Вы находитесь на странице: 1из 84

Coupling methodologies for integrating CFD and FEM codes to simulate structures in fire

Alan Chan

Doctor of Philosophy The University of Edinburgh 2 !

To dad and mum, Jik Bong Chan and Wing Ying Chan

Declaration

This thesis and the research described and reported within have been completed solely by Alan Chan under the supervision of Torero and "rofessor# !uresh %umar# r !tephen Welch, "rofessor# Jose $#

Alan Chan !eptember &''(

Abstract
Coupling methodologies linking Computational )luid ynamics *C) + and )inite ,lement -ethod *),-+ computer codes for assessing the performance of structures in fire has been developed# This pro.ect has provided an insight into the various strategies of coupling the fire development and the thermal response models# By constructing a software interface between the codes, transfer of results between the C) and ),- will take place automatically# -ethodologies based on one/way

coupling, with information passing from fire to solid/phase code, have an advantage of fle0ibility, particularly where omission of structural components from the fire development model can be .ustified i#e# where there is only a minor influence on the overall gas temperature and radiative e0change# Two/way coupling should be considered when there is a more significant interaction between the fire and the structure#

The work has e0amined coupling methodologies for a C)

code !1)2,345,

developed under the sponsorship of an international Consortium of fire research labs and a ),- code !A)263&5, developed by the 7niversity of $i8ge# The program pair was applied to two test cases9 a hypothetical benchmark scenario with localised fires and various steel components and an e0perimental scenario involving a loaded beam/ column assembly in a flashed/over fire#

Ac"no#ledgements

Table of contents
ABSTRACT............................................................................................................................................4 ACKNOWLEDGEMENTS...................................................................................................................5 TABLE OF CONTENTS.......................................................................................................................6 NOMENCLATURE...............................................................................................................................8 INTRODUCTION ...............................................................................................................................10 OVERVIEW OF HEAT TRANSFER TO STRUCTUREBACKGROUND TO MODELLING OF STRUCTURES EXPOSED TO FIRE.........................................................................................14 )26, -1
,$$2:;###############################################################################################################################4<

T=,6-A$ A:A$Y!2!##########################################################################################################################4> Heat Transfer................................................................................................................................15 Tabular Data.................................................................................................................................17 Simple Fire ModelsThermal Design Formulae............................................................................17 General (or Ad an!ed" Models.....................................................................................................17 !T67CT76A$ A:A$Y!2!#####################################################################################################################4? :7-,62CA$ -1
,$$2:; 2!!7,!########################################################################################################&&

#FD simulations...........................................................................................................................$$ Solid%phase simulation..................................................................................................................$& METHODOLOGY DEVELOPMENT TO INTEGRATE A CFD AND A FEM PROGRAM.... 4 C17"$2:; A""61AC=########################################################################################################################&< TY",! 1) C17"$2:;##########################################################################################################################&> $,@,$! 1) C17"$2:;#########################################################################################################################&( #oupling the fire de elopment to the thermal model....................................................................$' #oupling the thermal response to the me!hani!al response........................................................&$ #oupling the me!hani!al response to the fire de elopment.........................................................&& T(pes and le els of !oupling established in the F)*+ST*,# pro-e!t..........................................&. -,T=1 1$1;Y =2,6A6C=Y##############################################################################################################A> 1:,/WAY !1)2,/!A)26 C17"$2:;#################################################################################################<' General methodolog( of one%/a( !oupling to SAF)*...................................................................0 Transfer File..................................................................................................................................1 #artesian interpolation..................................................................................................................1 Spheri!al )nterpolation ................................................................................................................1

VALIDATION OF INTEGRATED METHODOLOGY.................................................................55 B,:C=-A6% CA!, -1


,$$2:; W2T= C17"$,

C) /),- -,T=1

1$1;Y####################################>>

*esults...........................................................................................................................................51 Sensiti it( studies..........................................................................................................................2$ CT2C- ,BT,6:A$ C1$7-: CA!, -1 ,$$2:; W2T= !1)2,/!A)26###############################################C> CONCLUSION.....................................................................................................................................!" REFERENCES#...................................................................................................................................80 APPENDIX...........................................................................................................................................8

$%ME$C&ATU'E
A a ak !p3 #p D h! ) 4 5f 6 l 7u 9r : :F :/ :f3d *e T Tf Tg t , u / ; ( < / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / area 3m&5 absorption coefficient banded absorption coefficient 3k bands5 specific heat capacity 3JDkgD%5 displacement 3mm5 convective heat transfer coefficient 3WDm&D%5 radiative intensity, i#e# energy per unit area per unit solid angle 3WDm&Dstr5 thermal conductivity 3WDmD%5 thermal conductivity of the fluid 3WDmD%5 characteristic length scale 3m5 latent heat 3kJDkg5 radiation transfer eEuation path length :usselt number *h!685+ 3/5 "randtl number *!p85+ *kinematic viscosityDthermal diffusivity+ 3/5 heat flu0 3WDm&5 outgoing heat flu0 3WDm&5 incident heat flu0 3WDm&5 design fire load density related to the floor area Af 3kJDm&5 6eynolds number *,8+ *inertial forceDviscous force+ 3/5 temperature 3%5 surface temperature 3%5 gas temperature 3%5 time 3seconds5 characteristic velocity 3mDs5 0/component of velocity 3mDs5 y/component of velocity 3mDs5 G/component of velocity 3mDs5 0/coordinate direction 3m5 y/coordinate direction 3m5 G/coordinate direction 3m5

Greek letters / convective heat transfer coefficient 3WDm&D%5 / elemental solid angle 3steradians5 / difference or change 3/5 / emissivity 3/5 / element angle 3radians5 / thermal conductivity 3WDmD%5 / coefficient of viscosity 3kgDmDs5 / pi constant, A#4<4>?&C>< 3/5 / angle of incidence 3radians5 / density 3kgDmA5

/ /

!tefan/BoltGmann constant 3>#CH04'/( WDm&D%<5 solid angle and normal to the surface 3steradians5 ambient convective

Subscripts a / ! /

(ntroduction
The aim of this "h is to establish a validated methodologye0amine methodologies ynamics *C) + and a )inite ,lement -odels of coupling a Computational )luid

-ethod *),-+ code to simulate structures sub.ected to fire# -any engineering problems necessitate multidiscipline physical simulations# !uch problems reEuire the solution of two or more distinctive physical subsystems that are tightly coupled# ,0amples can include fluid/structure interaction, thermo/mechanical interaction, as well as thermo/electric interaction# Attempts of coupling have been achieved in the past in the medical field linking together two leading commercial software, C)B for the C) and A:!Y! for the ),-, for blood flow problems3A5# @arious other code

pairs have also been successfully coupled inby other engineering fields and applied to numerous applications as shown in Table 4#

A$$%&'()&*+ Thermal loads in space re/entry vehicles Thermal interaction in turbines Thermal interaction in turbine blades ,ngine cooling Car underhood simulations Thermo/electrical coupling Biomechanical simulation

C*,- P(&. Tau/A:!Y! )luent/ABAI7! 2n=ouse/!emcef -ecano )luent/ABAI7! !tar/C /"ermas/"os6ad !tarC D)luent/A:!Y! !tarC /-ecano

T(/%- 1# Coupled codes and the associated applications3>5

The main reason for coupling different simulation codes is that in many real world applications the interaction of different physical phenomena must be taken into consideration in order to achieve more representative predictions# The ability to

couple C)

and ),- codes is yet to be fully established in the fire field# By

adopting a coupled methodology one can investigate the behaviour of fire/e0posed structures in order to gain an insight into the essential failure mechanism# )or e0ample, the effects of localised fires on structural components can also be effectively modelled by coupling the C) and ),- code, which allows one to break away from the restrictions imposed by using a standard temperature time curve# Typically, the structural performance of buildings under fire has been assessed on the basis of simplified models in which the compartment under fire is modelled using a standard time/temperaturen ,urocode parametric fire curve3<5# )urthermore, the behaviour of the structure has often been evaluated with respect to the performance of individual structural elements which have been assessed in standard fire resistance tests# etails of the structural behaviour of frames and sub/assemblies under the

impact of localised heating have therefore normally been bypassed# Considering the effects of natural fires and the response of the structure as a whole the actual behaviour might diverge significantly from these model predictions

The minimum legislative level of safety for structural fire design aims to provide an acceptable risk associated with the safety of building occupants, fire fighters and people in the pro0imity of the building# 6eflecting the concerns of the insurance industry, levels of safety can be increased to protect the building contents, the building superstructure, heritage, and business continuity, corporate image of the occupants or owner, and the environmental impact, with conseEuent further benefit to life safety#

The introduction of performance/based building regulations reEuires that the thermal response of real/life structures to fire can be accurately simulated# !imulation of the effects of severe fire on the performance of a structure of buildings reEuires a close coupling between the gas/phase energy release and transport phenomena and the

mechanical analysis in the load/bearing materials# The performance/based approach involves the assessment of three basic components comprising the likely fire development, heat transfer to the structure and the structural response# modelling of the fire can only be achieved via C) etailed

model and adeEuate

representation of the structural response reEuires use of a solid/phase model),-# C) calculations are used to simulate the fire development and the ),- calculations describe the behaviour of the structural members sub.ect to this fire# The main=ence, one of the priorityies is to e0ercise the coupled model in real/life scenarios to e0amine sensitivities of the methodology#

The differences in spatial and temporal length scales, differences in numerical techniEues, and the comple0ity of the computer codes makes the development of an efficient coupled analysis of fire/structure interactions a challenging task# The

ob.ective here will be to develop a treatment of the boundary conditions that will ensure appropriate heat transfer into the assembly# An added comple0ity to the problem is the inherent differences in the time scales of the fire affecting the assembly and that of the heat transfer through the different elements# Therefore, the issue of the timescales of the gas and thermal response, which can help establish the reEuired freEuency of interaction between each sub/model, has to be addressed at an early stage of the coupling#

The ultimate aim of the work, which forms also ran parallel topart of a ,uropean Community pro.ect funded by the 6esearch )und for Coal and !teel *6)C!+ known as )26,!T67C, is to apply a validated methodology to real/life scenarios so as to gain an improved understanding of the behaviour of structures in fire and, together with the study of coupling methodologies# C) codes were used for the fire

simulation, whilst both ),- and finite volume codes were used for the structural

modelling in the 6)C! pro.ect# The latter is therefore strictly referred to as solid/ phase, though ),- is also used as a short/hand form#

%vervie# of )eat Transfer to *tructure+ac"ground to modelling of structures e,posed to fire


This chapter introduces the methods available to resolve the thermal boundary conditions for a structural analysis# 2n the following, the three analysis components J fire modelling, thermal analysis and structural analysis are discussed9

Fire Modelling

Figure 1: Fire modelling in compartment[6]

The most advanced and sophisticated fire modelling techniEue is the use of C) models to predict fire growth and compartment temperatures# C) models have been shown to be successful in the modelling of smoke movement and have more recently been applied to the assessment of the thermal response of structural elements# They are capable of modelling pre/flashover and localised fires in comple0 geometries with smoke movement in multi/compartments#

According to ;#Co03(5, typical C)

models analyse systems involving fluid flow,

heat transfer and associated phenomena by solving the fundamental eEuations of the fluid flow# These eEuations represent the mathematical statements of the conservation laws of physics#

Basically, in a C)

model, the partial differential eEuations of the thermodynamic

variables *:avier/!tokes eEuations+ are solved in a very large number of points in the computational domain# -ost C) models for enclosure fires are appropriate for low/ speed, thermally/driven flow with an emphasis on smoke and heat release from fires# The specific challenges of fire modelling include the adeEuate representation of buoyant flows and incorporation of the effects of turbulent fluctuations#

The input and solution control reEuirement for C)

models is very demanding and

reEuires e0pertise in defining the correct parameters and assessing the feasibility *numerically and physically+ of the calculated results# 1n the other hand, the results are given with much greater detail than Gone models, providing the variables in all points of the compartments, such as temperature, velocity and chemical species concentration#

Thermal Analysis

Figure : Schematic diagram !or heat trans!er[6]

)eat Transfer

=eat transfer is the science of evaluating the energy transfer that takes place between material bodies as a result of temperature differences# The three modes of heat transfer are conduction, convection and radiation# The thermal analysis on structural fire problems can be divided into two parts9 - The heat transfer by convection and radiation across the boundary from the fire into structural membersK - The heat transfer by conduction within structural members# The net flu0, :c, due to convection applied to a surface of the element is given by3H59

: L hc *Tg/Tf+

*,Euation 4+

=ere, hc is the coefficient of heat transfer by convection and Tg and Tf are the gas and suface temperatures respectively# A common assumption used in thermal analysis of the structures is that the net radiant flu0 :r applied to a surface of the element is related to the gas temperature Tg and Tf by the following formula3H59

T< :r L g s **Tg+ </* Tf+<+ : =

*,Euation &+

=ere, is the configuration factor3 g, f are the gas and surface emissivity respectively and is the !tefan/BoltGmann constant *>#CH 0 4'/( WDm&%<+#

!urface to surface flu0 e0changes can also be calculated# Thus, given a spatially uniform enclosure temperature and a time/temperature curve for the thermal environment of the enclosure is specified, and attention can be confined to the calculation of the temperature and stress distribution in the structural elements# )or the calculation of the configuration factor a method is given in anne0 ; of ,urocode 43H5# The surface emissivity is obviously dependent on the type of material e0posed#

7nless the gases are very black, the emissivity value, g, would typically be less than 4 which greatly complicates the problem, as the temperature and optical properties are spatially variable This problem can be addressed by a C) /based approach#

Tabular Data
The thermal analysis in structural members can be e0tremely comple0, especially for materials that retain moisture and have a low thermal conductivity# The simplest method of defining the temperature profile through the cross/section is to use test data presented in tables or charts, which are published in codes or design guides# =owever, these test data are generally of restricted application being based on standard fire conditions# :evertheless, this data might be used in the pro.ectdesign as well as for verification of certain elements of the thermal modelling methodology#

*imple Fire ModelsThermal Design Formulae


!imple design eEuations are presented in codes and design guides to predict the temperature development in bare steel, using a lumped parameter approach# The approach considers both radiative and convective heat transfer and, although a spreadsheet is reEuired to solve the eEuation over the fire duration, it is simple to use# !imilar eEuations e0ist for protected steel sections 34'5K however the thermal properties of the proposed protection material are needed, which can be difficult to obtain#

-eneral .or Advanced/ Models


Computer programs are generally needed to provide a general solution to the governing heat transfer eEuations# 7se of these advanced numerical heat transfer models *e#g# finite element or control volume based approaches+ reEuires the relevant e0pertise to ensure the models are applied correctly and used within their limitations#

The advanced models for heat transfer problems are generally computer numerical models# 2n structural fire engineering problems, most of the time, the heat transfer is in transient/state condition, coupled with time/dependent boundary conditions and temperature/dependent material properties# ConseEuently, realistic solutions to this comple0 problem can only be provided using advanced methods for numerical heat transfer, e#g# ),- thermal analyses# The heat transfer analysis can be performed in one, two or three dimensions# )or the purposes of this pro.ect, it will generally not be adeEuate to model in ' *e0cept for the case of unprotected metal structures+ or in 4 *e0cept for the case of enclosure boundaries+34A5#

The general aspects for the modelling of heat transfer analysis include9

M-01&+2

The shape and dimensions of the structural model are modelled by a finite element mesh of general flow continuum elements, in the form of triangles, Euadrilaterals, wedges, or bricks# The boundary elements or interface elements can be line shaped elements for a & model, and triangular or Euadrilateral elements for a A solid/phase model# =eat sources can be represented by either temperature/time functions or heat flu0 in boundary elements# We generally need both# Convection and radiation at boundaries of the structural model can be modelled by by the heat transfer coefficient of boundary elements and radiation is best treated e0plicitlyimposing a net heat flu0 across boundary elements, which is a summation of the net convective and radiant flu0# The material can be isotropic, orthotropic or anisotropic# The material thermal properties of conductivity, specific heat and emissivity can be significantly temperature/dependent# -oisture evaporation and movement, hydration heat and change in boundary conditions may be modelled# Also intumescent effects may be relevant#

B*3+,(.4 C*+,&)&*+0

M()-.&(% P.*$-.)&-0 S$-'&(% F-()3.-0

The thermal and mechanical response of structures in fire can normally be performed by the computer packages for general finite element modelling, including9
-

ABAI7!

A:!Y! :A!T6A: 2A:A34<5 from T:1 *the :etherlands 1rganisation for Applied !cientific 6esearch+

There also e0ist a number of special/purpose computer programs for thermal fire engineering analyses, including9 - !A)26 from the 7niversity of $i8ge, Belgium *)ranssen, %odur M -ason &'''+ - @7$CA: from the 7niversity of !heffield and Buro =appold 4??(34>5 - )26,!/TA34C5 from :ational 2nstitute of !tandards and Technology, 7!A *2ding, Bresler M :iGamuddin 4?HH+ - TA!,)/& from $und 2nstitute of Technology, !weden *WickstrNm 4?H?+ - T=,$-A from the Building 6esearch ,stablishment *%irby &''&+

Structural Analysis
!tructural analysis is the process of understanding how applied loads on floors, walls or roofs of a building flow through the beams, columns and other structural members to reach the foundations# The building structure resists the applied loads by

deforming slightly under the load# The flow of the loads through the structure is accompanied by deformations and internal forces developed within each structural member# 2nternal forces come in the form of bending moments, a0ial forces or shear forces#

!tructural analysis is used to calculate the deformations of structures under the applied loads and internal forces in each and every member# The structural analysis of simple structures, e#g# simply/supported beams or columns, is carried out by hand

calculations from first principles of statics and mechanics and may often reEuire referencing to standard formulae# -ost structural analysis adopt the assumption that the structure is linear and elastic# The linear elastic assumption is good for most structures at low levels of load# =owever, non/linearities occur in structural analysis when large enough load levels are reached# ;eometrical non/linearities occur when deformations become so large that they induce additional internal loads resulting in larger deformations# -aterial non/linearities occur when materials are stressed

beyond the elastic range causing plastic behaviour with irreversible effects#

!tructural analysis for fire design is essentially the same process as structural design for normal temperature design#, but it is complicated by the effects of elevated temperatures on the internal forces and the properties of materials# )or simple

structural elements e0posed to fire, load carrying capacity can be calculated with simple hand calculation methods34'5344534&5# =and calculations are most

appropriate for single elements with simple supports especially where internal temperatures are uniform or where the temperature of one part of the member is critical# -any tools are available to calculate the structural fire performance of load bearing structures# These can be categoriGed into hand calculations, design charts and special purpose or generic computer programs# The coupling between fire dynamics and structural response in building fires are largely due to radiative heat transfer from combustion products to structural elements# =owever, convection heat transfer cannot be ignored during the early stages of a fire# "urpose/written or commercially/available finite element software could be used to assess the structural failure of more comple0 structures under fire conditions# At the

simplest level ),- codes can be used to post/process the C)

predictions, using

calculated flu0esDtemperatures as boundary conditions for the surfaces of e0posed structural elements# The C) boundary conditions are matched to the ),- mesh by interpolation methods# As with all design and validation methods, the use of the finite element method is an appro0imation of the real behaviour# Before such software is used the designer should be adeEuately e0perienced to identify the assumptions and appro0imations embedded within the software and the assumptions and appro0imations in the actual use of the software# Any software used must be able to model geometric and material non/linearities because deformations are larger and material strength is less than in normal temperature conditions## The first level of assessment of failure is to compare the analysis with the defined acceptable performance criteria# This could include9 - A limit on the ma0imum displacement to ensure compartmentation is maintained, protection of fire/fighting shafts and protection of escape routes - !tability of the structure - A limit on the ma0imum strains in the reinforcement The above limits are easily defined within the analysis# =owever, localised behaviour such as fracture of the reinforcement *if a smeared cracking model is adopted+ or connection failure can be more difficult to Euantify# 2f large displacements are acceptable, an assessment on the likely fracture of reinforcement or connection failure due to high tensileDcatenary forces should be conducted and if necessary more robust details should be adopted to ensure localised failure does not occur# When using finite element models to predict the structural response of a building to a given defined temperature distribution, a sensitivity assessment should be conducted#

This should include a sensitivity assessment on the mesh density and connection behaviour#

"umerical modelling issues


Advanced calculation methodologies should34'59 - Obased on acknowledged principles and assumptions of the theory of heat transferP - consider temperature/dependent material properties - influence of moisture Omay conservatively be neglectedP - be validatedion - on basis of relevant test results - sensitivity analysis on critical parameters

CFD simulations
The most basic consideration in the gas/phase model is that of the grid resolution# This factor should not be underestimated during the analysis and in theory a grid refinement e0ercise using at least three different grid resolutions should be carried out, preferably doubling the grid cells in all directions# )or buoyant flows, special attention should be given to the vertical mesh spacing# 2n practise, this will be a very demanding task but some attempt to e0amine sensitivity should be made# ,0tra attention should be paid to those regions within the C) domain where high

accuracy is important, for e0ample the fire source and near solid boundaries or openings in an enclosure# 2n transient simulations, be sure that the time/step is adapted to the choice of the grid and check the influence of the time/step on the final results# -ore iterations per time/step are generally needed in order to capture rapidly developing fires#

The C) user will have to contend with numerous convergence issues# :ormalised mass and momentum continuity errors, and residual errors for all solved variables should be less than about '#'4, preferably less than '#''4# 2n an ideal simulation the residual should decrease steadily# ,0amine convergence by following the data

during simulations, especially pressure, at critical locations, for e0ample in the fire plume, in the hot layer an in ventilation openings# @alues should gradually settle down to their converged levels and if not a small ad.ustment should be made to the residual error#

*olid0phase simulation
Assumptions, relating to connections, should be checked by carrying out a sensitivity analysis# The stiffness relating to a0ial and rotational movement should be varied and the impact on the overall response assessed# The mesh density of the mechanical model should also be taken into consideration# The same structural analysis should be carried out a number of times with the number of elements J ideally increasing by a factor of two# The procedure should be repeated, until the answers between successive analyses do not vary by more than >Q3C5#

Methodology development to integrate a CFD and a FEM program


#oupling Approach
The core work of the pro.ect "h involves the development of a methodology to

integrate the capabilities provided by models of fluid dynamics, thermal radiation and solid/phase thermal conduction# The methodology calls upon the capabilities of all models collectively to provide a representation of the response of structures e0posed to fire# This is broken down into the following stages of calculation9

4# Calculation of the time/dependant gas temperatures, velocities, density, specific heat capacity and chemical species within the C) containing the fire source &# Calculation of the radiation field i#e# incident radiative intensities A# Calculation of the boundary conditions on the gasDsolid interface i#e# surface temperatures, radiative and convective heat flu0es <# Calculation of the ),- solids*structure+ thermal response *rise in temperaturechange in temperature+ ># Calculation of the subseEuent mechanical response of the structure fluid domain

C)

codes were used for the fire simulation, whilst both ),- and finite volume

codes were used for the structural modelling# The latter is therefore strictly referred to as solid/phase, though ),- is also used as a short/hand form# :o single e0isting model can provide all the features listed above that can be applied to structural fire problems# By employing a combination of C) , radiation and solid/phase models,

with appropriate e0change boundary condition information, the best of all worlds are at our disposal in tackling the comple0 Euestions associated with the thermo/ mechanical response of structures# =owever with the multitude of models involved comes an increase in the number of calculations in a single process, which can subseEuently become too computationally demanding## These calculations may be considered independently, loosely coupled to one another, or tightly coupled into a single calculation# Typically, the C) models addresses stages 4 M & the first two

tasks and the solid/phase model the last twostages < M >#

Types o! #oupling
2n general, predicting the full thermo/mechanical response of a fire/e0posed structures follow three main stages9

4# !imulation of the fire within the compartment &# !imulation of the heat penetration into the structure *i#e# thermal response+ A# !imulation of the mechanical response of the structure

2n general, task 4 could be performed by C) codes Within the scope of the )26,!T67C pro.ect, C) codes were used to carry out step 4, , which are usually accompanied with a radiation sub/model and together perform the C) and thermal radiation calculations in unison# !tep Task & is can be carried

out either by a C) or a solid/phase code and step task A is addressed completely by the solid/phase code# 2n order to bridge the gap between the fire and structural parts of the simulation, boundary condition data must be e0changed between the C) and

solid/phase models# "otential errors are introduced as a result of this process# )or e0ample, due to the different grid systemsnumerical mesh adopted by all codes the

spatial scale of the thermal response models is typically an order of magnitude smaller than the scale of the fire and mechanical model# This in turn makes it practically impossible to achieve an acceptable match in the definition of the structural components between the fire model, where resolving the full geometry in detail can be computationally demanding, and the solid/phase model, which can resolve detailed geometry without greatly diminishingincreasing the computational effort# !tructural deformation, which can affect the development of the fire, cannot be easily accounted for as real/time modifications of the C) geometry must be

achieved# The array of comple0ities has inevitably led to the adoption of many different levels coupling and appro0imations# Three main stages of coupling can be established9 - Coupling between steps 4 and &, where gas temperatures, radiative and convective properties are passed from the fire development *gas/phase+ calculation to the thermal *solid/phase+ calculation# 2n turn, the thermal calculation responds by feeding surface temperatures back into the fire calculation# - Coupling between steps & and A, where temperature distribution within the structural elements is passed from the thermal calculation to the mechanical calculation# ,ffects such as cracks within the structure are fed back to the thermal calculation if a two/coupling approach is adopted# - Coupling between steps 4, & and A, where the effects of the deflection calculated from the mechanical model can interrupt the gas flow calculated in the fire development model# 2n the )26,!T67C pro.ect, various pairings of software packages were used# )or C) simulations @,!TA, !1)2,, JA!-2:, and ) ! were used# )or solid/phase 2A:A, !A)26, !T,$A and A:!Y! were used# 2t is via this "h

simulation

where that the coupling of the !1)2, and !A)26 codes have been established and subseEuently compared with coupled codes from other )26,!T67C partners#

$e%els o! coupling
Coupling the fire development to the thermal model
The interaction between the fire development model *C) + and the thermal response model *C) or ),-+ can be appro0imated at different levels of comple0ity#

Although the interaction between the fire development model and the thermal response model is essentially a back and forth process, i#e# a two/way relationship, a one way approach can be adopted and may prove to be advantageous under certain circumstances#

=ne%/a( and T/o%/a( !oupling 2n simple one/way interaction the fire development is calculated completely independently of the thermal response model# The temperature distribution as

calculated in the fire development is used as input data in the thermal response# The parameters to be transferred are the time and location dependeant gas temperatures, radiative flu0esDintensities and convective heat transfer coefficients# 2nterpolation of data between the different temporal and spatial scales used by the models is carried out# 2n turn, the thermal response model calculates the temperature distribution within the structure using the interpolated output from the fire development model# 1nly parts of the structure that significantly alter the flow characteristics are need to be present in the fire development model, i#e# compartmentation or enclosures, and structural members that impose less of an influence are omitted altogether, for e#g# shallow beams running along a ceilingcolumns that are not completely engulfed by the fire source# 2f the components are present, they could be represented by

isothermal surfaces, adiabatic surfaces or a 4/ simple heat transfer model#

Two/way interaction comprises the one/way interaction together with feedback from the thermal model to the fire# The surface temperature is solved in the thermal response model and fed back into the fire development model# The surface

temperature of the structure is calculated takingen the heat loss through the structure into account# The structure is e0plicitly modelled in the fire development model, but owing to the different grid scales in both models, this is typically represented in less detail than in the thermal response model# epending on the scenario, the surface

temperature of the structural component may or may not influence the progression of the fire# )or e#g#e0ample, if the structural component were large relative to the fire, it may have a large impact on the gas/phase temperatures, or, as is often the case, the component could be remote from a localised fire and the influence would be negligible#

=ence, many different levels of coupling can be distinguished# The following figure *)ig# A+ provides a diagrammatic representation of the coupling discussed so far# 1ne of the most fundamental choices, as mentioned above, is whether or not to include the structural members in the C) calculation *i#e# A versus B+# By omitting

these components there are many conseEuent advantages in terms of computational savings and fle0ibility# 2n this case, the C) results can be e0ploited generically in

con.unction with a variety of different structural arrangements in the ),- code, i#e# with no necessity to repeat that part of the analysis# This might be an appropriate strategy when the component is relatively small compared to the overall volume of the computational domain, e#g# when dealing with 2/section beams, but attention must be paid with to effects of elements such as deep 2/beams which may affect the thermal field due to the interference of ceiling .ets and formation of smoke reservoir# 2f the component is included, then there is another fundamental choice concerning the coupling of the thermal response back to the fire, i#e# methods paths C and , for one and two way coupling respectively#

A2 *tructure omitted 1 Fire de%elopment CFD model Gas phase "a%ier Stokes, radiation

+2 *tructure included 2.+/ Thermal response FEM model Solid phase 'etailed radiation !ield 'etailed con%ection calculation

2.A/ Thermal response FEM model Solid phase Appro*imate radiation !ield #orrelations !or con%ection C2 Flo#field disturbance

D2 Flo#field disturbance 3 thermal influence

Figure &: #oupling methods !or !luid #F' and solid(phase F)M simulations o! !ire(e*posed structures+

2n the one/way approach, an some degree of error is introduced in the fire development model as in reality heat will be lost through thermal penetration into the construction or gained where a0ial conduction is significant# This might be

represented appro0imately by applying a range of surface conditions such as isothermal, con.ugate, simple 4/ heat transfer model or adiabatic conditions# 2n

addition, the geometric model of the structural elements in the fire development model will in general be on a coaurser mesh, i#e# neglecting the finer detail such as flange and web dimensions, than in the thermal response model# An advantage is that the simulations of the fire development can be carried out completely independent of the thermal response model and any changes made in the construction do not lead to es not reEuire new simulations of the fire development, provided that the flow field is not significantly altered# 2n reality, this approach could potentially mean that the models can be operated by independent users, as long as a prior

agreement has been established by both parties with regards to the scenario layout and transferred thermal data#

Merged ersus segregated sol er The tightest interaction between the fire development model and the thermal response model*s+ that can be achieved is a merger of the system of eEuations of the fire development model and the thermal response model# 2n this case, both the fire development and the thermal response are calculated within the C) software# The

disadvantage here, however, is the increase of the siGe of the resulting system of eEuations, which maybe too much for one code to handle, taking into account the fact that the thermal responsefire development model uses a much finer grid relative to the fire development modelis solving multiple fields of parameters through out the fire domain, such as gas temperatures, velocity field and chemical species# 1n top of that, the thermal response model would be solving the solid temperatures on a finer mesh relative to the fire development model#

An alternative would be that the system of eEuations remains separate, whereby the C) code and the ),- code take responsibility for the fire development and the The only form of communication

thermal response calculations respectively#

between the two codes would be via the e0change of calculated thermal data at the boundaries# The data can be transferred after each iteration, after each time/step and even after several time/steps# The level of transfer detail transferred taken place is dependant dependent on how both the compartment in the fire development model and the structure in the thermal response model are defined spatiallyspatially discretised#

As the spatial scale of the thermal response model is of angenerally a couple orders of magnitude lower than the scalethat of the fire development model, generally

separate simulations are made for representative parts of the construction, for e#g# beams and columns, rather than one simulation for the heat penetration for the entire construction# 2ntroducing too high a number of cells in the fire development model to sufficiently capture the details of the structural geometry could easily produce a model that isbecome too large to handle#

Coupling the thermal response to the mechanical response


ifferent levels of coupling can also be established# The simplified approach would be a one/way interaction in which the mechanical response model reads the temperature distribution, which is initially calculated by the thermal response model# The data is only transferred in one direction from the thermal to the mechanical model without a reverse process, which means that the temperature distribution can be calculated over the entire time and spatial domain and transferred only once at the start of the mechanical response simulation#

An advanced two/way model would take into account of the effects of deformations and displacements of the construction on the thermal response model# 2nformation regarding the any alterations to the structural geometry, i#e# cracks developing in the compartment walls, will be relayed back from the mechanical to the thermal model so that the latter can register the presence of a potential source of heat loss and carry out appropriate calculations to account for this# =owever, this advanced level of coupling will not be covered in the work as it is difficult to predict the growth of a crack or gap during a fire in a reliable and accurate way# There is also the issue of the lack of e0perimental data regarding the effect of crack width and gap openings on the thermal response# =ence, only a one/way coupling of the thermal response model to the mechanical model has been considered#

Coupling the mechanical response to the fire development


This type of coupling addresses the issue of how for e#g#e0ample the deflection of a roof may alter the flow pattern of hot gases and smoke within a fire enclosure# To account for this effect, a C) program with a grid system that possesses the ability

to model nodal displacement is called uponnecessary for this task# The downside is that not all fire C) programs are capable of dealing with non/Cartesian grids as the ma.ority possess structured grid systems# The alteration of the mechanical properties of the construction necessitates an updated fire development simulation per grid change, making this type of coupling an impractical approach for design problems# =ence again, such a coupling will not be considered further in this pro.ect# =owever, for C) codes that can handle non/Cartesian grids, an interface file can potentially

be constructed matching the nodes from the mechanical model to the nodes of the fire development model# The C) grid shape would be updated in the thermal

response model at every iteration, each time/step or after several time/steps# The mechanical model will be updated with the subseEuent thermal response output#

Types and levels of coupling established in the F('E*T'UC pro4ect


2n order to investigate coupling methodologies, a number of code pair links were established for software packages that are used on a regular basis by various party members of the )26,!T67C pro.ect# JA!-2:,, !1)2,, @,!TA and ) ! were used for the fire simulation and 2A:A, A:!Y!, !A)26 and !T,$A were used for The main purpose of this "h is to

the solid/phase simulations 34H534(53&&5#

demonstrate e0amine the methodologiesy of coupling rather than to show how well a C) and solid/phase code can model fires and subseEuent structural

thermalDmechanical responses respectively# The specific code pair and coupling assessed is set out in the Table & below# )or pairings, the C) code is stated first

followed by the solid/phase code is on the right# All the solid/phase codes are based on the finite element method with the e0ception of !T,$A, which employs the finite volume method#

Type of coupling

-erged JA!-2:,/!A)26 )ire development to !1)2,/!A)26 @,!TA/ 2A:A JA!-2:,/!T,$A thermal response !1)2,@,!TA/!A)26 ) !/A:!Y! !A)26*4+ Thermal response to 2A:A*4+ :one none mechanical response A:!Y!*4+ T(/%- # Types and levels of coupling established in the )26,!T67C pro.ect3&&5 4+ ,0isting technology before )26,!T67C

1ne way

$,@,$ 1) C17"$2:; Two way !egregated

Methodology -ierarchy
The study of the coupling methodologies has focused on the implementation and testing of a representative set of coupling approaches in a hierarchy labelled A/), as summarised below# 2n all levels of coupling, information is converted by the C) and solid/phaseD),- codes, and passed between each other# The methods differ in the type of data e0changed and subseEuently the way the data is interpreted and converted by each code#

T/o%/a( #oupling (Methods A and >" Tight coupling of the fire development model and the thermal response model is achieved when via a two/way segregated approach, as mentioned earlier, which is demonstrated by -ethods A and B# These methods are based on the use of pre/ calculated interpolation factors for the conversion of the thermal data, generated before the start of each C) and ),- iteration# These methods are Euite similarK

however, in -ethod A, effort on interpolations and radiation calculations is shared between both codes, whereas all the work is done in the C) code in -ethod B# A

flow/chart diagram of method B is provided in appendi0 A# The implementation of -ethod B is described as follows9

4# The C) code generates a file with a description of the surface mesh of the C) solids in a specified format# The surface mesh is made up of so/called C) faces, with each face representing one mesh node# ,ach face is described by the position of its vertices# &# The ),- code generates a file with a description of the surface mesh of the ),solids in the same format as used to describe the C) solids#

A# A program called CT2:" computes the reEuired conversion factors for the C) code# The conversion factors *interpolation factors and view factors+ are written to a file in a specified format# <# 1n initialiGation, the C) code reads the file with the conversion factors# 5. The C) code computes, after each C) iteration, the values of the

environmental gas temperature, T3en 3-, and environmental heat transfer coefficient,

en -, for each surface mesh node - of the ),- solid# The calculation is based on
the gas temperature, Tg3i, the convective heat transfer coefficient, !3i, and the incident radiation flu0, :g3rad3i, on each surface mesh node i of the C) solid

using3 the pre/calculated conversion factors# The values are written to a file in a specified format# 6. The ),- code reads the values of Ten 3- and en 3- before each ),- iteration and applies these values as boundary conditions for the temperature computation# H# After each ),- iteration the ),- code computes the values of the surface temperatures, Ts3i, for the C) solid based on the conversion factors# The values

are written to a file in a specified format# The values of the emission coefficients of the surfaces are also written to the file to account for temperature/dependent emission coefficients calculated by the ),- code# (# The C) code reads the values of the surface temperatures, Ts3i, on the C) mesh before each C) iteration and applies these values as boundary conditions for the flow computation# 2n -ethod B interpolation factors are read by the C) code to convert calculated

gas temperatures, convection coefficients and incident radiant flu0es from the C) mesh to the ),- mesh and also to convert receiving surface temperatures from the ),- mesh to the C) mesh#

=ne%/a( !oupling (Methods #%F" $oose coupling is achieved when the fire development model and the thermal response model is linked together in a one/way segregated approach# 2n some cases, structural elements can be introduced as either simplified heat transfer ob.ects or .ust as flow blockages if the elements are seen as a potential influence to the fire development# The inclusion of the element can account for shadow effects and the disturbances in the flow of hot gases# 2n some one/way methods it may be possible to neglect the structural elements altogether during the C) analysis# -ethod C is a subset of -ethod A but without the feedback of data to the gas/phase# A flow/ diagram illustrating -ethod C is provided in Appendi0 B# The method consists of the following steps9

4# The fire development is calculated by the C)

code neglecting the structural

elements and assuming certain thermal boundary conditions for walls, ceiling and floor# &# 2n the ),- code each structural element is modelled as a O),- profileP# A ),profile may consist of several segments with the same cross/section connected to each other along the a0is of the profile# ,ach segment is assumed to have infinite conductivity and therefore a single uniform temperature# A# The geometric description of all ),- profiles is written to a file# The description includes the specific location of the ),- mesh nodes on the a0is of each profile, the orientation of each profile and the shape of the cross/section of each profile# 2t also contains the absorptivity of the elements# 4. The C) code *or C) postprocessor+ reads the file with geometric data made in step A and converts the thermal results of the whole *time/dependent+ C) run to the specific locations of the mesh/nodes *segments+ of each ),-/profile# The converted thermal results consist of Tg3-, !3-, and :g3rad3-#

># The ),- code reads the file with thermal data made in step < and computes the time/dependent temperatureDmechanical response using this data as a boundary condition#

=owever, in -ethod C geometric data, in terms of specific location of the ),mesh nodes of each structural profile and the orientation and cross/section of each profile are passed to a C) post/processor from the ),- code# With prior

knowledge to the geometry at the stage of post/C)

analysis, gas temperatures,

incident radiant flu0es and convection coefficients are calculated and passed on to and read by the ),- code#

-ethod

is an e0tension to -ethod C and is also based on passing component

geometry information in the form of ),- profiles at specific locations to the C) model from the ),- model# 2nstead of transferring the thermal data from the C) code to the ),- code using a formatted A!C22 file, the thermal response calculations are carried out via a simple thermal module embedded within the C) code# As a result, it is the temperatures of the solid structural elements that are passed directly on to the structural ),- model#

-ethod , is similar to -ethod C, but instead of calculating the incident radiant flu0es in the C) code, this is carried out by the ),- code using the directional code# The cross/section and

radiant intensity values transferred from the C)

orientation of the structural profile is not reEuired by the C) code as the location of the profile is enough# A flow/diagram illustrating -ethod , is presented in

Appendi0 C# The method consists of the following steps9

4# The fire development is calculated by the C)

code neglecting the structural

elements and assuming certain thermal boundary conditions for walls, ceiling and floor# &# 2n the ),- code each structural element is modelled as a O),- profileP J when reEuired in more detail as in -ethod C, i#e# with finite heat conductivity# A# The specific location of the ),- mesh nodes on the a0is of each profile is written to a file as so/called O),- linesP# 4. The C) code *or C) postprocessor+ reads the file made in step A and converts the thermal results of the whole *time/dependent+ C) run to the specific

locations of the mesh/nodes *segments+ of each ),- line# The converted thermal results consist of the gas temperature, Tg3-, the convective heat transfer coefficient, !3-, and directional radiance values, )53-, on each mesh/node -# ># The ),- code reads the file with thermal data made in step < and computes the time/dependent temperatureDmechanical response using this data as boundary condition# The ),- code must be able to handle the shadow effect when appropriate#

=owever, the amount of thermal data to be transferred to the ),- code is much less in -ethod C, with .ust three variables transferred per mesh node, while in -ethod , depending on the number radiant intensities, :2, per mesh node the values of *& F :2+ variables have to be transferred#

2n -ethod ) the same thermal data as in -ethod , is interpolated on to a coaurse C) mesh, or at targeted sampling points, and saved in an e0change file accessible

by the ),- code# These sampling points can be positioned at every cell or second cell along in all three directions within the C) domain# 2nterpolating the data on to a coarser mesh reduces the siGe of the e0change file, however, data is lost in supplying radiative intensities to ),- mesh, which can subseEuently reduce the

accuracy of the radiative flu0 calculations# 2nterpolation of the thermal data is carried out entirely by the ),- code# The thermal data transferred are gas

temperatures, convective coefficients and direction radiant intensities per course coarse mesh node# The fle0ibility of -ethod ) is that it saves storage memory because the original C) solution can be deleted once the data has been interpolated on to the courser coarser mesh#

The couplings implemented for the code pairs from the, listed in Table &,9 !1)2,/ !A)26 are based on the one/way procedure of -ethod )K the ) !/A:!Y! and JA!-2:,/!T,$A couplings use -ethod ,K @,!TA/ 2A:A coupling is based on the two/way coupling -ethod B#

.ne(/ay S.F0)(SAF01 coupling


1ne/way coupling has been established to study how two programs that have been developed under different engineering disciplines can be linked and moreover to investigate the specific method of coupling known in this work to be -ethod ), as described previously#

-eneral methodology of one0#ay coupling to *AF('


The two programs must have a common understanding of the system of co/ordinates in which the fire compartment and the structural components involved have to be described# The simplest approach is for both programs to adopt the same system of co/ordinates# 2t is also possible for the two programs to use different system of co/ ordinates as long as sufficient information has been passed on to the ), user as to

what system has been used in the C) analysis, in order to define the translation and rotation to be applied to each position in order to obtain the same position in their own system#

)rom the fire development point of view, the C)

user makes the numerical model

with all the choices about mesh refinement, boundary conditions, time/steps, turbulence model, combustion model, radiation model and the eventual presence of the structure or part of it# The model is run and the results are verified# 1ne piece of information from the ), analysis that the C) user must have is the order of

magnitude of the fire resistance time that is e0pected for the specific fire scenario# This crucial piece of information should be provided by the ), user, as for e#g# the C) user may be use to analysing the response of heat sensors that activate within

minutes after the ignition of the fire, so the ), user should make aware to the C) user that the structural fire resistance will be in the range of a couple of hours# The definition of the fire scenario must also involve the ), user because the critical fire scenario may be different when it comes to the fire resistance of structure than when the investigation deals with smoke movement# This flow of information from the ), user to the C) user is crucial from the beginning before the C) analysis is carried out as it has a strong influence on the outcome of the thermal response calculations# 2mportant as it is, this type of information can easily be e0changed in a simple conversation or via electronic mail#

Transfer File
The transfer file is written by the C) user via a post/processor, after the C)

analysis has been carried out# This file could be written automatically by C) program during the analysis, but it is preferable that the C) user first runs the C) simulation and analyses the results because the user is then in a better position to

decide at which time/step and location the C) the transfer file#

thermal data should be passed on to

The time/steps that are present in the transfer file can be separated byhave a much longer duration than the time/steps used in the C) analysis# The C) user should

decide whether an interpolation in time that will be made by the ), user based on the information given at certain time/steps be sufficient to represent the evolution of the situation in the fire compartment# A safe assumption would be that the ), analysis would adopt a liner interpolation of time# The fire scenario has thus to be

represented or, in other words, written into the transfer file at or around the time/step that marks a significant transition in the conditions in the compartment, for e#g# before and after the occurrence of flashover, at the time when the fire intensity starts to decrease# The presence of a structure within the compartment does not influence this decision#

The spatial structure of the points at which the information is written is governed by the C) results and hence must be chosen by the C) user, for e#ge0ample# the

location of the fire source or the eventual interface location of the hot gas layer# The mesh density of transfer points is normally much coaurser than that of the mesh reEuired in the C) analysis in order to reduce the siGe of the e0change file# The

transfer points in the transfer file follow the global system of co/ordinates used by the C) program#

The information that has to be present in the transfer file is as follows9 - Coordinates of the transfer points with respect to the C) system irections of the radiative intensities, in the form of directional vectors, at every transfer point global coordinates

- Time steps at which the radiative intensities are given - At each time/step per transfer point9 the gas temperature, the radiant intensity in the relevant directions and a convective heat transfer coefficient, which can assumed to be a fi0ed value# The set of radiative intensities are calculated by post/processing the stored C) solution data, using the stored gas

temperature, radiation absorption coefficient and solid surface temperature, which are the temperatures of the main building compartment boundaries derived by the C) programRs wall heat transfer model# The convective heat transfer coefficient may be estimated by using a suitable correlation of the stored C) velocity field#

2n the )26,!T67C pro.ect, a specific format *table A+ of the transfer file has been defined for the purposes of the one/way coupling method# 2f this format is followed, this allows the combination of any the !1)2, C) ), programcode# program code to the !A)26any

The locations of the transfer points in the compartment, as shown in the following figure, must allow for sufficient spatial interpolation to be carried out in the ), analysis# 2t is therefore preferable to have points to cover the most e0ternal positions of the compartment, i#e# on the internal surfaces of the compartment boundaries as illustrated below, so that every point of the structure is surrounded with transfer points and interpolation can be carried out#

beam included only in solid(phase model


(not included in CFD or post-processing analysis) (surface fluxes are calculated by solid-phase model using radiant intensities and directional ectors pro ided by the post-processing analysis)

5 , y

Fire

post(processing intensity 2point3 elements


(radiant intensities transferred to the solid-phase model) (may need to interpolate (in solid-phase model) data from >1 points)

Figure 4: A sample o! locations o! trans!er points /ithin the compartment

The number of directions of radiative intensities per transfer point cannot be underestimated, especially if the fire source is localised# 2t has been found,This is and discussed further in the validation section, that the number of directions does not have to be in the order of magnitude of 4'', yet the minimum number of C is simply not sufficient to capture the radiative field#

Convection heat transfer coefficient estimation The coupling between fire dynamics and structural response in building fires are largely due to radiative heat transfer from combustion products to structural elements# =owever, convection heat transfer cannot be ignored during the early stages of a fire# The convective heat transfer coefficient, dependent on the local properties of the flow i#e# the velocity field, is calculated per transfer point per time/ step of the C) in the C) analysis# But as the structural member is not e0plicitly represented

analysis, the local flow is not realistic# Certain means have to be taken

up in order to appro0imate the convection coefficient, which is heavily dependent on the conditions in the thermal boundary layer, with the lack of a surface# Therefore, a simple assumption is used, knowing that in many cases, the radiant intensities play a

much more significant role in the heat transfer to the member#

epending on the

emissivity and the value of the convective heat transfer coefficient, convection predominates at low temperatures *below &''oC+ but it is not until <''oC when radiation becomes increasingly dominant3?5# The contribution of convective heat transfer to the steel temperature can be estimated from a correlation that involves the :usselt number, 7u# With knowledge of this dimensionless parameter the convection coefficient h! may be found and the convective heat flu0 calculated# The :usselt number is eEual to the dimensionless temperature gradient at the surface and gives an estimated measure of the convective heat transfer occurring at that surface3&A59

7u

5 f T S h6 =+ 5f 6 ( S

*,Euation &+
( S ='

6earranging the above eEuation gives h9

5 7u h = f 6
!

*,Euation A+

where 6 is the steel characteristic length and 5f can obtained from the "randtl number3&A59 9r = #p 5f *,Euation <+
*

The A general e0pression for forced convection coefficient over a flat plate, dependeant on whether it is a laminar or turbulent flow, due to Jakob M h! = '#'AH 5 6e<D>"r4DA *6eT>04'>+ h! = '#CC< 5 6e4D&"r4DA 6 6 and Johnson M 6ubesin *4?<?+, is the following respectively3&<53&A59 ow *4?<C+

*6eU>04'>+

h! = '#CC< 5 6e4D&"r4DA 6

*6eT>04'>+

*,Euation >+

h! = '#'AH 5 6e<D>"r4DA 6

*6eU>04'>+

*,Euation C+

where 6e is the dimensionless 6eynolds number that determines if a flow is laminar or turbulent3&A53&A59 *e =

,6

*,Euation H+

thus,

h! = #p 7u 6 9r

*,Euation (+

There are several options when it comes to deciding a value for the steel characteristic length, 6# 2t can range from the thickness of the steel flange to the length of the steel member with orders of magnitude difference# =owever, it can be seen that the above eEuation implies that smaller length scales would generate bigger convection coefficients and since the flow over the members are not e0plicitly modelled a compromise has been made# The proposed appro0imation would be to take the width of the steel memberRs cross/section as the characteristic length#

The lack of an element also means that the flow field is left undisturbed, when supposedly a structural element is present, during the C) analysis# =ence an

appro0imation of the velocity per transfer point is made# The characteristic velocity, if an element surface is present, is given by9 , L *u$ F $+0.5
5 5

*,Euation ?+

# u v .a/ y y , u v .b/ ,

Figure 5 6 7elocity %ectors used /ith2a3 and /ithout2b3 a sur!ace

When element,

there the

is

no three , L *u$ F
$

F /$+0.5

*,Euation 4'+

components of velocity are taken into account9

6adiation transfer eEuation solution The radiative intensity distribution along the discretised ray path towards a transfer point is solved with the recurrence relation3&C59 )n?1 = ) e/al F T .*4/e/al+ *,Euation 44+ n g The initial value of each ray is evaluated at the originating surface element, i#e# the C) domain boundary, and is given by3&C59 )0 = :? *,Euation 4&+

with

:? = Tf.F *4/+ :%

*,Euation 4A+

The surface emissivity, , solid surface temperature, Tf, and incident heat flu0, :/ are all e0tracted from the stored !1)2, C) solution#

The radiation transfer recurrence eEuation, as above, can simply be solved with a spatially OlumpedP absorption coefficient, a, with an e0act ray path length, l#

L 4 J e/al

*,Euation 4<+

The main most detailed !1)2, models for solving the same eEuation are based around a representation of the absorptionDemission characteristics of the gases and

soot known as Omi0ed grey gas modelP# The basic concept of such models is to represent the gas properties in a series of bands, 5# The models of interest generally have A or < bands, which might be taken as clear gas, optically thin, optically intermediate, and optically thick, though the introduction of soot can double the number of bands reEuired# 2n the most detailed model, the banded weighted sum of grey gases *W!;;+, the radiative transfer eEuation is evaluated on a band by band basis using the corresponding banded absorption coefficients and the e0act path length of ray passing through each cell# appro0imated by W!;;9 2n this case, the emissivity can be

/ (T)
kL'
5

(1 e%a5(T"l+

*,Euation 4>+

where /5 and a5 are the weight factors and banded absorption coefficients respectively# The weight factor is provided to determine the proportion of black/ body radiation actually emitted in this band, which is dependant on the medium temperature# This fThe weight factor and banded absorption coefficient is are

represented as a polynomial function of the local temperature# A collection of the polynomial coefficients for varying mi0tures of water vapour, carbon dio0ide and soot has been given by various authors such as Truelove, Taylor M )oster and !mith3&>5# Applying this detailed emissivity into the radiation transfer eEuation, as mentioned above, gives the following9

*)n?1+5 =) e/a5l F / (T ) T .*4/e/a5l+ n 5 g g

*,Euation 4C+

The total intensity per point, n, along the ray is simply calculated by summing the banded intensities at each point9 )n L

*) +
kL'

5 n

*,Euation 4H+

A study has been conducted to compare the lumped absorption model against the banded absorption model# etails are discussed further in the validation section#

The previous tasks until the writing of the transfer file were part of the C) analysis# The ), analysis starts by reading the transfer file# The !1)2,/!A)26 one/way coupling methodology has e0ploited the interpolation work carried out by "rofessor Jean/-arc )ranssen and his colleagues from the 7niversity of $i8ge#

Cartesian interpolation
The locations of the transfer points have been chosen in the compartment by the C) user# The ), user needs to interpret these points on other locations in the ), coordinate system, depending on the position of the structural component in the compartment# These locations reEuired by the ), analysis may be at the surface of a beam member or as an appro0imation, on the a0is of the member# The interpolation has to be carried at every time/step and for every piece of information stored in the transfer file, i#e# gas temperature, radiant intensities and convection coefficients# 2t is called Cartesian interpolation because, generally, the global system of coordinates is a Cartesian system# When the Cartesian interpolation has been done, the gas

temperature, radiant intensities and coefficient of convection are known at each structural point of interest#

*pherical (nterpolation
The gas temperature and the coefficient of convection are scalars# They can,

therefore, be used at the structural points of interest once the Cartesian interpolation has been carried out# The radiation intensities come with directional vectors defined in the Cartesian system of coordinates# These intensities have to be integrated at the surface of the structure in order to calculate the impinging heat flu0# The integration

is done on the hemisphere defined by the e0ternal normal perpendicular to the surface of a component# The eEuation of this integration is9 EL

) !os
i
i i

*,Euation 4(+

A certain number of directions, i, have to be selected and the radiant intensity in these directions )i, must be evaluated# These directions are e0pressed in a local system of coordinates that has to be placed at the structural point, that being on the surface of the structural element#

The directions in which the intensities have to be evaluated for the numerical integration are not the directions in which the intensities are given at the structural points after the Cartesian interpolation# A new interpolation has to be performed in order to find, from the intensities given in the directions of the transfer file, the intensities in the directions reEuired by the numerical integration at each structural point#

The integration has to be repeated at each time/step of the thermal calculations but, for each structural point, the direction of the surface system of coordinates will not change if the thermal calculation is made in the unchanged configuration of the structure# These coefficients of interpolation can, therefore, be calculated only once at each structural point if the information is stored to be retrieved at each time/step#

2t has to be noted that, generally speaking, the surface system of coordinates is not parallel to the global system of coordinates# A rotation of a0es may eventually be reEuired in order to find, from the direction of the longitudinal a0is of the structure and from the shape of the cross/sectinal area of the element, the orientation of surface system of coordinates# A rotation of a0es is also reEuired to e0press the directions of the transfer file in the surface system of coordinates# 1nly when these

two types of rotation have been carried out can the spherical interpolation be performed#

The spherical interpolation is made by the Cssgrid *Cubic !pline !phere ;62 er+ solver from the :;-AT=34?5 library based on the work of 6obert 6enka and his packages !T62"AC%3&'5 and !!6)"AC%3&45# The method uses a tension spline interpolation algorithm to fit a function to input dataK the can be specified on a randomly spaced set of latitudeDlongitude coordinates# The spherical interpolation is made, for each structural point, according to different orientations of the local system of a0is depending on the orientation of the surface of the section#

*tructure of the transfer file


)or use in one/way !1)2,/!A)26 coupling, as well as other code pairs that uses -ethods C/), a generic interface file specification has been established# The file is not formatted, which means that any format can be used for e0amplee#g# A#>,F'A as well as A>''# ,ach piece of data in the file has to be separated from other data by a separator, at least one blank character#

At the beginning of each data block a heading or a keyword in capital letters is used# 2t is possible to add comments after the heading as long as it is separated by a space# The comment must be written on the same line as the heading and cannot go beyond the end of that line#

The structure of the file3&&5 is given below via a hypothetical e0ample# 2t is given for the general case where heat flu0es are calculated by the C) analysis as oppose

to the ), analysis calculating them from the post/processed radiative intensities, which is also shown in the e0ample# This is .ust to show that this e0change file can accommodate one/way coupling methodologies from C to )# :) or :2 can be given the value of Gero if reEuired in the transfer file if information about the flu0 or intensity respectively is not available#

)2$,:A-, Cfd#t0t :!T,"! C T2-,! A'# testVe0ampleVA's#simd C'# testVe0ampleVC's#simd ?'# testVe0ampleV?'s#simd 4&'# testVe0ampleV4&'s#simd A''# testVe0ampleVA''s#simd C''# testVe0ampleVC''s#simd

;ives the name of the file ;ives the number of time/steps in the file $ists the times where the flu0es andDor intensities are given, optional name of the files after each time.

:) &(( BYWV)$7B,! 4'#('> 4<#( 4# '# /4# '# 4'#? 4<#( 4# '# /4# '# 44# 4<#( 4# '# /4# '# 44#4 4<#( 4# '# /4# '# 44#4?> 4<#( 4# '# /4# '# ### :" 4<< BYWV2:T,:!2T2,! 4'#H 4<#H& 4# 4'#? 4<#H& 4# 44#4 4<#H& 4# 44#A 4<#H& 4# 4'#H 4<#?& 4# 4'#? 4<#?& 4# 44#4 4<#?& 4# 44#A 4<#?& 4# 4'#H 4>#4& 4# 4'#? 4>#4& 4# 44#4 4>#4& 4# 44#A 4>#4& 4# ### :2V 26,CT21: &' BYWV 26,CT21:! '#'''' '#A>C( '#?A<& '#'''' /'#A>C( /'#?A<& '#'''' '#A>C( /'#?A<& '#'''' /'#A>C( '#?A<& '#?A<& '#'''' '#A>C( /'#?A<& '#'''' /'#A>C( ### T2-, A' )$7B,! A''# 4'# &>''# A''# 4'# &>''# A''# 4'# &>''# A''# 4'# &>''# A''# 4'# &>''# A''# 4'# <''# A''# 4'# <''# ###

;ives the number of transfer points where the flu0 is given *:) points+ :) lines, each containing the B, Y, W location of a transfer point and the components *B, Y, W+ of the outward normal to the surface in the ;lobal system of coordinate#

gives the number of transfer points *:" points+ where the radiant intensities are given :" lines# ,ach line gives the location B, Y, W of the transfer point where the radiant intensities are given#

;ives the number of radiant intensity directions *:2+ :2 lines# ,ach line contains B, Y, W component of an incident radiant intensity direction viewed from point XoutwardsX#

;ives the time/step where the subseEuent values of flu0es and intensities are given# This block and the two blocks that follow )$7B,! and 2:T,:!2T2,! are repeated :!T,"! time# :) lines# ,ach line contains gas temperature in %, convection heat transfer coefficient and incident radiation flu0#

2:T,:!2T2,! A''# 4'# 4&># HC'# ### 4&># A''# 4'# 4&># HC'# ### 4&># A''# 4'# 4&># HC'# ### 4&># A''# 4'# 4&># HC'# ### 4&># A''# 4'# 4&># HC'# ### 4&># A''# 4'# 4&># HC'# ### 4&># ###

:" lines# ,ach line contains gas temperature in %, convection heat transfer coefficient and :2 incident radiant intensities#

Table & 6 Generic inter!ace !ile !ormat !or one(/ay coupling[18]

6alidation of methodology

integrated

@alidation of the !1)2,/!A)26 one/way coupling methodology described above has been performed via a couple of modelling e0amples# This analysis has included the comparisons of predicted and e0perimentally measured temperatures and displacements of a steel beam and column structures e0posed to different wood crib fire scenarios# The validation stage also provided an opportunity to compare the performance of different coupled C) /),- codes with respects to modelling a hypothetical benchmark fire scenario involving a selection of e0posed steel columns and beams# This is reported directly below along with the following section that reports the work undertaken to model the e0ternal column e0periment#

9enchmark case modelling /ith coupled #F'(F)M methodology


Thermal response calculations were carried out for a series of steel tubes inside a rectangular compartment A'm by &'m in floor area with an (m high ceiling, as illustrated in the following figure9

Figure 6 6 Solid and transparent %ie/s o! the hypothetical benchmark case

A fast growing fire reaching A'-W in ('' seconds was specified, as illustrated in figure 44, so that a hot smoke layer would develop under the entire ceiling and a cold clear layer would be maintained above the ground# The performance of the !1)2,/ !A)26 coupled methodology was analysed with respects to beams immersed in the hot ceiling layer and columns partially engulfed in the hot smoke and heated lower down by radiation both from the fire and from the above smoke layer# Adiabatic conditions were applied to the compartment boundaries so that radiation reflected from these surfaces could also act as a means of heat transfer#

!imulations were carried out on hollow column and beam elements# different steel sections modelled are given below9

etails of the

- A A'm long beam with a hollow section of >''mm height 0 &''mm width 0 4mm thickness - A &'m short beam with a hollow section of 4'''mm height 0 <''mm width 0 4mm thickness - A (m sEuare column with dimensions <''mm depth 0 <''mm width 0 4mm thickness The section thickness of 4mm was assumed in order to eliminate the transient heating effects# The column and beams do not touch, so that heat transfer via conduction from one steel member to the other is eliminated# There was also no heat transfer in the form of convection and radiation /ithin the hollow section as the internal surfaces were also set to adiabatic#

Two fire locations were considered as follows9 - $ocation A9 )ire is furthest away from the pair of column - $ocation B9 )ire is closest to the pair of columns and is placed under the centre/line of the long beam#

The following diagram gives a summary of the compartment geometry, steel member dimensions and fire locations# 2t only acts as an indication of where the relative steel elements are placed within the compartment for the C) user, since the !1)2,/

Figure : 6 ;lan and side ele%ation o! benchmark case /ith locations o! !ire source, columns and beams[15]

!A)26 one/way methodology will omit these during the C) e0plicitly represent them during the solid/phase calculations#

calculations, but will

!teel temperatures were calculated at various locations along the columns and beams as shown in the figures (, ? and 4'#

6esults were generated by one/way coupling using the C)

program !1)2, and

solid/phase ), model !A)26# The )26,!T67C pro.ect provided the opportunity to compare the outcome with the results generated by two/coupling using @,!TA and 2A:A and by one/way coupling with ) ! and A:!Y! and JA!-2:, and !T,$A#

a3

f3

a4
c d Am b e Am a f &m
FI RE

f2

f4

a1

f1

Figure < 6 ;lan and side ele%ation o! benchmark case /ith locations o! !ire source, columns and beams[15]

<m

h (m 4>m &'m i .

g2

g3

g1

FI RE

Figure 8 6 ;lan and side ele%ation o! benchmark case /ith locations o! !ire source, columns and beams[15]

k >m

>m

k2

k3

k1

FI RE

Figure 1= 6 ;lan and side ele%ation o! benchmark case /ith locations o! !ire source, columns and beams[15]

,urocode34'5 properties were assumed for the steel, and the surface emissivity was '#H# The fire was to be modelled as burning heptane, with a soot yield of '#'((# The heat release rate, as mentioned previously, grew to A'-W as shown in )igure 4'#

&5 &=

-eat 1elease 1ate 2M>3

5 = 15 1= 5 = = == 4== 6== <== 1=== 1 ==

Time 2seconds3

Figure 11 6 -eat release rate o! benchmark !ire

)or one/way coupling where the beams and columns were not included in the C) the convective heat transfer coefficient was given by the following correlation9 hc L ScpS:usselt numberDcharacteristic steel lengthD"randtl number The gas density at the location of the solid surface and the gas speed at the surface were introduced into the coefficient via the calculation of the dimensionless :usselt number#, shown previously in eEuation &#

'esults
The results presented in the following two figures show the temperatures of the steel members e0posed to )ire A at specific locations both within and below the hot layer calculated by the different code pairs mentioned previously in Table &#

4==

Temperature 2o#3

&==

S.F0)(SAF01 ( g& S.F0)(SAF01 ( l& ?ASM0")(ST)$A ( g& ?ASM0")(ST)$A ( l& 7)STA('0A"A ( g& 7)STA('0A"A ( l& F'S(A"S@S ( g& F'S(A"S@S ( l&

==

1==

= = == 4== 6== <== 1=== 1 == 14== 16== 1<==

Time 2seconds3

Figure 1 6 ;redicted steel temperatures in hot layer at locations g& and 0& o! long and short beam respecti%ely
4== &5= &== S.F0)(SAF01 ( ! S.F0)(SAF01 ( !4 ?ASM0")(ST)$A ( ! ?ASM0")(ST)$A ( !4 7)STA('0A"A ( ! 7)STA('0A"A ( !4 F'S(A"S@S ( ! F'S(A"S@S ( !4

Temperature 2o#3

5= == 15= 1== 5= = =

==

4==

6==

<==

1===

1 ==

Time 2seconds3

Figure 1& 6 ;redicted steel temperatures belo/ hot layer at locations ! and !4

Both two/way and one/way give generally reasonable results albeit with some obvious differences that are apparent in the steel temperatures predictions# 2n the hot layer, figure 4&, at locations gA and lA the differences are very small with the temperatures of the two/way method e0ceeding the temperatures of the one/way method# 2n each case, besides !1)26/!A)26, the temperatures at gA are higher than

those of at location lA, which is e0pected as location gA is closer to the fire# Temperatures of both gA and lA calculated from the !1)2,/!A)26 method seem to produce identical results# The reasons for the differences between code pairs are

mainly attributable to variations in the underlying gas phase temperature predictions# The following figure *fig# Y+ illustrates the differences in the gas temperature predictions at locations gA of the benchmark test produced by the C) JA!-2:, and !1)2,# codes

5==

4==

Temperature 2 o#3

&==

==

1== ?ASM0") S.F0) = = == 4== 6== <== 1=== 1 == 14==

Time 2s3

2n the cold layer, figure 4A, at locations f& and f<, a bigger difference is observed between the coupled codes# =ere again, the two/way method produces higher steel temperatures than those of one/way and the temperatures at the opposed surface of f& is generally higher than those of f<, which is e0pected# The sensitivity to the e0posure differences varies between the code pairs, as the ) !/A:!Y! results are

shown to remain high yet the !1)2,/!A)26 results remain low# The reasons for these discrepancies may lie in the capabilities of the thermal model to redistribute the energy within the solid or in the differences in the way the C) radiation# codes account for

The respective one/way methods are distinguished by a fundamental

difference in this treatment, with incident radiative flu0es to solid surfaces computed directly by the C) radiation model in -ethod ,, but more ambitiously, a field of

directional radiative intensities passed on a mesh of sampling points in -ethod )# The former approach may generally be more robust, but with the latter the storage of memory is saved#

*ensitivity studies
The effect of the resolution of the directional information was e0amined, from &< to 4'' rays were found to give acceptable results in the !1)2,/!A)26 coupled method# 2t is observed in the following figure that C rays are .ust not sufficient in solving the radiation field#

&==

5=

Temperature2 o #3

==

15=

1== 1== rays 4< rays 4 rays 1 rays 6 rays = == 4== 6== <== 1=== 1 ==

5=

Time 2seconds3
Figure 14 6 Temperatures at location ! dependant on the number rays

A study was undertaken to look at the differences in the treatment of radiation in terms of absorption coefficients# The following three figures illustrate the influence the two different radiative transfer models have on the steel temperatures at different locations within the fire compartment#

&==

5=

Temperature2 #3

==

15=

1== ! 2banded3 !42banded3 ! 2lumped3 !42lumped3 = == 4== 6== <== 1=== 1 == 14==

5=

Time2s3
Figure A 6 The e!!ects o! banded and lumped absorption coe!!icients at locations ! and !4

5=

==

Temperature2 #3

15=

1==

5= l& 2banded3 l& 2lumped3 = = == 4== 6== <== 1=== 1 == 14==

Time 2s3
Figure A 6 The e!!ects o! banded and lumped absorption coe!!icients at location l&

#T0#M e*ternal column case modelling /ith S.F0)(SAF01

The test described here, which is referred to as Test 4A, is taken from a series of &A steel column tests carried out at CT2C-, )rance, in 4?H?# The aim of these tests, as part of a research program, was to determine the fire resistance of loaded and non/ loaded structural members outside, as well as inside a building# The fire in the building was produced by the free combustion of a mass of woodwood fuel load#

,0perimental setup A diagram of the e0perimental compartment is given in the following figure9

&+1=m

.pening +65 m &+65 m

1+&6m 1+4:m &+65 m

Figure 15 6 .%er%ie/ o! the e*perimental set up !or e*ternal column test[15]

!ide walls and the back wall of the fire compartment was made of two layer of bricks9 an e0ternal layer of ordinary 44#>cm thick bricks and an internal layer of $,;6A$ 4(#Hcm thick bricks The front faZade containing the opening is composed of cellular concrete blocks of &'cm think# The ceiling is in 4H#>cm thick cellular concrete slabs# A A metre high fire proof screen e0tends the front/wall above the furnace#

According to the test 4A report provided by CT2C-, the fire load density is eEual to A?kgDm&# This corresponds roughly to >H'-JDm& assuming a heat of combustion of 4<#C-JDkg#

The structural members are =,A&&' for the column and =,B&&' for the beam, which were both loaded#

The unprotected column is located ><>mm from the faZade of the e0ternal wall, centred on the opening with the 2/section flanges parallel to the faZade of the fire compartment# 2t is < metres high between top and bottom simple supports# An a0ial load was applied by means of a 4'' tonnes .ack placed at the bottom + The

compressive load applied on the column eEuates to CA'k:# The temperature of the column was measured by &? thermocouples distributed in ( cross/sections down the column, as shown in figure 4C# At cross/sections &, < and C, atmosphere

thermocouples were located at a distance 4'cm from the profile surface in order to measure the ambient temperature, as shown in figure b#

F01)

2 6

Figure 16 6 ;osition o! the thermocouples in the column

F01)

=+1 m

=+1 m

Figure 1: 6 Thermocouple distribution around column

The temperature of the internal beam was measured with thermocouples placed web and top and bottom flange of ht beam# The beam comprises three Gones, as shown

in figure 4(# 2n Gone 4, the steel length varies from >>'mm/H>'mm depending on the test and is not protected e0cept by the presence of the front wall that shields it for roughly 4>'mm# 2n Gone &, the steel stretches for A&H'mm and is protected with a 4>mm thick layer of asbestos cement and a <' mm thick layer of vermiculate, as shown in figure 4?# 2n Gone A, the steel is not influenced directly by the flames but is affected by heat transfer due to conduction from Gone &# A two/point load was provided by two .acks 4>''mm from either ends of the beam# The two concentrated loads applied on the beams are eEual to 4'Ck: on each one#

Figure 1< 6 ;osition o! the thermocouples in the beam

Figure 18 6 ;rotection o! beam

Fire de%elopment The mass of the combustible fire load, in this case a wood crib, was monitored during the test so that an estimation of the heat of release rate can be calculated and subseEuently used as input in the fire development model# The following figure shows the wood mass loss through time#

6==

5==

#rib Mass 2kg3

4==

&==

==

1==

Figure A: Mass loss e%olution o! /ood crib during compartment !ire


Time 2seconds3

5==

1===

15==

===

5==

Figure = 6 >ood mass loss during the compartment !ire

#F' simulation !or the #T0#M !ire de%elopment

Figure 1 6 S.F0) #F' model o! the e*terior and interior o! the #T0#M test

Based on the setup of the test, i#e# the compartment and component dimensions, and the measured mass loss rate, a C) fire development model was established in

!1)2, as shown in the above figure, but for the purposes of testing the one/way methodology the structural steel component has been omitted during the simulation# The following figure shows a comparison between the e0perimental measurement and C) calculated average compartment temperatures#

1 ==

1===

Temperature 2 o#3

<==

6==

4==

== )*perimental S.F0) simulation = = == 4== 6== <== 1=== 1 == 14==

Time 2seconds3

Figure

6 #F' and e*perimental a%erage compartment temperatures

-eat trans!er 2t is the role of the thermal response model, as part of the one/way coupling methodology, to recover the calculated radiative intensities and corresponding directional vectors and applying them to the ), model as radiative flu0es in order for the latter model to calculate the temperature distribution in the solid# The following figure illustrates the location along the sections where comparison between the e0periment and simulations are made#

*8 *1 *2 1m

1m +4 5m 4+<m *7

Figure & 6 $ocations o! sections /here comparison is done

-eat trans!er results 1nce the heat transfer calculations are carried out by the thermal model in !A)26, the temperature distribution within the solid is available for the beam and the column# !teel temperature curves are determined from the !A)26 calculations for each section presented in figure &A# The abbreviations used for the temperature curves in the following figures are9 - 7)9 average temperature on the une0posed flange - W9 temperature on the web - ,)9 average temperature on the e0posed flange

The simulation for the insulated part of the beam, as shown in figure &<, is not really good# The ma0imum temperature, Tma0, reached in the simulation in all regions of the beam is higher than the ones obtained with the test# This is due to the fact that the thermal properties of the insulation are not well defined in the simulation . A

4== BF2test3 >2test3 )F2test3 BF2sim3 >2sim3 )F2sim3

&==

Temperature #3 2

==

1==

= = == 4== 6== <== 1=== 1 == 14==

Time2s3

Figure 4 6 Section S1 o! the beam

compromise has also been made with the geometry of the protection#

The temperatures in the unprotected part of the beam, shown in figure &>, are better# The heating stage seems to be more important in the simulation than observed in the fire test# The Tma0 reached in all regions in the simulation are higher than that of the fire test because the thermal calculations were carried out separately for both elementsK therefore conduction between the elements is neglected in the simulation# 2t should be noted that correlations, figure &H, for section !< of the column follows the same reasoning, however, a closer correlation with the test is observed#

:== 6== 5== )F2test3 BF2test3 >2test3 )F2sim3 BF2sim3 >2sim3

Temperature2 #3

4== &== == 1== = = == 4== 6== <== 1=== 1 == 14==

Time2s3
Figure 5 6 Section S o! the beam

The calculated temperatures of the section, !A, directly in front of the opening fit between the bounds of the test range# A good correlation is observed, in figure &C, for the web and the une0posed flange of the section# =owever, the thermal

calculations have failed to register the intense rise in temperature of the e0posed flange# This may be attributed to how the method captures radiation, as it has struggled to go beyond >''oC, so crucial radiative information may have been lost when the C) stored data was transferred to a coaurser mesh#

<== )F2test3 )F2sim3 >2test3 >2sim3 BF2test3 BF2sim3

6==

Temperature2 #3

4==

==

= = == 4== 6== <== 1=== 1 == 14==

Time2s3
Figure 6 6 Section S& o! the beam

4==

&==

Temperature2o#3

==

1==

)F2test3 )F2sim3 >2test3 >2sim3 BF2test3 BF2sim3 = == 4== 6== <== 1=== 1 == 14==

Time2s3
Figure : 6 Section S4 o! the beam

Mechanical calculation The mechanical model uses the steel temperature distribution calculated in the thermal model as input in order to calculate the mechanical response# The

temperature distribution profiles are transferred on to the mechanical model via integration points along the finite beam elements# There are two integration points per beam element in the !A)26 mechanical model# 2n addition to the thermal loads, mechanical loads and boundary conditions are applied to the mechanical model, and are presented in the following figure9
6&=C" y D 1==C" Fi*ed 1==C" -inge

1+5m

&+5m -inge

Figure < 6 Mechanical loads and boundary conditions applied to the mechanical model

The mechanical effect of the compartment faZade on the above set up is taken into account in the simulation, i#e# displacements are blocked in the B and W direction near the front end of the beam, highlighted with a red circle#

Mechanical calculation results The ma0imum displacement in the y/direction *Dy,ma0[&'mm+, figure &?, is reached on the column below the beam/column connection in front of the compartment opening# This part of the structure is very critical due to several effects such as

thermal effects, as this part is in front of the opening of the fire compartment, transverse loading due to the thermal e0pansion of the beam and small effects of buckling due to the vertical loading of CA'%:#

A ma0imum displacement in the G/direction of appro0imately 4Amm, figure A', is reached at the mid/span of the beam due to & point vertical loads of 4''%:# A thermal gradient, T with reference to figure &<, of appro0imately 4>'oC is Euite important in this part of the mechanical model as the hotter e0posed flange will induce and increase the bending of the beam#

The displacements have been e0tracted from the mechanical model in order to make a comparison with the test results, as shown in both figures &? and A'# The figures show the displacement results obtained and also illustrate where they were e0tracted from along the beam and column#

The steel temperature curve of the & section show good correlation with the test up to C''oC but suddenly diverges away from the test results# This may be due to the thermal properties of the insulation not being appropriately defined in the thermal model resulting in a higher thermal gradient and hence higher displacements# This has inadvertently caused large displacements in the column, e0pansion# 4, due to thermal

y D

+
1m

5 Test Simulation =

'isplacement 2mm3

(5

(1=

(15

( =

( 5 = == 4== 6== <== 1=== 1 == 14==

Time2s3
Figure 8 6 #olumn displacement, '1, comparison bet/een !ire test and SAF01 simulation

y D +5 m

14 1 1=

'isplacement 2mm3

< 6 4

Test Simulation = == 4== 6== <== 1=== 1 == 14==

Time 2s3
Figure &= 6 9eam displacement, ' , comparison bet/een !ire test and SAF01 simulation

Conclusion
2t has been found that the number of directions does not have to be in the order of magnitude of 4'', yet the minimum number of C is simply not sufficient to capture the radiative field#

'eferences2
345 Computer models for fire and smoke# n#d# 3online5# 3Accessed &C th )ebruary &''H5# Available from the WWW9 http9DDwww#firemodelsurvey#comDpdfD!1)2,V&''H#pdf 3&5 )ranssen# J#-, %odur# @#%#6, -ason# J# ,ser@s manual for SAF)* $007, 7niversity of $i8ge &''> 3A5 J-T "enrose, 6 =ose, CJ !taples, 2! =amill, 2" Jones and !weeney, &'''# )luid and ),# 4( th CA /),- 7sersR -eeting,

!tructure interactions9 Coupling of C) &'/&& !eptember &''' 3<5

,uropean Committee for !tandardiGation *C,:+# 4??<# B! ,: 4??4/&/&94??<# ,urocode 49 Basis of esign and esign Actions on !tructures# Brussels9 B!2

3>5

Coupled -ulti/"hysics Applications# &''<# 3online5# 3Accessed C th April &''>5# Available from WWW9 www#scai#frauhofer#deDmpcciVapplications#html

3C5

"rofessor Bailey, C# n#d# 1ne !top !hop in !tructural )ire ,ngineering 3online5# 3Accessed &>th July &''>5# Available from WWW9

http9DDwww#mace#manchester#ac#ukDpro.ectDresearchDstructuresDstrucfireD esignDperfo rmanceDfire-odellingDdefault#htm 3H5 ,uropean Committee for !tandardiGation *C,:+# &''&# B! ,: 4??4/4/&9&''&# ,urocode 49 ;eneral actions J Actions on structures e0posed to fire# Brussels9 B!2 3(5 3?5 Co0# ;# 4??># Combustion )undamentals of )ire, $ondon9 Academic "ress rysdale, # # 4??(# 2ntroduction to fire dynamics# !econd edition# Chichester9

John Wiley and !ons 34'5 ,uropean Committee for !tandardiGation *C,:+# &''># B! ,: 4??A/4/&9&''># ,urocode A9 esign of !teel !tructures# Brussels9 B!2 3445 ,uropean Committee for !tandardiGation *C,:+# 4??A# B! ,: 4??&/4/&94??A# ,urocode &9 esign of Concrete !tructures# Brussels9 B!2 34&5 ,uropean Committee for !tandardiGation *C,:+# 4??<# B! ,: 4??>/4/&94??<# ,urocode >9 esign of Timber !tructures# Brussels9 B!2

34A5

Cooper# $#Y, )ranssen# J#-# 4???# A basis for using fire modelling with 4/ thermal analyses of partitions to simulate &/ and A/ structural performance in real fires# )ire !afety Journal, AA pp#44>/4&(#

34<5

T:1

2A:A / A T:1 Company n#d# 3online5# 3Accessed 4( th :ovember &''>5#

Available from http9DDwww#tnodiana#com 34>5 !heffield 7niversity ,nterprise $imited *!7,$+# n#d# 3online5# 3Accessed 4( th :ovember &''>5# Available from http9DDwww#suel#co#ukDsuelDcompaniesDvulcan#html 34C5 2ding# 6, Bresler# B# M :iGamuddin# W# 4?HH# )26,!/TA9 A Computer "rogram for the )ire 6esponse of !tructure/Thermal *Three/ imensional @ersion+, B)6$ :2!T 34H5 Breunese, A#, )ellinger, J#, %enyon, Y# M $emaire, A# *&''<+ OA coupled C) /),A study of a tunnel structure e0posed to a fireP, 9ro!. 10th )7T+*F6AM #onf., pp# 4A<?J4AC'# 34(5 Whao, B#, esanghere, !# M arche, J# *&''H+ OC) and ),- coupling with

computer programs ) ! and A:!Y! for fire resistance assessmentP, 9ro!. 11th )7T+*F6AM #onf., pp# 4<C4J4<CC# 34?5 Cssgrid, a cubic spline interpolation package for random data on a sphere# &''<# 3online5# 3Accessed &&nd !eptember &''H5# Available from http9DDngwww#ucar#eduDngmathDcssgridDcsshome#html 3&'5 6enka# 6# J# *4??H+# OA$;162T=- HH&# !T62"AC%9 @oronoi elaunay Triangulation and

iagram on the !urface of a !phereP, A#M Transa!tions on.

Mathemati!al.Soft/are, @ol# &A, :o# A, !ept# 4??H, pp# <4C/<A<# 3&45 6enka# 6# J# *4??H+ OA$;162T=- HHA# !!6)"AC%9 2nterpolation of !cattered ata on the !urface of a !phere with a !urface under TensionP, A#M Transa!tions on. Mathemati!al. Soft/are, @ol# &A, :o# A, !ept# 4??H, pp# <A>/<<&# 3&&5 %umar, !#, -iles, !# Welch, !#, @assart, 1#, Whao, B#, $emaire, A# #, :oordi.k, $#-#, )ellinger, J#=#=# M )ranssen, J#/-# *&''(+ )26,!T67C J 2ntegrating advanced three/dimensional modelling methodologies for predicting thermo/ mechanical behaviour of steel and composite structures sub.ected to natural fires, 6)!/"6/'&44', ,uropean Commission#

3&A5

2ncropera, )#"# 4??C# )undamentals of =eat and -ass Transfer# )ourth edition# Jon Wiley and !ons

3&<5

,ckert, ,#6#;# M rake, 6#-# 4?>?# =eat and -ass Transfer# !econd edition# :ew York9 -c;raw/=ill,

3&>5

$allemant, :#, !ayre, A# and Weber, 6# *4??C+# O,valuation of emissivity correlations for =&1/C1&/:&Dair mi0tures and coupling with solution methods of the radiative transfer eEuationP# "rog# ,nergy Combust# !ci#, vol# &&, pp# ><A/>H<

3&C5

@eresteeg, =#%#, -alalasekera, W#, &''H# An introduction to Computational )luid ynamics J The )inite @olume -ethod# !econd edition# ,sse09 "earson ,ducation limited#

Appendi,

CTI NP interpret interpret

generate

generate

generate

Mesh on CFD solid

(nterpol factors CFD to FEM interpret interpret

(nterpol factors FEM to CFD

6ie# factors

Mesh on FEM solid

generate

interpret

generate generate (ncident rad9 flu, generate

CFD

calculate calculate

(ncident rad9 intensity

FEM

interpret CFD

-as temp: FEM 2-*5-).4 generate conv htc interpret calculate I+'&,-+) .(,. *urface temp9 emission generate coeff9 I+'&,-+) .(,. 6%37 generate

Env temp: env htc generate FEM generate 0).3')3.(%

P*0) CFD

interpret

FEM )1-.5(%

Fig+A+ Method 9 diagram interpret


G(0 )-5$8 '*+9. 1)'

Fig+A+ Method # diagram

Mid(span Euarter(span

'isplacement 2mm3

(5

(1=
FEM %&+-0

interpret

(15
CFD

I+'&,-+) .(,.

generate

calculate P*0) P.*' FEM 0).3')3.(%

1=

&=

4=
generate

generate 5= I+'&,-+) .(,. 6%37

6=

interpret interpret

Time 2min3

G(0 )-5$8 : '*+9. 1)'

Fig+A+ Method ) diagram

Вам также может понравиться