structures is attractive since they have
the potential to reduce computation time dramatically, which could be of significance if large
numbers of results need to be generated for design documents. The analyses are all performed
using a finite element program, called MasSET (Masonry Subject to Elevated Temperatures),
whose most recent development has been the capability to take into account the interface
element between the masonry wall and the slab. Separation, sliding, and opening and closing
of initial gaps at the interface between the masonry wall and the reinforced concrete slab are
accounted for by adjusting the properties of interface elements.

Key words: fire resistance, finite element, structure and design

Introduction
Building regulations require that the structural integrity of a building is maintained for a
specified period in the event of a fire. Design codes for structural masonry have traditionally
adopted a prescriptive approach for determining fire protection requirements, although
more recently some calculation methods [1, 2] have been introduced for estimating the fire
resistance. These are generally concerned with isolated elements and are based mainly on
the results of experimental studies. Large-scale structures are rarely fire tested because of
the cost and the physical limitations of standard furnaces. Hence there is little experimental
evidence relating the fire resistance of individual members to that of a complete building
frame assembly, even though there are indications that continuity between adjacent elements
may significantly influence fire survival. A way of overcoming size limits on testing is the
use of numerical simulations. There has, therefore, been much interest in the development

∗ Correspondence should be addressed to: University of Ulster, FireSERT, Jordanstown, Belfast, BT37 0QB,

Northern Ireland, United Kingdom E-mail:a.nadjai@ulst.ac.uk


212 Fire Technology

of computer programs, which model the behaviour of structural frames, and sub-assemblies
in fire.
Prediction of the structural response of building frames to fire involves a series of highly
iterative non-linear calculations at different temperatures, and analysing extensive frames
is often very time consuming. In practice, fires are often localised by compartmentation to
stay within certain areas of a building, and this poses the question of whether a study of the
complete frame is really necessary. The use of limited subframes has long been accepted
as sufficiently accurate for structural analysis in ambient-temperature design, but little has
been done to investigate the validity of this type of simplified model in analysis for fire
engineering.
In fire separating elements, such as masonry walls, heat is usually exposed to one side of
the element only. This is particularly important in the case of brick masonry due to its low
thermal conductivity, producing high thermal gradients over the cross section [3]. Thermal
bowing is therefore produced due to differential thermal expansion. With the hot face of
the wall expanding more rapidly than the cool one, the wall will tend to bow towards the
fire (see Figure 1). The fire-exposed face of the wall will also experience a considerable
reduction in mechanical material properties, which effectively can be represented as a
reduction in thickness of the hot face [4]. As a result of the change in thickness, any applied
load will have moved towards the fire exposed face and must be recognised as having the
advantageous effect of maintaining structural stability by counteracting the thermal bow.
The current design codes do not provide concise calculation models for the determination of
fire resistance times. Some design guidance is available based entirely on isolated standard
fire tests, which relate minimum periods of fire resistance to required wall thickness [1]. The
thermal displacements together with material property degradation may result in structural
collapse of the wall depending on boundary conditions, magnitude of the applied load and
geometry of the wall [5, 6].
A finite element model called MasSET (Masonry Subject to Elevated Temperatures)
has been developed [6] taking into consideration the material and geometric non-linearity,
the cracking and crushing using the stress-strain failure criterion for masonry walls under

Applied Load

Fire Side cracks

Restrained boundary
(a) (b)

Figure 1. (a) Horizontal-cracking apprising in an axially loaded


masonry wall, (b) deformation and crack patterns in a non-loaded
masonry wall.
Compartment Masonry Walls in Fire Situations 213

elevated temperature. The inclusion of the geometric non-linearity is considered as an


important factor not to be neglected. The benefit of it is to accommodate the important P-
effects in an axially loaded wall, which can largely contribute to the structural collapse.
The interface element, which can predict wall-slab interaction behaviour under elevated
temperatures, is also taken into consideration, in which cracking or debonding of the
mortar layer on the tension side is a dominant mode of failure. Therefore, the tensile
properties of the wall must be based on the bond strength of the mortar layer in the
brick wall.
This paper compares the effectiveness of analysing different two-dimensional sub-
assemblies to predict the behaviour in fire of a full plane composite frame. The comparison
includes deflections, internal forces and limiting temperatures.

Finite Element Idealisation


The temperature distributions over the thickness of masonry walls are generally curvi-
linear, giving rise to a non-linear application of thermal strains. However, due to the shear
fibres in a continuum material, the strain distribution may only occur linearly, and indeed
the shape functions adopted are formulated to adhere to this stipulation. Therefore the free
thermal strains εth , associated with the curvi-linear temperature profile are incompatible
with actual linear flexural strain distributions [7], thus, at any points, thermal expansion
cannot occur freely and is internally restrained producing thermal stresses. These thermal
stresses are self-equilibrating both axially and rotationally and hence do not contribute to
external displacements. The distribution of thermal strains through the thickness of a wall
heated on one face is generally similar to that shown in Figure 2. Assuming that there are no
external loads, the resulting thermal stresses will cause compression on the outer sections
and tension in the center of the wall. These tensile stresses may be of a magnitude to cause
cracking in the region.

Mesh Density Convergence Studies


It was necessary to ascertain the optimum mesh density in obtaining an accurate solution
to thermo-structural analysis. In masonry walls the magnitude of thermal bowing is largely
dependant on the temperature distribution through the wall thickness. It has already been

Free thermal
strain, εth Wall thickness
Sign
- ve convention,
tension +ve
FIRE 0 0
SIDE
+ve

Actual thermal Resulting thermally induced


strains, εath stresses σtis, caused by
restrained thermal strains.

Figure 2. Self-equilibrating thermal stresses caused by a non-linear


temperature distribution.
214 Fire Technology

highlighted that the thermal profile in the wall thickness is generally curvilinear, creating
thermal stresses. Given these factors the controlling factor in mesh descretisation was the
number of elements through the wall thickness. An increased number of elements would
represent the thermal profile and capture thermal stresses more accurately. The number of
elements along the height of the wall was therefore determined based on acceptable limits
of aspect ratio.
It was recognised that the number of elements required through the wall thickness was
also dependent on the degree of non-linearity of the temperature profile. The non-linearity
of the thermal profile becomes more exaggerated with increasing wall thickness and heating
rate. A mesh density convergence study was conducted investigating the optimum number
of elements to employ through the wall thickness, subject to varying degrees of temperature
non-linearity. Analysis was carried out based on the free thermal bowing of an arbitrary
cantilever wall.
Figure 3 illustrates three temperature profiles at a snap shot in time over three respective
heating regimes (corresponding to variable wall thickness). Temperature case I was based
on recorded experimental data from a 100 mm thick concrete masonry wall exposed to
the BS476 standard fire curve. The temperature information from case I was then directly
mapped on to walls of 200 mm and 300 mm thick to form temperature cases II and III (i.e.
the experimental 100 mm temperature profile formed the first 100 mm of the temperature
profile in cases II and III, with a linear extrapolation being assumed between 100 mm and
the unexposed face of the wall).
Figure 4 shows the top horizontal displacements according to temperature cases I, II, and
III. The effect of increasing elements through the wall thickness was immediately apparent.
With temperature cases I and II, two elements thick gave appropriate accuracy, however as
the degree of temperature non-linearity increased (case III) convergence of results required
at least three elements thick.

900 Temperature non-linearity 80.0 Number of elements thick


800 1
I 70.0
Top horizontal displacement (mm)

2
700 II 60.0 3
600 III
Temperature (C)

4
50.0
500
40.0
400
30.0
300
200 20.0

100 10.0

0 0.0
0 20 40 60 80 100 0 5 10 15
Wall thickness (mm) Time (mins)
(a)

Figure 3. Temperature profiles, with varying degrees of


non-linearity
Compartment Masonry Walls in Fire Situations 215

60.0 70.0 Number of elements thick


Number of elements thick
1
Top horizontal displacement (mm)

60.0

Top horizontal displacement (mm)


50.0 1 2
2 3
3 50.0
40.0 4
4
40.0
30.0
30.0
20.0
20.0

10.0
10.0

0.0 0.0
0 5 10 15 0 5 10 15
Time (mins) Time (mins)
(b) (c)

Figure 4. Sensitivity to thermal bowing to number of elements


through the wall thickness (a) Temperature case I, (b) Temperature
case II, (c) Temperature case III.

Eight node isoparametric elements, with quadratic serendipity shape functions, were used
to model the masonry wall. In the finite element idealization, temperatures were assigned
to each node. The masonry panels were commonly descritised into three elements deep
over the thickness of the wall (see Figure 5) giving seven nodal points to represent the
curvi-linear temperature profile, thus enabling the effects of thermally induced stresses to
be calculated and accounted for.
Material Modelling
The two phase masonry material of brick unit and mortar joint exhibits distinct non-linear
stress - strain behaviour. Each material component may be modelled separately providing
a more accurate description of the macro level behaviour between the two constituents.
Alternatively, an equivalent homogenised material may be employed, giving more flexibility
in the finite element mesh descritisation.

σc εσ 3
c = .  3 (1)
σult εult
2+ εσ
εult

Figure 5. Three eight noded isoparametric elements used across the


thickness of the wall.
216 Fire Technology

σ cult

εn εcrac
εσ
σt
ε ult ε cu
Figure 6. Stress-strain relationship of masonry materials in
compression and tension.

c
Where; σult = Ultimate compressive strength, σ c = compressive strength, εult = Ultimate
strain, εσ = stress relation strain.
Material non-linearity was approached using the tangent modulus method. The tem-
perature dependent parabolic stress-strain relationship for a concrete material proposed
in Eurocode 4 [8] was adopted with a linear strain-softening branch. In order to relate
the uniaxial stress-strain relationship to a two dimensional continuum material, the well
established biaxial failure envelope for concrete [8] was employed.
The brittle tensile nature of the masonry materials was accounted for by the inclusion of
a smeared cracking model. If the maximum principle stress at any gauss point exceeded the
user specified tensile failure stress, then a crack was assumed to develop in a plane normal
to the direction of the maximum principle stress over the associated gaussian integration
area. The crack was simulated by setting the Young’s modulus in the direction normal to the
crack as zero. The stress in the direction normal to the crack was not set to zero immediately
but reduced according to a linear crack softening relationship. The effects of crack closure
were also incorporated by restoring the compressive strength of the material in the relevant
direction. A certain amount of shear capacity was retained, but reduced as the crack width
increases. The σ − ε relationship of the cracking model employed is shown in Figure 6.
Crushing of a material was modelled by reducing all strength properties and stresses in
an integration area to zero when the effective biaxial strain at that particular gauss point
exceeded the user specified crushing strain.

Elevated Temperature Effects


The effects of thermal expansion were included by the application of thermal strains εth ,
which were defined by the εth -temperature relationship adopted by Eurocode 4 [8]. Thermal
strains are indeed an initial strain component, and are therefore independent of stresses.
The nature of the finite element method written in displacement formulation prevents a
direct explicit displacement-strain relationship. Therefore any deflection sought after (in
this case thermal bowing) must be achieved from an equation of the form;

{P }j = [K]i {a}i + {f }j (2)


Compartment Masonry Walls in Fire Situations 217

where {P} is the vector of externally applied nodal loads, [K] is the stiffness matrix, {a} is
the vector of nodal displacements and {f} is the vector of loads due to element body forces.
Superscripts i and j denote iteration and increment, respectively.
Temperature changes result in an initial strain vector of the form;
 
 εxth 
{εth }i = εyth (3)
 
γxyth

where in an isotopic material γ xy th = 0, as no shear strains are caused by thermal dilation


The thermal strains must be converted to equivalent nodal loads to act as body forces, {fth },
which are capable of creating the state of thermal strains in the element, when applied to
that element.

T
{fth }j = − [B]i [D]iT {εth }j dV (4)
Ve

The thermal stains are converted to imaginary thermal stresses, {σ }th , using the current
tangential elasticity matrix, [D]T . The thermal stresses are then integrated over the element
volume, therefore the thermal strains must be evaluated at the gauss points to facilitate the
gauss quadrature integration. Temperature, Ti , is assigned to each nodal point, and with use
of the element shape functions Ni ; the temperature at the gauss points Tgp , is determined,
 
 N1 
{Tgp } = [Ti ] {Ni } = [T1 , T2 , . . . . . .] N2 (5)
 
:

then using the explicit εth —Temperature relationship employed, the thermal strains are
evaluated. Rearranging Equation (2) gives;

{P }j − {fth }j = [K]i {a}i

and by substituting Equation (4);


T
{P }j + [B]i [D]iT {εth }j dV = [K]i {a}i (6)
Ve

The equilibrium equations are solved for nodal displacements, a, and it is apparent that
when the external applied nodal loads are zero, the displacements are entirely due to thermal
effects. Upon determining the nodal displacements, the total strains εtot , are calculated;

{εtot }i = [B]i {a}i (7)


218 Fire Technology

The individual components of the total strain were recognised in order to determine the
stress related strain component for the calculation of internal forces. Equation (8) describes
each strain component;

{εtot } = {εσ }j + {εtr }j + {εth }j (8)

where; εtot = total strain, εσ = stress related strain, εtr = transient strain, εth = thermal
strain
The transient strain component has been reported to be significant in concrete materials
that were loaded and experiencing heating for the first time. This strain component has the
effect of reducing and redistributing thermal strains in loaded elements, preventing excessive
damage by thermally induced stresses [9]. In the present analysis the transient strain model
proposed by Anderberg [10], Equation (9), has been employed, and is dependent on both
thermal strain and stress level.

j j −1 j
εtr = −2.35 σσ c .εth
j
ult
j −1 (9)
εtr = −0.0001.T . σσ c
ult

Where; σ j −1 = stress in the element from the previous increment; σult c


= Compressive
strength at ambient temperature; T = element temperature; εth = thermal strain.
From Equation (9), as the stress level increases, so does the transient strain component,
thus reducing the effective thermal strains.
Modification of material strength and stiffness with increase in temperature was incor-
porated which is particularly influential in the case of loaded masonry walls. Upon the
c
application of temperature, the material properties of compressive strength σult , ultimate
strain εult , crushing strain εcu , and tensile strength σ t , were modified for each integration
c
area according to the temperature at the gauss point. Both σult and εult were modified ac-
cording to the Eurocode 4 recommendations. A tri-linear decay model for σ t based on test
results by Thelanderson [11 ] was employed, while the linear crushing strain model used
by Anderberg [10] was implemented.

Modelling the Behaviour at the Interface


The behaviour between the wall and the rest of the bounding structure is simulated using
a modified form of the joint element originally proposed by Goodman et al. [12]. The
element has six nodes (Figure 7). The degrees of freedom are the relative displacements,
u and v, of opposite nodes. The relative displacements along the element are related to
the relative displacements at the nodes by

 

 u1 − u4 

 v1 − v4 
 







u N1 0 N2 0 N3 0 u2 − u5
= (10)
v 0 N1 0 N2 0 N3 
 v2 − v5 


 


 u3 − u6 

 
v3 − v6
Compartment Masonry Walls in Fire Situations 219

Figure 7. Finite element model of masonry and interface element


with slab.

where

1 1
N1 = ξ (ξ − 1), N2 = (1 − ξ 2 ), N3 = ξ (ξ + 1)
2 2

The relative normal and tangential movements, n and t , respectively, can be calculated
from the relative displacements as




t cos θ sin θ u
= (11)
n − sin θ cos θ v

where θ is the angle between the local and global axes.Thus the strain matrix can be written
as


N1 0 N2 0 N3 0 α 0 −α 0
{ε} = {U } (12)
0 N1 0 N2 0 N3 0 α 0 −α
 
100
in which α =  0 1 0 and the displacement nodes are given as
001
220 Fire Technology
 
{U } = u1 v1 u2 v2 u3 v3 u4 v4 u5 v5 u6 v6

Equation (12) may, formally, be written as

{ε} = [B] {u} (13)

where [B] is the strain nodal displacement matrix, and {u} is the nodal displacement vector.
The stiffness matrix of the boundary element is given by

[K] = [B]T [D] [B] dx (14)

in which matrix D relates the normal and tangential stresses at the interface to the relative
normal and tangential movements. The dimension of the stiffness matrix resulting from
Equation (14) is 12 × 12. It can be used directly between two eight node isoparametric
elements.
The stresses and strains (the relative movement) at the interface may, in general, be
related through a constitutive relationship [13, 14] such as




σt Ktt Ktn t
= (15)
σn Knt Knn n

In this work however, the relative shear and normal displacements are assumed to be
uncoupled. Therefore Equation (15) is reduced to




σt Kt 0 t
= (16)
σn 0 Kn n

where Kt and Kn are the tangential and normal interface stiffness coefficients, respectively.
Different interface conditions can be modelled by adjusting these coefficients. When the
normal strain is in tension, i.e. n > 0.0, separation is assumed to be taking place, and
the normal stiffness, Kn , is taken as zero. Since a separated interface cannot take up any
shearing stress the shear stiffness, Kt , is also taken as zero. If n ≤ 0.0, the interface is
assumed to be in contact, and a very high value is assigned to Kn .
The tangential stress-strain relationship is assumed to be elastic-perfectly plastic using
a Mohr-Coulomb yield criterion with zero cohesion. When |σt | < |µσn |, where µ is the
coefficient of friction of the interface, firm contact is assumed and Kt is assigned the value
of the slope of the tangential stress-relative displacement curve resulting from shear box
tests [15 ]. Slip takes place whenever |σt | exceeds|µσn |. In such a case, σ t is reduced to
µσ n and Kt is taken as zero.
Values of the stiffness coefficients Kt and Kn used in the analyses for different interface
conditions are listed in Table 1.
Compartment Masonry Walls in Fire Situations 221

Table 1.
Selection of interface stiffness coefficients

Interface conditions Kt Kn

Firm contact n > 0.0 and Experimental value Very high value
|σt | < |µσn |

Contact interface with slip Very low value Very high value
n = 0.0 |σt | ≥ |µσn |

Separation or initial lack of fit Very low value Very low value
n > 0.0

P- Effects
It was recognised that the problem at hand contained certain aspects of geometric non-
linearity. Changes in geometry due to deformation significantly influence the predicted
response of a masonry wall [9]. The total lagrangian descriptor was used to redefine
the structural geometry, and account for the additional stresses experienced in a laterally
deforming axially loaded masonry wall. Green’s strain tensor was used to include higher
order strain terms. The membrane stresses, which exist in an axially loaded slender structure,
may actually have a stiffening effect and have been accounted for by the inclusion of the
geometric stiffness matrix.

Validation of the Finite Element Model


Full Scale Model
A full-scale axially loaded masonry wall with a dimension of 3.0 m × 3.0 m × 90 mm thick
wall panel constructed of clay masonry units test was conducted by Gnanakrishnan [16].

0.014

0.012

0.01
Clay
Thermal Strain

0.008 Mortar

0.006

0.004

0. 002

0
0 200 400 600 800 1000 1200
Temperature (C)

Figure 8. Thermal strain relationship of mortar and clay according


to Gnanakrishnan [16].
222 Fire Technology

Exposed face temperature (C)

0 200 400 600 800 1000


0
Mid wall horizontal displacement (mm)

-10 000

-20
-30

-40

-50 Test
No rotational restraint
-60
20% Rot.
-70
-80

Figure 9. Effect of top rotational restraint on the mid wall


horizontal displacements.

The top and bottom boundary conditions were laterally and rotationally restrained, while
the vertical edges were unrestrained with sufficient clearance for free thermal expansion.
Sandwich plattens constructed of a ceramics fibre core and laminated by two steel plates,
were used between the wall top and bottom boundaries and the loading plates. Their purpose
was to distribute the load, but also exhibited a high degree of permanent strain allowing
rotation of the top and bottom boundaries. The wall was subject to the Australian standard
time-temperature curve, and an axial applied stress of σ y = 0.877 N/mm2 .
The 3000 mm high × 90 mm thick wall was modelled using MasSET, and was descritised
using 3 elements thick by 75 elements high. The clay masonry unit and Ordinary Portland
Cement mortar joint materials were individually represented, and compressive strengths
of 59.4 and 8.2 N/mm2 as recorded by Gnanakrishnan [16], were utilized, respectively.
The secant coefficients of thermal expansion for both clay and mortar were measured
by associated testing, and were translated to the thermal strain-temperature relationship
as is shown in Figure 8. Both relationships were implemented during analysis. Interface
elements were introduced between the top and bottom boundary conditions and their loading
surfaces. Sensitivity analysis to varying degrees of rotational stiffness was conducted since
the behaviour of the sandwich plattens was not quantifiable.
Relevant mechanical and thermal material properties were determined by Gnanakrishnan
[16] and used directly in the numerical modelling of the MasSET. Figure 9 illustrates the
numerical results yielded by MasSET, and clearly shows that increasing rotational stiffness
produced a decrease in the extent of thermal bowing. With free rotation, failure due to
buckling was experienced, and occurred prematurely in comparison to experimental results.
Best overall agreement was achieved when employing 20% top and bottom rotational
restraint [6], although numerical displacements were much smaller than experimental after
800◦ C. The value of rotational stiffness was assumed to be constant in the analysis: however
the sandwich elements used at the boundaries in test were reported to have a high degree of
Compartment Masonry Walls in Fire Situations 223

Deflection
L/4 Temperature Wall Thickness
L /3

L/4

L L /3
Fire side
L/4

L/3
L/4
-50 -34 0 +34 +50

(a) (b)

Figure 10. (a) Measurement positions. (b) Position of


thermocouples through wall and reference system.

permanent strain. Given this, the value of rotational restraint may have been overestimated
at higher rotations, hence resulting in decreased numerical displacements.

Half Scale Model


A half scale model wall was selected from the experimental work tested at the University
of Ulster [17]. The model walls were 430 mm wide by 1330 mm high strip columns, with a
thickness of 50 mm (see Figure 10(a) subjected to 50% of the design ultimate load.. Masonry
units consisted of concrete with a specified crushing strength of 20 N/mm2 , bedded on a
1:3 mortar mix, using Ordinary Portland Cement. The bases of the walls were built on a flat
steel plate and the applied axial load was distributed over the width of the panel top through
a steel loading plate which was restrained against rotational movements. The vertical
edges of the walls were unrestrained with sufficient space for free thermal expansion. The
axial load was kept constant during all tests using an ‘Enerpac’ hydraulic system and was
monitored using load cells. Lateral deflections were measured at support positions (top and
bottom) and at midspan using LVDT’s. Temperatures were recorded using thermocouples
arranged in two sets (5 per set) at vertical third points of the wall. At each location one
thermocouple was placed in the centre of the wall, two were placed on each face and the
remaining two at distances of d/6 and 5d/6 into the wall measured from the wall face
(see Figure 10(b).
The cross-section of the wall was modelled using eight noded isoparametric elements, 3
elements wide and 69 elements high, creating a mesh size of 207 elements with 766 nodes.
The height of the brick unit and mortar joint used in analysis was proportionally larger
than in the actual test for the purpose of reducing the problem size to save computational
time. Seven nodal points over the wall thickness enabled the curvi-linear temperature
distribution to be simulated [5]. The steel loading plate was also modelled and prevented
numerical instabilities caused by local crushing of the top bricks. Contact elements were
also introduced between the top of the wall and the loading plate and at the bottom of
224 Fire Technology

Temperature, Tef (°C)


0
0 200 400 600 800 1000

-10
Central Lateral Deflection (mm)

-20
Analysis
Test
-30

-40

-50

-60

Figure 11. Comparison between experimental and analytical


Central deflection V’s Exposed face temperature for FW1.
the wall. These elements where given zero tensile strength in the Y-direction to ensure no
tensile bonding between surfaces.
The test duration was 28 min, at which point complete structural collapse was experi-
enced. During the test the wall clearly bowed towards the fire. The central deflection (cx )
versus exposed face temperature (Tef ) relationship from both experiment and analysis is
shown in Figure 11. The results are in reasonable agreement, where the onset of large lateral
displacements occurred at approximately 812◦ C.
Several authors have reported from fire wall tests that failure due to buckling tended to
occur at a central deflection of 90% the thickness of the wall [16]. The tested wall failed at a
central deflection of 42.5 mm (85% wall thickness), therefore the failure mode was almost
definitely one of buckling instability. The large deflections created tensile stresses on the
exposed face, which produced horizontal cracking in the mortar joints followed by almost
immediate structural collapse. Figure 12 demonstrates the crushing of the mortar joint at
the base of the wall on the fireside by using the MasSET. This effectively relieved some of
the rotational restraint at the bottom of the wall facilitating the onset of large displacements.
This hypothesis could not be compared to experimental events, as the fireside of the wall
could not be seen.

Subframe Analysis
Two particular subframe assemblages were considered as shown in Figure 13. The 100 mm
thick wall supported 150 mm thick concrete slabs which were protected from elevated
temperature during the analysis. The 3000 mm high walls resulted in a slenderness ratio
of 30. Each subframe was subject to the BS6399 design load, with the self weight of both
slab and wall also being considered in the analysis. The magnitudes of applied loading,
subsequent stresses, and stress levels are given in Table 2, while referring to Figure 13.
Compartment Masonry Walls in Fire Situations 225

Figure 12. Failure mode, deflections are magnified by a factor of


4.0.

L L
L

d d
h
t BS. Design Load h
t
BS. Design Load

d d
h = 3000 mm
h h t = 100 mm
t d = 150 mm
t L = 2500 mm

(a) (b)

Figure 13. (a) Subframe 1, (b) Subframe 2.


226 Fire Technology

It was recognised that in real structural situations, the end conditions of a masonry wall
may not actually be bonded to the adjacent surfaces, but simply in contact. The thermal
movements caused by fire environments may cause slip or separation of the masonry wall
contact at the boundaries, as previous testing has shown. The partial separation of a boundary
contact due to rotation may result in a redistribution and concentration of an axially applied
load towards one face of the element. These effects can be quite significant and it has been
proclaimed that they enhance structural stability in some cases by counteracting the thermal
bow [3]. To accommodate these effects, the six noded interface element has been employed
between the wall and slab interface.
The cross section through the thickness of the wall components was modelled using eight
noded isoparametric elements, adopting a mesh density of 3 elements wide and 75 elements
high. The seven nodal points over the wall thickness enabled a non-linear temperature
distribution to be simulated. Previous mesh density convergence studies using MasSET has
shown that with a curvilinear temperature distribution, 3 elements through the thickness
of the wall is sufficient for capturing the thermal stresses created and material property
degradation effects. Since the slab elements were protected form high temperatures, a
discretisation of one element thick giving ten elements wide was adequate. For the purposes
of computational efficiency a homogenised masonry material was adopted in analysis, with
typical values assumed for all properties.
Since MasSET is not currently coupled with a thermal analysis module, the temperature
distributions recorded during firewall testing at the University of Ulster [17] were utilised
in the analysis. In these tests the elements were subject to the heating regime of the BS476

Table 2.
Loading Conditions.

Axial stress, σ A Stress Level


BS6399 N/mm2 σ A /σ c ult
Loading
kN/m 1st 2nd 1st 2nd

Subframe 1 6.32 1.285 0.645 0.064 0.032


Subframe 2 6.32 0.713 0.357 0.036 0.018

Table 3.
Failure Modes of Each Analysis.

Fire location

1st 1st + 2nd 2nd

Subframe 1 Crushing of bricks at Crushing of bricks at Lateral instability


boundaries in bottom boundaries in bottom buckling of top wall
wall. wall
Subframe 2 Lateral instability Lateral instability Lateral instability
buckling of bottom wall buckling of top wall buckling of top wall
Compartment Masonry Walls in Fire Situations 227

fire curve. In both subframes several scenarios of fire location were considered, i.e. fire on
1st floor only, fire on both floors, and fire on the 2nd floor only.
Figures 14 and 15 illustrate the mid wall lateral displacements, with the exposed face
temperatures, for each of the three fire situations in subframe 1, and subframe 2 respectively.
The observed critical failure mode for each of the six analysis conducted is summarised
in Table 3. The failure temperatures and times are also given in Table 4. Figures 16 and
17 show the deformed structure and the various strain components described in section
elevated temperature effects, at an instance immediately prior to failure for subframes 1
and 2.

Figure 14. Displacement v’s exposed face temperature for


subframe 1.

Figure 15. Displacement v’s exposed face temperature for


subframe 2.
228 Fire Technology

Figure 16. Deformations and strain components prior to failure for


subframe 1, with fire on both floors.

It transpired that the thermal bowing of a particular wall was unaffected by the occurrence
of fire in an adjacent upper or lower compartment. The lateral displacements of the top walls
in both subframes was consistently larger than the bottom walls since the stress level in the
bottom walls were higher (see Table 2) leading to an increased transient strain component.
However the structural failure mode and critical temperature/times were affected by the
fire location as shown in Tables 3 and 4, respectively. Where the fire was confined to

Table 4.
Failure times and temperatures of each analysis.

Fire location

1st 1st + 2nd 2nd

Time (min) Temp (◦ C) Time (min) Temp (◦ C) Time (min) Temp (◦ C)

Subframe 1 40 762 33 670 72 930


Subframe 2 74 932 64 893 64 893
Compartment Masonry Walls in Fire Situations 229

Figure 17. Deformations and strain components prior to failure for


subframe 2, with on 2nd floor only.

one compartment, the displacements of the wall in the neighbouring compartment were
insignificant (see Figure 17) and are omitted from Figures 14 and 15.
In subframe 1, the most critical condition was that of fires on both floors, where failure
occurred in the bottom wall by crushing at the boundaries. The occurrence of fire in the
top compartment only, produced the most favourable fire resistance time, creating failure
of the top wall by P- effects. No early failure by crushing was experienced in subframe 2
(since the applied loading was smaller than in subframe 1), and in all circumstances, failure
was due to buckling effects. Subsequently, the most prolonged failure time was given by
fire in the bottom compartment only, since lateral displacements were decreased due the
increased stress level.
The incorporation of the interface elements at the boundaries played a crucial role in
representing real structural end conditions. From Figures 16 and 17, the separation of
the interface elements on the unexposed face is immediately apparent. Thus, the applied
loading from the slab elements became eccentric towards the fire exposed face, serving to
greatly enhance fire resistance in terms of structural stability

Conclusions
In general, the finite element model MasSET proved to be successful. Areas where ex-
perimental simulation proved difficult came as a result of undetermined experimental con-
230 Fire Technology

ditions, in particular complex boundary conditions. However where controlled rotational


stiffness was present, MasSET produced close similitude.
The limited studies described here suggest that the behaviour of masonry walls in a fire can
be adequately represented by appropriate subframes similar to those used in conventional
structural design. The indicative study for frames with protected or unprotected masonry
walls demonstrates quite different behaviour, and clearly shows the necessity for covering
all possibilities in the choice of subframes to be examined. However, the subframe models
chosen are capable of representing the behaviour of the masonry walls at different levels
without introducing unrealistic conditions or restraints. In particular, boundary conditions
should be avoided which cause artificial restraint to axial thermal expansion.
It is possible that subframes could even be used as the basis for a manual calculation
method for limiting temperatures, although it would be preferable to test a wider range of
structural parameters before such methods are established.
The analytical method which has been used to perform these studies is under development
to allow for the effects of plastic strain reversal. The ability to include the material unloading
characteristics will enable the prediction of permanent deformations and residual forces
resulting from a fire. This feature would be of importance when considering the structural
integrity and safety of a building for access after the event of a fire.

References
1. CP 121, Code of Practice for Walling, Pt 1: Brick and Block Masonry, British Standards Insti-
tution, London, 1973.
2. EUROCODE No. 6. Design of masonry structures. Pt 10: Structural fire design, Commission of
European Communities, 1990.
3. G.M.E. Cooke, “Thermal bowing and how it affects the design of fire separating constructions,”
Interflam 88, Fourth Int. Fire Conf. 22–24, March, Church College Cambridge, 1988, pp. 230–
236.
4. S.J. Lawerence and N. Gnankrishnan, “ The fire resistance of masonry walls ”, Fire National
Structural Engineering Conference, Melbourne, 1987, pp. 431–437.
5. A. Nadjai, M.E. O’Gara, and F. Ali, “Finite element modelling of compartment masonry
walls in fire,” International Journal of Computer and Structures, vol. 81, 2003, pp. 1923–
1930.
6. A. Nadjai, M.E. O’Gara, F. Ali, and D. Laverty, “A numerical model for the behaviour of masonry
under elevated temperatures,” International journal of Fire and Materials, vol. 27, 2003, pp.
1–20.
7. D.J. O’Connor and B.W. Scotney, “Determination of equivalent thermal response parameters
for evaluating the structural response of beams subjected to transient thermal environments,”
International Journal of Mathematics, Education, Science and Technology, vol. 26, 1995, pp.
111–130.
8. EUROCODE No. 4. Design of composite steel and concrete structures. Pt 1.2: Structural fire
design (second draft). Commission of the European Communities, 1992.
9. G.A. Khoury, B.N. Gringer, and P.J.E. Sullivan, “Transient thermal strain of concrete: Literature
review, conditions within specimen and behaviour of individual constituents,” Magazine of
Concrete Research, vol. 37, 1985, pp. 131–144.
10. Y. Anderberg and S. Thelanderson, “Stress deformation characteristics of concrete at high
temperatures,” Division of Structural Mechanics and Concrete Construction, Lund Institute of
Technology, No. 15, 1976.
Compartment Masonry Walls in Fire Situations 231

11. S. Thelanderson, “Effect of high temperature on tensile strength of concrete,” Division of Struc-
tural Mechanics and Concrete Construction, Institute of Technology, Neostyled, No. 27, 1971.
12. R.E. Goodman, R.L.Taylor, and T. L. Brekke, “A model for the mechanics of jointed rock,” Soil
Mechanics and Foundations Division, ASCE, vol. 94, 1968, pp. 637–659.
13. A. Nadjai,“The behaviour of steel frames with semi-rigid joints containing unreinforced infill
panels,“ PhD thesis, The University of Sheffield, 1993.
14. I.M. May and J.H. Naji, “Nonlinear analysis of infilled frames under monotonic and cyclic
loading,” International Journal of Computer & Structures, vol. 38, 1991, pp. 149–160.
15. G.J.W. King and P.C., Pandey, “The analysis of infilled frames using finite elements,” The
Proceeding of the Institution of Civil. Engineers, vol. 65, 1978, pp. 749–760.
16. N. Gnanakrishnan, The Effect of end Restraint on the Stability of Masonry Walls Exposed to Fire,
National Building Technology Centre, Technical Records No. 531.
17. D. Laverty, A. Nadjai, and D.J. O’Connor, “Modelling of thermo-strutural response of concrete
masonry walls subjected to fire,” Journal of Applied Science, vol. 10, 2001, pp. 3–19.

View publication stats

Potrebbero piacerti anche