15
Comparing simple and advanced tools for structural re safety engineering Nicolas Henneton* ,, Christophe Renaud and Bin Zhao CTICM (Centre Technique Industriel de la Construction Métallique), Saint-Aubin, France SUMMARY The present paper gives an overview of the actual tools available for the estimation of the re development and of the resulting thermal actions on structural members. A case study is developed on the basis of the Fire Safety Engineering methodology, respectively with two different approaches, one based on advancedtools and another one based on simpliedtools. Indeed, three categories of re models are used in this study, each of which corresponding to a different level of precision and complexity: hand calculations, zone models, and computational uid dynamics (CFD) models. The case study is relative to the calculation of the heating of a portal frame in a gymnasium, under localised real re conditions. It is shown through comparisons that, in this case, predictions of analytical methods are, to certain extent, in good agreement with predictions of the CFD model. In particular, it is demonstrated the relevance of using a simplied method of EN 1991-1-2 to predict thermal actions to vertical members. The obtained results also highlight the need to develop more relevant analytical methods in order to predict the temperature eld during a re in a large volume. Copyright © 2014 John Wiley & Sons, Ltd. Received 27 September 2013; Revised 12 March 2014; Accepted 21 August 2014 KEY WORDS: re safety engineering; analytical model; CFAST; FDS; localised re; heat transfer 1. INTRODUCTION In the structural Eurocodes, standard procedures for the prediction of thermal actions on structures are dened. The way to evaluate the thermal actions in case of real res is described in Eurocode 1 part 12 [1]. There are basically three main approaches to deal with re development and the resulting thermal actions [24]. The simplest is to use analytical methods, which can be set up quickly and are easy to use because of the limited parameters involved. These analytical formulations and empirically based correlations have been used for more than ve decades [5]. However, results may only be correct within a narrow range of applicability. Nevertheless, these models can serve as a rst step for more sophisticated computer modelling. A more complex approach is the use of zone models [6], which assume the compartment as being composed of a plume region and two gaseous layers, both interacting through the governing equations. Zone models consider the temperature in these layers to be time dependent but uniform in space. Most existing zone models have been developed for single compartments or series of compartments whose size is representative of domestic rooms, small ofces or small industrial units. For this kind of compartments, the zone models predict reasonable results. Despite being classied as advanced modelsin Eurocode 1 part 12, they are rather simple to set up and the computational time is of the order of seconds on a modern desktop PC. However, when zone models are used for the design of structural members, the effects of the radiation from a localised re to the structure *Correspondence to: N. Henneton, CTICM (Centre Technique Industriel de la Construction Métallique), Saint-Aubin, France. E-mail: [email protected] Copyright © 2014 John Wiley & Sons, Ltd. FIRE AND MATERIALS Fire Mater. (2014) Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/fam.2268

2014_Zhao_Comparing Simple and Advanced Tools for Structural Fire Safety Engineering

Embed Size (px)

DESCRIPTION

Numerical tools for structural fire engineering

Citation preview

  • FIRE AND MATERIALSFire Mater. (2014)Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/fam.2268Comparing simple and advanced tools for structural firesafety engineeringNicolas Henneton*,, Christophe Renaud and Bin Zhao

    CTICM (Centre Technique Industriel de la Construction Mtallique), Saint-Aubin, FranceSUMMARY

    The present paper gives an overview of the actual tools available for the estimation of the fire developmentand of the resulting thermal actions on structural members. A case study is developed on the basis of the FireSafety Engineering methodology, respectively with two different approaches, one based on advanced toolsand another one based on simplified tools. Indeed, three categories of fire models are used in this study,each of which corresponding to a different level of precision and complexity: hand calculations, zonemodels, and computational fluid dynamics (CFD) models. The case study is relative to the calculation ofthe heating of a portal frame in a gymnasium, under localised real fire conditions. It is shown throughcomparisons that, in this case, predictions of analytical methods are, to certain extent, in good agreementwith predictions of the CFD model. In particular, it is demonstrated the relevance of using a simplifiedmethod of EN 1991-1-2 to predict thermal actions to vertical members. The obtained results also highlightthe need to develop more relevant analytical methods in order to predict the temperature field during a fire ina large volume. Copyright 2014 John Wiley & Sons, Ltd.

    Received 27 September 2013; Revised 12 March 2014; Accepted 21 August 2014

    KEY WORDS: fire safety engineering; analytical model; CFAST; FDS; localised fire; heat transfer1. INTRODUCTION

    In the structural Eurocodes, standard procedures for the prediction of thermal actions on structures aredefined. The way to evaluate the thermal actions in case of real fires is described in Eurocode 1 part 12[1]. There are basically three main approaches to deal with fire development and the resulting thermalactions [24]. The simplest is to use analytical methods, which can be set up quickly and are easy touse because of the limited parameters involved. These analytical formulations and empirically basedcorrelations have been used for more than five decades [5]. However, results may only be correctwithin a narrow range of applicability. Nevertheless, these models can serve as a first step for moresophisticated computer modelling.

    A more complex approach is the use of zone models [6], which assume the compartment as beingcomposed of a plume region and two gaseous layers, both interacting through the governingequations. Zone models consider the temperature in these layers to be time dependent but uniform inspace. Most existing zone models have been developed for single compartments or series ofcompartments whose size is representative of domestic rooms, small offices or small industrial units.For this kind of compartments, the zone models predict reasonable results. Despite being classifiedas advanced models in Eurocode 1 part 12, they are rather simple to set up and the computationaltime is of the order of seconds on a modern desktop PC. However, when zone models are used forthe design of structural members, the effects of the radiation from a localised fire to the structure*Correspondence to: N. Henneton, CTICM (Centre Technique Industriel de la Construction Mtallique), Saint-Aubin, France.E-mail: [email protected]

    Copyright 2014 John Wiley & Sons, Ltd.

  • N. HENNETON, C. RENAUD AND B. ZHAOshould be considered additionally with help of analytical methods [7]. Besides, as for analyticalmethods, it is not possible with zone models to predict precisely fire development along andbetween combustible materials, and the user needs to specify heat release rate (HRR) curves.

    The most complete approach is the use of field models (or CFD models), which solve numericallythe 3D governing equations for a fire-driven flow in their differential form with varied levels ofcomplexity. Fire development can be simulated on the basis of the thermo-physical properties of thematerials rather than the simply prescribed HRR history.

    Unfortunately, there is no guidance in Eurocode 1 part 12 on how to select the appropriate type ofmodel according to the encountered case study and no recommendations on the scope and limits ofsuch models. Moreover, although sophisticated models have found increasing application in firemodelling and thus in fire safety engineering, it is clear that a large number of practitioners,including fire engineers, and fire control authorities still lack of in-depth awareness of the firemodelling techniques. Indeed, these sophisticated models consider in detail the most important firemechanisms, which imply a high number of fundamental parameters and require therefore furthercalibration and validation. Another point is that these complex models may be limited by theimpossibility to obtain the inputs required [4].

    However, there exist specific cases where hand calculations or simpler models can be used as agood approximation of more complex models, provided that assumptions are correct. In thiscontext, this paper aims at studying the heating of a portal frame of a gymnasium in real fireconditions, respectively with two different approaches, by using the fire safety engineeringmethodology [8]. This is the specific case of a localised fire in a large building because allcombustible materials cannot be involved simultaneously in the fire. Research in this topic is ratherlimited because design fire scenarios most commonly used for the structural design of buildings arefrequently based on post-flashover fires, with the assumption of uniform temperature conditionsthroughout a compartment, regardless of its size ([11, 12]). This engineering approach will follow thesteps below:

    Description of the gymnasium, including the structural members to be analysed and fuel loads, Selection of the design fire scenario challenging the structure, Determination of the thermal actions on the structural members, both with simplified andadvanced tools,

    Evaluation of the thermal response of the structure, both with simplified and advanced tools.2. DESCRIPTION OF THE CASE STUDY

    Figure 1 gives an overview of the gymnasium and of the different activities planned within. Thefloor area is 34 84m2, and the height of the gymnasium is 11m. There are 10 smoke vents(activated at a temperature of 93 C), representing 1% of the roof area, and 2 smoke screens of2m high (all are visible in Figure 1). The vents will be considered in both the CFD model andthe zone model.Figure 1. Description of the gymnasium (snapshot from the FDS simulation).

    Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • COMPARING SIMPLE AND ADVANCED TOOLS FOR STRUCTURAL FIRE SAFETY ENGINEERINGIn order to select the appropriate design fire scenario for the evaluation of the portal frame, fuel loads needto be determined. Many activities that take place in the gymnasiummay involve combustible materials: shotput, high jump, pole vault, climbing, sprint, and other accessories such as a wooden podium, chasubles,balls, hoops, etc. Besides, the bleachers consist of glue-laminated benches. To be on a safe side, it will besupposed that all devices are set up at the same time. Table I summarizes the corresponding fire loads.The foam gym mats represent almost half of the total fire load, which is approximately 400GJ.

    The load-bearing structure of the gymnasium consists of unprotected steel portal frames with weldedsections (PRS), spaced at 6m intervals. The span of the portal is 33.7m, and the grade of structuralsteel is S235. The characteristics of the structural members are summarized in Figure 2.

    The portal frame that is studied by the fire safety engineering methodology is located close to thehigh jump and shot put activities, as shown in Figure 1. The columns are not embedded in thewalls, so the column close to the high jump activity is likely to be engulfed by the fire.3. METHODOLOGY FOR EVALUATING THE THERMAL ACTIONS

    As mentioned in the introduction, two approaches have been considered to assess the heating of the portalframe in real fire conditions and will be compared: an approach based on advanced tools (later called in thetext advanced approach) and another based on simplified tools (later called simplified approach).

    3.1. Advanced approach

    The first approach uses advanced tools, namely the fire dynamics simulator (FDS) code (version 5.5.3)developed by National Institute of Standards and Technology [13] to simulate fire development andpropagation of smoke in the gymnasium and the FEM code ANSYS [14] to calculate the heating ofTable I. Estimated fire loads in the gymnasium.

    Location/activity Combustible material Fire load (MJ)

    Bleachers Glue-laminated benches 51 000High jump Polyester foam mats 33 000Pole vault Polyester foam mats 61 200Sprint Hurdles (PVC) 2 550Sprint Starting block (PE) 2 400Sprint Polyester foam mats 17 000Shot put Safety net (polyethylene rope) 2 000Shot put Standing support structure (wood) 15 000Shot put Polyester foam mats 21 300Climbing Climbing wall (wood) 136 000Climbing Polyester foam mats 32 400Miscellaneous accessories Pommel horse, podium, hoops, balls, chasubles 30 500

    Figure 2. Portal frames.

    Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • N. HENNETON, C. RENAUD AND B. ZHAOthe steel members. A coupling procedure between FDS and ANSYS developed within the scope of theEuropean research project FIRESTRUC [15, 16] is used to predict the thermal actions to the portalframe. This is a one-way coupling procedure developed to take into account the radiation effect forany type of structural members cross section shape, in particular when the shadow effects aresignificant. It consists of the following steps:

    1. In the FEM code ANSYS, members of the portal frame are modelled using brick elements(SOLID70). Each structural member is considered independently, and its connection with othermembers is neglected. In such a way, each member will have its own heat transfer model. Thermalactions and radiation effect between the walls of structural members are managed with specificplane elements (SURF152).

    2. The geometric description of all FEMmembers is written to afile. For eachmember, the number of com-posing finite elements is given. For each finite element, location, normal and solid angle seen are given.

    3. A FDS simulation is conducted, and a result file is generated, containing gas temperature, theconvective heat transfer coefficient and average radiant intensities along every direction, for eachcell in 3D user-defined zones.

    4. A post-processing program reads the file with geometric data made in step 2 and converts thethermal results of the whole (time dependent) FDS run (written in the file made in step 3) to thespecific locations of the mesh nodes of each FEM member.

    5. ANSYS reads the file with thermal data made in step 4 and computes the time-dependenttemperature response using this data as a boundary condition.

    3.2. Simplified approach

    The second approach is based on simplified tools. Various methods and formulae are used, dependingon the situation.

    3.2.1. Thermal actions resulting from a localised fire to the horizontal structural members of theportal frame. Thermal actions resulting from a localised fire to the horizontal structural members ofthe portal frame are calculated according to simple analytical formulae given in EN 1991-1-2,namely Hasemis and Heskestads methods [1].

    In Heskestads method, the flame height Lf is first given by

    Lf 1:02D 0:0148Q2=5 m (1)where D is the diameter of the fire [m] and Q the HRR of the fire [W]. Then, the temperature (z) alongthe axis of the plume is estimated by

    z 20 0:25Qc2=3 z z0 5=3 900C (2)where Qc is the convective part of the HRR [W], z the height [m] along the flame axis and z0 the virtualorigin [m]. In this study, the resulting net heat flux per unit area hnet received by a horizontal structuralmember will be finally calculated by

    hnet f m zLf 273 4

    m 273 4h i

    c zH m

    W=m (3)

    where is the configuration factor, f is the emissivity of the fire (equal to 1), m is the emissivity of thestructural member (equal to 0.7), is the Stephan Boltzmann constant, c is the heat transfer coefficient(equal to 35W/m2K), H is the distance between the fire source and the ceiling [m], and m is the surfacetemperature of the structural element [C].

    Therefore, in this case, the radiative part of the heat flux is calculated from the temperature at theheight of the flame axis (a disc with a diameter identical to the diameter of the fire is considered forthe configuration factor) and the convective part of the heat flux is obtained from the temperature ofgases under the ceiling.

    In Hasemis method, the net heat flux per unit area hnet is calculated from a horizontal flame lengthLH and the horizontal distance between the vertical axis of the fire and the point along the ceiling wherethe thermal flux is calculated, r (refer to EN 1991-1-2 for more details of the calculation procedure).Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • COMPARING SIMPLE AND ADVANCED TOOLS FOR STRUCTURAL FIRE SAFETY ENGINEERINGIn theory, Heskestads method should be used if the flame is not impacting the ceiling and Hasemismethod in the other case (flame impacting the ceiling). In this study, the height under the ceiling is11m; therefore, flames are not likely to impact the ceiling. However, the two methods will be used,regardless of this limitation, in order to compare the results.

    3.2.2. Thermal actions resulting from a localised fire to the vertical structural members of the portalframe. Thermal actions resulting from a localised fire to the vertical structural members of the portalframe are calculated according to a simplified model based on the concept of solid flame. As shown inFigure 3, the flame envelope is modelled as a radiative area (semi-cylinders and rectangular planes),discretized in bands of constant temperature, whose properties evolve with time according to theHRR and depend on the height considered. The height of the flame is calculated from Eqn (1), andthe temperature of each elementary band (10 cm high) is given by Eqn (2). Then, the radiative heatflux received by an elementary face of a columns section is the sum of radiative heat fluxes emittedby each elementary band. More details for this simplified model can be found in Thauvoye et al. [17].

    3.2.3. Thermal actions resulting from the hot smoke layer on structural members far from fire.Thermal actions resulting from the hot smoke layer on structural members far from fire arecalculated using the temperature of the hot gas layer and the smoke layer height predicted by thetwo-zone model CFAST [18] and Eqn (3). Actually, at each time step,

    Radiative and convective heat fluxes will be calculated to horizontal structural members, and alsoto vertical structural members if they are in the hot gas layer,

    Radiative heat fluxes will be calculated to vertical structural members if they are in the cold gas layer.

    Figure 4 gives an overview of the CFASTmodel. Smoke vents and natural ventilation are simulated, withequivalent geometric dimensions to those used in the FDS simulation. Because the assumption of uniformproperties within a layer may be challenged for spaces that are very long as a result of cooling and loss ofbuoyancy far from the fire, that may not be accounted for in the zone model [9], the gymnasium issubdivided into three virtual rooms connected by full height and width openings in order to bettersimulate variations in gas temperature and layer height across the area of the large space. The virtualopenings between the compartments are located under the smoke screens, as can be seen in Figure 4.

    3.2.4. Heating of the portal frame. The combination of both above models allows the determinationof the temperature field near and far away from the fire. In consequence, the thermal actions for all thestructural members are obtained by taking the most important value predicted by above models.

    For the heating of the structural members, a uniform temperature distribution in the cross section isconsidered, and the increase of temperaturem during a time intervalt is calculated from the equationFigure 3. Model of solid flame.

    Figure 4. CFAST model.

    Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • N. HENNETON, C. RENAUD AND B. ZHAOm Am=V =C hnett (4)

    where ,C and (Am /V), are, respectively, the density, specific heat, and section factor of the steel element.4. DESIGN FIRE SCENARIO AND DESIGN FIRE

    4.1. Design fire scenario

    A fire scenario is a qualitative description of the course of a particular fire with respect to time and space [10].The design fire scenario is the fire scenario selected to be the more challenging for the structure subject toanalysis. The factors considered in order to select this more challenging fire scenario concern thereforethe fire itself (distribution and type of fuel, fire load density, fire spread) and the position of the firerelative to the portal frame. In this study, the selected design fire scenario consists of a fire starting on apiece of foam gym mat near the portal frame studied and then spreading to the neighbouring mats used inactivities such as high jump or shot put (refer to Figure 1). Indeed, these foam mats

    Represent an important fire load density under the portal frame studied, Are very close to the column 1, Are composed of materials (mainly polyester) that will support a rapid fire growth.

    4.2. Design fire

    Once the design fire scenario is selected, the characteristics of the design fire have to be determined. Adesign fire is a simplified but still representative description of the complex physical and chemicalprocesses occurring in a fire. Design fire is usually characterized in terms of time-dependentvariables, such as the HRR. Fire can grow from ignition to a fully developed stage and finally decayand eventually burnout [10]. In the gymnasium, it can be supposed that the fire would rapidly reacha fully developed stage where the rate of combustion is limited by the fuel. The burning rate of fuelcontrolled fires is dependent upon the nature and surface area of the fuel.

    In fire safety engineering, there are three main approaches to determine the HRR as a function time:

    Analysis of experimental data and definition of a simplified HRR curve that is conservative. Construction of each part of the curve (growing, stationary and decay phases) by using analyticalrules and parameters depending on the type of building, such as the commonly used relationshipsgiven in EN 1991-1-2. In this approach, the user needs to estimate a priori the amount ofcombustible that will be involved in the fire. A classical hypothesis is to consider that allcombustible materials within a compartment are burning after the flashover has occurred.

    Use of fire modelling to assess the HRR. It leads to a more specific fire development, dependingon the rate of fire growth of the combustible materials.

    4.2.1. Advanced approach. In this approach, the FDS code is used to estimate the fire growth and theresulting HRR curve is calculated by simulating the flame spread along the combustible surfacesinvolved in the fire. The flame spread is modelled by heat transfer through convection and radiation.As a surface reaches the specified ignition temperature, that part of the surface ignites and burnsfollowing a pre-defined constant HRR per unit area (HRRPUA).

    In this case, the design fire consists in

    Defining the characteristics of the combustible materials (specific heat, density, conductivity,ignition temperature, HRRPUA, heat of combustion);

    Placing a small radiating panel in front of a foam gym mat in order to ignite the first solidmeshes (Figure 1).

    The remaining surface of the foam mat and all other combustible materials will ignite when theirsurface temperature reach their ignition temperature. Consequently, the fire will propagate to othercombustibles by heating as a result of direct heat radiation or convection from the flames.Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • COMPARING SIMPLE AND ADVANCED TOOLS FOR STRUCTURAL FIRE SAFETY ENGINEERINGIt is worth noticing that adjustments of some of the thermo-physical characteristics may be necessary ina CFD model, depending on the grid size. In large domains, a compromise is needed between flowresolution and computational time. The use of a coarse grid (in the FDS model the mesh used is50 50 50 cm) may result in a lack of accuracy of the heat transfer calculations along combustiblesurfaces. As a rule of thumb, the key parameters (ignition temperature and HRRPUA) can be adjustedby preliminary comparisons with the expected fire growth rate of the material involved (slow, medium,fast or ultra-fast [19]). For example, an ignition temperature of 350C and a constant HRRPUA of400 kW/m2 have been chosen for the foam mats. These values have been calibrated in preliminarysimulations with FDS of a foam mat in fire, in order to have a fast fire growth, as can be seen inFigure 5a and b. One may notice in Figure 5b that between 3 and 6min, the fire growth is rather ofmedium type. This is because of the fact that the fire area is quickly limited laterally by the width ofthe mat. With initial values found in literature (HRRPUA of 400 kW/m2 [20] and ignition temperatureof 400C [21]), the simulated fire spread was too slow (Figure 5b).

    Such calibrations have been achieved for all types of combustible materials: glue-laminatedbenchers (slow fire growth), polyethylene ropes (ultra-fast fire growth), etc Values used aresummarized in Table II.Figure 5. (a) Preliminary simulation with FDS of a foam mat. (b) Resulting HRR in FDS, comparison witht2 curves.

    Table II. Values used in the FDS simulation.

    Combustible material HRRPUA (kW/m2) Ignition temperature (C)

    Glue laminated 175 350Wood 250 340PVC 200 360Polyester 400 350Polyethylene 700 360

    Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • N. HENNETON, C. RENAUD AND B. ZHAO4.2.2. Simplified approach. For both analytical methods and CFAST, the HRR curves are constructedwith the relationships given in EN 1991-1-2. First, it is necessary to determine the fire load, that is tosay the total energy able to be released in case of fire. The fire is supposed to start on a piece of foamgym mat near the portal frame studied and then will spread to the neighbouring mats used in high jump,shot put and sprint activities because polyester foam will support a rapid fire growth. Glue-laminatedbenchers are more difficult to ignite, so it is considered a priori that only benchers located atproximity of the high jump activity will ignite. Finally, the fire load of some miscellaneousaccessories is added in order to be on a safe side.

    Then, in order to meet the applicability requirement of the analytical methods given in EN 1991-1-2,especially with respect to an equivalent diameter of the fire less than 10m, two zones are considered(refer to Figure 6):

    Z1: zone including foam gym mats used for high jump and some glue-laminated benchers at prox-imity, giving a fire load of 51GJ.

    Z2: zone including the shot put activity (foam gym mats, the safety net in polyethylene rope andthe standing support structure in wood), foam mats used at the end of the sprint, and some miscel-laneous accessories, giving a fire calorific load of 62.5GJ.

    Besides, these two zones are more relevant for the solid flame models used to evaluate the thermalactions to the vertical structural members of the portal frame. Figure 7 illustrates the solid flame foreach zone, composed of semi-cylinders and rectangular planes.

    In the simplified approach, the design fire scenario is described as follows: Fire starts in zone Z1 andthen propagates to zone Z2, 5min later. At this time, the HRR is 4MW in zone Z1, corresponding to afire area of 10m2, and the fire will therefore propagate to zone Z2 because the foam mat in zone Z1 isonly 3m wide.Figure 6. Zones considered in the simplified approach.

    Figure 7. Schematic above view of incident heat fluxes and solid flames for zones Z1 and Z2 in the simpli-fied model.

    Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • COMPARING SIMPLE AND ADVANCED TOOLS FOR STRUCTURAL FIRE SAFETY ENGINEERINGThe growing phase for each zone is defined by the expression

    Q 106 t=t W (5)

    where t is the time needed to reach a HRR of 1MW, depending on the fire growth rate. It is supposeda fast fire growth rate for zone Z1 (reasonable assumption for a polyester foam mat), and an ultra-fastfire growth for zone Z2, considering that the fire would be already fully developed in zone Z1, thusleading to high incident heat fluxes to foam mats of zone Z2.

    The growing phase is limited by a horizontal plateau corresponding to the stationary state and to avalue of Q given by multiplying the maximum HRRPUA with the maximum area of the fire, definedby the layout of the combustible materials. For each zone, the max HRRPUA is chosen to be400 kW/m2 [20] (because polyester foam is the dominant combustible), and the calculated maximumfire area is also similar, of approximately 100m2.

    Finally, the horizontal plateau is limited by the decay phase that starts when 70% of the total fireload has been consumed. The decay phase may be assumed to be a linear decrease starting when70% of the fire load has been burnt and completed when the fire load has been completely burnt [1].

    Figure 8 gives the resulting HRR curves. It should be mentioned that the recommendations on therange of applicability of analytical methods given in EN 1991-1-2 are met:

    The equivalent diameter of the fire in each zone is 10m, The maximum HRR in each zone is 50MW.5. RESULTS

    5.1. Results of the advanced approach

    As can be seen in Figures 9 and 10, in the FDS simulation, there is a rapid fire development between 7and 8min, with a maximum HRR of 125MW during a few seconds, corresponding to the fullydeveloped fire in the shot put area, and especially because of the very fast combustion of thepolyethylene ropes surrounding the foam mats. This behaviour is typically difficult to predict in thesimplified approach, when constructed the HRR curve.

    The HRR then stabilizes at 82MW between 15 and 25min after the fire starting, which is verysimilar to the maximum value (80MW) of the HRR curve constructed in the simplified approach. Itallows a relevant comparison between the two approaches, especially at the height of the fire. After40min, there is a slow combustion of wooden combustibles in FDS, so results between the twoapproaches are no longer comparable.

    The maximum gas temperature under the ceiling is approximately 800C and is reached in thecorner of the gymnasium (Figure 11).Figure 8. HRR curves constructed with EN 1991-1-2.

    Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • Figure 9. HRR curve in FDS, comparison with the RR curve used in the simplified approach.

    Figure 10. Snapshots from the FDS simulation.

    N. HENNETON, C. RENAUD AND B. ZHAOThe maximum heating of the portal frame is reached at about 30min after the fire starting. InFigure 12, the snapshot from ANSYS at this time shows the heating of the column 1 close to thefire and part of the horizontal members. The maximum temperature is approximately 700C and isreached within the column 1 at 1m high. The temperature of the members decreases with the height,and in the upper part of the column, the temperature is only 350C. Concerning horizontal membersof the portal frame, the maximum temperature is about 500C and reached above the high jump area.Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • Figure 11. Temperature of gases (C) under the ceiling, t = 20 min. Value of 500C in black.

    Figure 12. Heating of the portal frame (C) obtained with ANSYS, at t = 30 min.

    COMPARING SIMPLE AND ADVANCED TOOLS FOR STRUCTURAL FIRE SAFETY ENGINEERING5.2. Comparison between simplified methods

    As mentioned previously, various methods and formulae are used in the simplified approach,depending on the situation:

    Heskestads and Hasemis methods for calculating thermal actions to the horizontal structuralmembers,

    A simplified model based on the concept of solid flame for calculating thermal actions to the ver-tical structural members,

    The CFAST model for calculating thermal actions resulting from the hot smoke layer on structuralmembers far from fire.

    In the Figure 13 are compared the incident heat fluxes to the horizontal structural members obtained byHeskestads method (for each zone Z1 and Z2), Hasemis method (for each zone Z1 and Z2, calculatingheat fluxes on the vertical axis of the fire, r=0) and the CFAST model. At the height of the fire, between15 and 25min, results obtained with the Hasemis method are by far the most important. Consequently, inorder to have a safe approach, thermal actions predicted by the Hasemis method will be used to study theheating of the portal frame, even though flames are not impacting the ceiling.Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • Figure 13. Comparison of incident heat flux to the horizontal structural members obtained in the simplifiedapproach.

    N. HENNETON, C. RENAUD AND B. ZHAO5.3. Comparison between simplified and advanced approaches

    The Figure 15 gives heating curves of the portal frame for both simplified (dotted line) and advancedapproaches (full line). The selected measurement points are indicated in Figure 14.

    These results allow the following analyses:

    For the horizontal structural member above the fire in the high jump area (x = x1 in Figure 14), thetemperatures are very similar (Figure 15a). The simplified approach (i.e. Hasemis method) givesrealistic results and seems to be conservative, despite the fact that in theory, the method should notbe used because flames are not impacting the ceiling. At the opposite, the Heskestads methodwould give underestimated and not conservative results. This first result highlights the need toextend the scope of application of these methods for large compartments.

    For the horizontal structural member nearly above the fire in the shot put area (x = x2 in Figure 14),the temperature predicted by Hasemis method in zone Z2 is also realistic and conservative incomparison to the advanced approach (Figure 15b). It should be mentioned that these results havebeen obtained using r= 0 in Hasemis method, that is to say calculating heat fluxes at the vertical axisof the flame. This is relevant for x= x1 (this point is at the centre of Z1) but not for x2. In fact, tem-peratures are quickly reduced as the horizontal distance from the axis of the flame increases, which isnot on the safe side. For example, the maximum temperature calculated in the steel member is 550Cabove the axis of the flame in zone Z2 (r=0), 420C with r= 2m and 340C with r=5m.

    For the horizontal structural members very far from the fire (x= x4), results of both approaches arevery close (Figure 15d). In the simplified approach, thermal actions are calculated from the hot gaslayer predicted by CFAST, which seems to be relevant in that case.

    For the horizontal structural members a bit far from the fire (x= x3), still mainly subjected to thehot gas layer, the simplified approach (i.e. CFAST model) underestimates the heating of themembers (Figure 15c). This is because of the fact that the assumption of a uniform hot gaslayer in space is not accurate, in such a large compartment, in the transition zone between the fieldclose to and the field far from the fire. The average gas temperature in the upper layer calculatedby the zone model is not sufficient because there is a vertical thermal gradient within the hot gaslayer at this location.Figure 14. Locations of measurement points for the comparative study.

    Copyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • Figure 15. Heating of the portal frame. Comparison between advanced approach (full line) and simplifiedapproach (dotted line). (a) x = x1. (b) x = x2. (c) x = x3. (d) x = x4. (e) Column 1, z = 1 m. (f) Column

    1, z = 4 m. (g) Column 1, z = 8 m. (h) Column 2, z = 4 m. (i) Column 2, z = 9 m.

    COMPARING SIMPLE AND ADVANCED TOOLS FOR STRUCTURAL FIRE SAFETY ENGINEERING

    Co For the vertical structural members close to the fire (column 1, z = 1m and 4m), the simpli-fied model based on the concept of solid flame gives results very similar to the advancedapproach (Figure 15e and f). There is a tendency to overestimate the incident heat fluxesat z = 4m (Figure 15f) because the flame envelope is modelled as a cylinder, while a coneshould be more appropriate, leading to lower heat fluxes to the column. One can also findthat there is a slight delay between the two approaches, during the first 20min of fire.Actually, in the simplified approach, the solid flame covers a surface area of 100m2 fromthe starting of the fire (as in Figure 7). This is not a correct assumption because the useof Eqns (1) and (2) leads to very low heat fluxes before 15min, when the HRR is notsufficiently important. The model will be improved in the future by considering a solid flamewhose surface area evolves with time according to the HRR.

    For the vertical structural members close to the fire but in the hot gas layer (column 1, z = 8m), thesimplified approach underestimates the temperature within the member (Figure 15 g). In fact, thispyright 2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • N. HENNETON, C. RENAUD AND B. ZHAO

    Copart of the column is only subjected to the hot gas layer computed by CFAST because the flameheight is found to be approximately 5m both in the FDS simulation and with the Heskestadsmethod (the latter being used for the model based on the concept of solid flame, refer to 3.2.2).

    For the vertical structural members far from the fire and in the cold gas layer (column 2, z = 4m),results between the simplified and advanced approaches are very close (Figure 15 h). In thesimplified approach, the members located in the cold gas layer received heat fluxes from boththe upper layer and the solid flames.

    For the vertical structural members far from the fire and in the hot gas layer (column 2, z = 9m),results are also very similar (Figure 15i).6. CONCLUSIONS

    In this comparative study, hand calculations and a zone model, on one hand, and a CFD model, on theother hand, have been used to estimate thermal actions to a portal frame of a gymnasium in real fireconditions. The aim was to use the various tools proposed in the structural Eurocodes and to investigatetheir limits. Two distinct methodologies to assess the HRR of the fire have been compared. It is shownthat in spite of the significant differences in the sophistication of these modelling approaches, in case oflocalised fires, predictions of analytical methods can be in good agreement with predictions of the CFDmodel. In particular, it is demonstrated the relevance of using a simplified method based on the conceptof solid flame to predict thermal actions to vertical members.

    It must be noticed that, in case of localised fire flames not impacting the ceiling, surprisingly, theHasemis method leads to realistic thermal actions for horizontal members if heat fluxes are calculatedat the vertical axis of the flame, while theoretically, it should not be applicable. The obtained resultsalso reveal some limitations in the use of actual simplified tools proposed in EN 1991-1-2 (both zonemodels and analytical methods) to predict the thermal field in the particular transition zone between thefield close to and the field far from the fire, in such a large building. This finding shows the need todevelop more relevant analytical methods to fill the gap in order to prepare the future revision of thestructural Eurocodes. Because such experiments are very rare, an alternative way to validate the newformulae may be to perform parametric studies with help of CFD models.NOMENCLATURE(Am /V)pyright 20Section factor of the steel element (m1)

    C Specific heat (J/kgK)

    D Diameter of the fire (m)

    H Distance between the fire source and the ceiling (m)

    hnet Net heat flux per unit area received by a horizontal structural member (W/m

    2)

    Lf Flame height (m)

    LH Horizontal flame length (m)

    Q Heat release rate (W)

    Qc Convective part of the heat release rate (W)

    r Horizontal distance from the vertical axis of the fire (m)

    t Time needed to reach a heat release rate of 1MW (s)

    t Time interval (s)

    x Measurement point on the portal frame

    z Height (m)

    z0 Virtual origin (m)GREEK SYMBOLSc Heat transfer coefficient (W/m2K)

    f Emissivity of the fire14 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam

  • COMPARING SIMPLE AND ADVANCED TOOLS FOR STRUCTURAL FIRE SAFETY ENGINEERINGmCopyright Emissivity of the structural member

    Temperature along the axis of the plume (C)

    m Surface temperature of the structural element (C)

    m Increase of temperature (C)

    Density (kg/m3)

    Stephan Boltzmann constant (W/m2K4)

    Configuration factorREFERENCES

    1. EN 1991-1-2. Eurocode 1: Actions on structures - Part 1-2: General actions - Actions on structures exposed to fire.European Committee for Standardisation (CEN), Brussels, Belgium, 2002.

    2. Mowrer FW. The Right Tool for the Job. Fire Protection Engineering Magazine 2002; 13:3945.3. Rein G, Bar-Ilan A, Fernandez-Pello AC, Alvares N. A comparison of three models for the simulation of accidental

    fires. Journal of Fire Protection Engineering 2006; 16:183209.4. Torero JL. Scaling-Up fire. Proceedings of the Combustion Institute 2013; 34:99124.5. Beyler CL. Fire plumes and ceiling jets. Fire Safety Journal 1986; 11:5375.6. Olenick SM, Carpenter DJ. An Updated International Survey of Computer Models for Fire and Smoke. Journal of

    Fire Protection Engineering 2003; 13:87110.7. ProfilARBED, Universit de Lige, CTICM, TNO, Labein. Development of design rules for steel structures

    subjected to natural fires in Large Compartments. ECSC Agreement No 7210-SA/210/317/517/619/932. FinalReport EUR 18868 EN, 1999.

    8. ISO/TS 24679:2011. Fire safety engineering - Performance of structures in fire. ISO, Geneva, Switzerland, 2011.9. ISO/TS 13447:2013. Fire safety engineering - Guidance for use of fire zone models. ISO, Geneva, Switzerland, 2013.

    10. ISO/TS 16733:2006. Fire safety engineering - Selection of design fire scenarios and design fires. ISO, Geneva, Switzerland,2006.

    11. Stern-Gottfried J, Rein G. Travelling fires for structural designPart I: Literature review. Fire Safety Journal 2012;54:7485.

    12. Hopkin DJ. Testing the single zone structural fire design hypothesis. Interflam 2013, Proceedings of the 13th

    International Conference, University of London, UK, 2013; 139150.13. McGrattan K, McDermott K, Hostikka S, Floyd J. Fire Dynamics Simulator (Version 5) - Users Guide. NIST

    Special Publication 1019-5, SVN Repository Revision: 6701, September 3, 2010.14. ANSYS. ANSYS Users Manual for Revision 14.0 - Volume IV Theory. Swanson Analysis SYSTEM, INC.,

    Houston, USA, 1992.15. Zhao B, Desanghere S, Darche J. CFD and FEMCouplingWith Computer Codes FDS and ANSYS For Fire Resistance,

    Interflam 2007, Proceedings of the 11th International Conference, University of London, UK, 2007; 14611466.16. BRE, Arcelor Profil Luxembourg, CTICM, TNO Efectis, Universit de Lige. FIRESTRUC - Integrating advanced

    three-dimensional modelling methodologies for predicting thermo-mechanical behaviour of steel and compositestructures subjected to natural fires. Final Report of RFCS Project, Contract No RFS-CR-030300, 2007.

    17. Thauvoye C, Vassart O, Zhao B. Analytical method for the calculation of heat fluxes received by a vertical membersubjected to a localized fire in an open car park. Interflam 2013, Proceedings of the 13th International Conference,University of London, UK, 2013; 10831094.

    18. Jones W, Peacock R, Forney G, Reneke P. CFAST - Consolidated Model of Fire Growth and Smoke Transport(Version 6) - Technical Reference Guide. NIST Special Publication 1026, 2005.

    19. NFPA 92B. Guide for Smoke Management Systems in Malls, Atria, and Large Areas, 2000 Edition. NFPA.20. Mouritz AP, Mathys Z, Gibson AG. Heat release of polymer composites in fire. Composites: Part A 2006;

    37:10401054.21. Braun E, Levin BC. Polyesters: A Review of the Literature on Products of Combustion and Toxicity. Fire and

    Materials 1986; 10:107123.2014 John Wiley & Sons, Ltd. Fire Mater. (2014)DOI: 10.1002/fam