Fire engineering and modeling

Abstract:

In this project, the burning of Heptane Sample in the fire rig, used in the University of Leeds labs, was simulated and studied. The aim was to draw attention to the discrepancies between the expected and the observed results, to the engineers designing the CFD simulation software for fire modelling, suggesting them to give more consideration to the small pool fires growth design. In the present models, the various features in this area like the ignition and convection in the pool, are over predicted and very much simplified, giving inaccurate results.

The project was done using the Fire Dynamics Simulator, FDS, one of the most advanced and powerful CFD programs; and Pyrosim, an interactive tool for FDS developed by thunderhead engineering team. Using these two, first the fire rig was designed employing the best possible accuracy, and then the burning of the Heptane sample was simulated. Furthermore, the air inlet, the extraction system, and the temperature thermocouples were also simulated. Finally, the results of the thermocouples were compared to the actual experiment done by one of the universities researchers.

Introduction

Fire Safety Engineering is the branch of engineering which uses science and engineering concepts for the protection of people and their property, whether living or non living against the hazardous effects of fire and smoke. While the most destructive fire and the beginnings of Fire safety can be traced back to ancient Rome, this discipline emerged as a separate branch of engineering only in the 20th century. This was because of the industrial revolution and a new era in Fire, which would lead to massive destruction of life and property.

Presently the study of behaviour of fire is carried out using Computational Fluid Dynamics, CFD, which is considered to be the most sophisticated and advanced technique in this area. CFD is used to predict Fire growth, smoke flow movement and also compartment temperatures, and hence can be used to develop complete fire models. They can also be used to model pre-flashover conditions and also localized fires in complex geometries with smoke movement in multi-compartments. They can also be used to model pre-flashover conditions and also localized fires in complex geometries with smoke movement in multi-compartments. In a CFD model, basically the partial differential equations are solved for a very large number of points in the compartments. Most CFD models for enclosure fires are appropriate for low-speed, thermally-driven flow with an emphasis on smoke and heat transport from fires.

There are certain limitations however. First of all, the input requirement for CFD models is very demanding and requires expertise in defining the correct input parameters and assessing the feasibility of the calculated results. In addition all computer generated models have limitations too. For example, correlations of flame heights and smoke entrainment rates assume an axis-symmetric, unobstructed fire plume (Hurley 2006). While the correlations could be used in other cases, for example, fires against a wall or in a corner, they require more than simple input of variables. And as the models have increased in sophistication, so have the types of limitations of which users should be aware.

The primary aim of this project is to touch upon these limitations and specifically concentrating on the Pool fire models, describing in detail, the effects and possible causes for the same

BackGground

Computational Fluid Dynamics (CFD) in Fire Safety Engineering

Computational Fluid Dynamics (CFD) is the quantitative prediction of fluid flow characteristics, using modelling techniques and numerical methods. CFD is used to find the approximate solution to the highly coupled differential flow equations for mass, momentum and energy transport.

While the study of fluids can be traced to distant past, historically only Analytical Fluid Dynamics, AFD, and Experimental Fluid Dynamics, EFD were used. The CFD technique was made possible with the advent of digital computers and the phenomenal improvement in their computational capability. The technique originated in the aerospace sciences in the 1960’s. It has since been developed and applied to an increasingly diverse range of problems, including automotive, nuclear engineering, biomedical field, HVAC, Power generation, Oil & Gas, sports and Fire safety engineering (Xing & Stern 2006)

Application of CFD modelling in fire safety science was attempted for the first time in the early 1980s. In 1988 BRE carried out an investigation into the King’s Cross underground station fire in London. As a part of this work CFD modelling was used to understand the behaviour, making it one of the earliest examples of the successful application of CFD modelling. The National Bureau of Standards, now NIST in 1980s started the development of CFD models for various fire applications. The present Fire Dynamics Simulator, FDS, is actually a consolidation of these models (Smardz 2003).

CFD technique is flexible, which means it can predict both smoke and flame movement in case of Fire. As these CFD models, or field models, are based on the fundamental laws of physics rather than empirical correlation i.e. based on set experimental conditions, they are the most versatile approach available with Fire Safety Engineers. CFD models are now being increasingly used in fire protection engineering to predict the movement of smoke in complex enclosed spaces such as atria, shopping malls and warehouses. (Gobeau et al. 2002)

Theory

The partial differential equations governing the fluid motion are also known Navier-Stokes equations. They cover three fundamental principles of:

• Conservation of Mass
• Conservation of Momentum
• Conservation of energy

Where, ρ is the fluid density, V is the fluid velocity vector, τij is the viscous stress tensor,

p is pressure, F is the body forces, e is the internal energy, Q is the heat source term, t is time, Φ is the dissipation term, and .q is the heat loss by conduction (Ashgriz & Mostaghimi n.d.)

For solving these equations, they are first made discrete to produce a numerical analogue. Then the entire domain is divided into small grids or elements. Finally, the initial and the boundary conditions of the specific problem are used to solve these equations.

CFD Result Analysis

CFD is a powerful technique since it can provide an approximate solution of the three-dimensional equations that govern fluid flows. However, it remains an approximation and in no circumstances is the solution obtained exact.

Numerical and physical errors are introduced at virtually all the stages of a CFD simulation since assumptions and approximations are made. Also, as the data generated by a CFD simulation can be huge, typically hundred of thousands of values for each time step recorded, it can be quite challenging to extract the specific information required. The CFD user has to use a post processor to analyze this result, in turn using interpolation methods, and this introduces some numerical errors in this stage too.

However, it is possible to control these errors with a good understanding of both fire science and CFD, i.e. a thorough knowledge of the capabilities and limitations of the physical and numerical sub-models as applied to fires and smoke movement. Also, the accuracy required from a CFD simulation depends, on the purpose of the simulation. It can be used for finding the direction of smoke, or quantitative values like heat flux etc.

Standards For the CFD fire Models

Presently, CFD models are widely use by the fire engineering people in solving practical problems in fire safety design of buildings. Hence, it becomes crucial that the accuracy of the predictions and also the appropriateness of a CFD model for a particular fire engineering problem can be reliably assessed (Smardz 2006).

For this a uniform standard needs to be developed, and already efforts to develop a uniform system of standards for fire field models are under way. For example, in Europe, a program was started by the Fire Research Division of the Office of Deputy Prime Minister in the UK (Smardz 2006).

In the United States also, an ASTM standard for evaluating fire models has been developed. This standard provides guidance on methodology for evaluation the predictive capabilities of fire models, including CFD models (Smardz 2006).

FDS (Fire Dynamics Simulator)

This section gives a brief description of the FDS tool that has been used in this project. The information brief given here is taken from the FDS User’s Guide and Technical reference manual.

About FDS

The complete name of the FDS is the NIST Fire Dynamics Simulator. FDS is a Fortran 90 computer program that solves the governing equations of fluid dynamics, and Smokeview is its companion program written in C/OpenGL programming language that produces images and animations of the results. The first version of FDS released in 2000. Since then several major improvements and new features were implemented in the program. (McGrattan 2006)

FDS is a CFD model of fire-driven fluid flow. The software solves numerically a form of the Navier-Stokes equations appropriate for low-speed, thermally-driven flow with an emphasis on smoke and heat transport from fires. Its companion program, Smokeview is a visualization program that is used to display the results of an FDS simulation (McGrattan 2006)

FDS Can be Used to Model the Following Phenomena:

• Low speed transport of heat and combustion products (mainly smoke) from fire
• Heat transfer between the gas and solid surfaces
• Pyrolysis
• Fire growth
• Flame spread
• Activation of sprinklers and heat detectors
• Fire suppression by sprinklers

FDS is widely used by fire safety professionals. One of the major applications of the program is for design of smoke control systems and sprinkler activation studies. FDS was also used in numerous fire reconstructions including the investigation into the World Trade Centre disaster (McGrattan 2006)

FDS Features

This section presents the theoretical basis for FDS. The physical model is first presented and then the numerical algorithm used to solve this equation is given

Hydrodynamic Model – An approximate form of the Navier-Stokes equations appropriate for low Mach number applications is used in this model. The approximation involves the filtering out of acoustic waves while allowing for large variations in temperature and density. This gives the equations an elliptic character, consistent with low speed, thermal convective processes.

FDS solves this type of equation by using an algorithm, whose core is based on an explicit predictor-corrector scheme, second order accurate in space and time. The computation can either be treated as Direct Numerical Simulation, DNS, in which the dissipative terms are computed directly, or as Large Eddy Simulation (LES), in which the large-scale eddies are computed directly and the sub-grid scale dissipative processes are modeled.

The choice of DNS vs. LES depends on the objective of the calculation and the resolution of the computational grid. If, for example, the problem is to simulate the flow of smoke through a large, multi-room enclosure, it is not possible to resolve the combustion and transport processes directly. However, for small-scale combustion experiments, it is possible to compute the transport directly and the combustion processes to some extent.

Combustion Model – There are two types of combustion models used in FDS, depending on the resolution of the underlying grid:

Mixture Fraction Combustion Model – It is sased on the assumption that large-scale convective and radiative transport phenomena can be simulated directly, but physical processes occurring at small length and time scales must be represented in an approximate manner. This uses LES calculation, and is used for most applications

Finite rate chemical reaction – In a DNS calculation, the diffusion of fuel and oxygen can be modeled directly, thus it is possible to implement a relatively simple one-step chemical reaction. The implementation of one-step reaction schemes is still very much a research exercise because it is not universally accepted that combustion phenomena can be represented by such a simple mechanism (McGrattan 2006)

Thermal radiation Model – Radiative heat transfer is included in the model via the solution of the radiation transport equation for a non-scattering gray gas, and in some limited cases using a wide band model. The equation is solved using a technique similar to finite volume methods for convective transport, thus the name given to it is the Finite Volume Method (FVM) (McGrattan 2006)

Geometry and use of multiple meshes – FDS approximates the governing equations on a rectilinear grid. The user prescribes rectangular obstructions that are forced to conform to the underlying grid. The grid cells can either be uniform in size or they can be stretched in one or two of the three coordinate directions. It is possible to use more than one rectangular mesh in a calculation. This allows creation of an efficient computational domain for geometries which can not be easily fitted into a single rectangular grid. It also allows using regions with different grid resolutions within one computational domain.

Both the grid stretching and the use of multiple meshes allow the user to apply better grid resolutions in critical areas i.e. near the fire without unnecessarily increasing the demand for computational power by applying fine mesh to the entire computational domain.

Boundary Conditions – All solid surfaces are assigned thermal boundary conditions, plus information about the burning behavior of the material. Usually, material properties are stored in a database and invoked by name. Heat and mass transfer to and from solid surfaces is usually handled with empirical correlations, although it is possible to compute directly the heat and mass transfer when performing a Direct Numerical Simulation, DNS

Input data required to run the model – All of the input parameters required by FDS to describe a particular scenario are conveyed via one or two text files created by the user. These files contain information about the numerical grid, ambient environment, geometry of the problem modelled, material properties, boundary conditions and the fire itself. The input file should also contain information about the desired output quantities (McGrattan 2006)

Liquid Fuels in FDS

For a liquid fuel, the thermal parameters are the same as those of a thermally-thick solid. The evaporation rate of the fuel is governed by the Clausius-Clapeyron equation. The only drawback of this approach is that the fuel gases burn regardless of any ignition source. An example of a liquid fuel is methanol.

The rate at which liquid fuel evaporates when burning is a function of the liquid temperature and the concentration of fuel vapor above the pool surface. Equilibrium is reached when the partial pressure of the fuel vapor above the surface equals the Clausius-Clapeyron pressure

Where, hv is the heat of vaporization, Mf is the molecular weight, Ts is the surface temperature, and Tb is the boiling temperature of the fuel.

For simplicity, the liquid fuel itself is treated like a thermally-thick solid for the purpose of computing the heat conduction. There is no computation of the convection of the liquid within the pool.

PyroSim

PyroSim is an interactive, graphical user interface for the Fire Dynamics Simulator, FDS. PysoSim gives the flexibility to the user, to work in either metric or English units, which can be switched at any point (PyroSim 2006)

PyroSim solves the Navier-Stokes equations appropriate for low-speed, thermally-driven flow and emphasizes smoke and heat transport in fires. Since the approach is based on fundamental physics, it can be applied to fires ranging from stove-tops to oil storage tanks. The PyroSim interface provides immediate input feedback and ensures the correct format for the FDS input file. Pyrosim is closely integrated with both FDS and SmokeView. (PyroSim 2006)

PyroSim was developed under the guidance of Dan Gemeny, Vice President, Fire Engineering, of Rolf Jenson and Associates, who provided that fire modeling skills. The software was developed by D. Swenson and B. Hardeman of Thunderhead Engineering consultants USA. (Interflam 2007)

Pool Fires

Pool fires have been an object of intensive studies over few decades. Essential and the most important feature of pool fire behavior is flame feedback to the surface of the liquid fuel. This feedback supports sustained flame above the pool surface and controls the burning rate of the fuel. For further development of CFD models, detailed data are needed on the structure of the flame feedback.

CFD modeling of pool fires has been attempted in a number of studies. These investigations demonstrated encouraging predictions for both confined and unconfined pool fires. For example, Sinai and Owens obtained good predictions of flame shapes for kerosene pool fires, subjected to cross-wind. However, most of the computational studies specify fuel release rate as a model parameter. Alternatively, simplified empirical correlations can be used to estimate fuel burning rate. Comprehensive CFD model of pool fire must predict fuel burning rate as part of the overall solution. Therefore, full coupling between the gas and liquid phases is required. Also, in large pool fires, soot plays a major role in the overall heat transfer through radiation and therefore, in the dynamics of the fire.

In this project the sample calculations are performed for kerosene and heptane pool fires.

Mathematical Model

• Gaseous Phase – The principal governing Favre-Averaged Navier-Stokes equations, which describe the turbulent flow of air, fuel and combustion products using the k-ε closure, assume the general form
• Combustion rate of fuel in the gaseous phase is modeled via an Eddy Break-up expression which involve Favre-Averaged mass fractions of reactants
• For radiation modeling, the Discrete Transfer Method is used, and the following RTE, Radiative Transfer Equation is solved for the radiation intensity:
• The above equation describes the change in radiation intensity with distance due to absorption and emission of radiation by the gas and soot particles along the path.
• Equation (3) describes the change in radiation intensity with distance due to absorption and emission of radiation by the gas and soot particles along the path.

Where, s is a specified soot conversion factor.

Liquid pool vaporization- At the time of pool fire, liquid fuel is vaporized from the surface before it mixes with the surrounding air and burns as a diffusion flame. To obtain the fuel release rate, the following one-dimensional heat transfer equation is solved in the direction normal to the liquid’s surface.

Here, liquid surface regression is taken into account to solve the equation in the region with the moving boundary. The surface regression rate is taken to be the same across the whole pool surface area, and is calculated as

Where, S is the pool surface area (Novozhilov & Koseki 2004)

Phase coupling conditions – The main method to provide sustainable combustion is the flame heat feedback to the surface, as shown in figure below. This has radiative and convective components. Re-radiation from the surface is considered as a part of the net radiation flux.

The energy which is transferred to the surface is partly spent on fuel vaporization, and partly is conducted to the inner layers of liquid. In this case, temperatures of liquid and gas must be equal at the interface. This model assumes variable fuel surface temperature; hence, two additional conditions at the interface are needed to close the problem.

The first condition is the heat balance across the interface, which is written in the form:

Where, Qf is a total (convective and radiative) feedback from the flame. The re-radiation component is also included in Qf. The total flame feedback is found upon resolving flow and radiation fields in the gaseous phase.

Fir deriving the second condition, local saturation equilibrium is assumed at the fuel’s surface. This means that the fuel vapor pressure at the interface is a given function of the surface temperature

Saturation pressure can be found using interpolation through look-up tables for any particular fuel. The mass concentration of fuel vapor at the surface can be found as

Compartment Fire Model – In order to calculate or predict the temperatures generated in a compartment fire, a description or model of the fire phenomena must be created. Such a model is, an idealization of the compartment fire phenomena. Consider a fire which starts at some point below the ceiling and releases energy and products of combustion. The rate at which energy and products of combustion are released may change with time. The hot products of combustion form a plume, which, due to buoyancy, rises toward the ceiling. As the plume rises, it draws in cool air from within the compartment, decreasing the plume’s temperature and increasing its volume flow rate. When the plume reaches the ceiling, it spreads out and forms a hot gas layer which descends with time as the plume’s gases continue to flow into it. There is a relatively sharp interface between the hot upper layer and the air in the lower part of the compartment. The only interchange between the air in the lower part of the room and the hot upper layer assumed is through the plume. As the hot layer descends and reaches openings in the compartment walls (e.g., doors and windows), hot gas will flow out the openings and outside air will flow into the openings. This description of compartment fire phenomena is referred to as a two-layer or zone model. The basic compartment fire phenomena are shown schematically in the figure below.

The two-layer model concept assumes that the compositions of the layers are uniform. That is, the temperature and other properties are the same throughout each layer. Although the temperature of the lower layer will rise during the course of the fire, the temperature of the upper layer will remain greater and is of the most importance in compartment fires. The assumptions may be less valid for very large spaces or for long, narrow spaces such as corridors and shafts (Drysdale 1999)

Ventilation Controlled Compartment Fire

Fires usually have either an excess of air or an excess of fuel. When there is an excess of air, the fire is considered to be fuel controlled. When there is more fuel present than air, a condition that occurs frequently in well-developed room or compartment fires, the fire is considered to be ventilation controlled (Drysdale 1999)

In a ventilation-controlled compartment fire, the combustion inside the compartment will be incomplete. The burning rate will be limited by the amount of air entering the compartment. This condition will result in unburned fuel and other products of incomplete combustion leaving the compartment and spreading to adjacent spaces. Ventilation-controlled fires can produce massive amounts of carbon monoxide.

If the gases immediately vent out a window or into an area where sufficient oxygen is present, they will ignite and burn when the gases are above their ignition temperatures. If the venting is into an area where the fire has caused the atmosphere to be deficient in oxygen, such as a thick layer of smoke in an adjacent room, it is likely that flame extension in that direction will cease, although the gases can be hot enough to cause charring and extensive heat damage (Drysdale 1999)

Ghosting Flames

Also known as dancing flames, these are not attached to the fuel source and move around an enclosure to burn where the fuel/air mixture is favorable. Such an occurrence in an under-ventilated situation is a sure sign that precedes back draft. This phenomenon was observed only for Heptane, when it was burnt under restrictive ventilation conditions, typically Kin = 0.06% and 0.14%, not for any other fuels. (Chitti 1994)

During the latter stages of the fire, typically around the time of peak fire intensity, flames were observed to flow over the sides of the vessel walls and travel along the weighing platform towards the air distribution slits on the floor, leading to the air distribution plenum. In some cases the flames detached from the pool altogether i.e. vertical flaming had been replaced by an unstable quasi-horizontal flaming).

Similar behaviours were reported by Audouin et al. (1997) who observed that, the entire flame detached from the surface of the pool and travelled to an opening, where it burnt at the entrance. Sugawa et al. (1991) describe a flame detaching itself from the fuel source and burning just under the ceiling as a pale blue aurora, approximately circular and flat, which in some cases reattached to the fuel source. Researchers such as Rasbash and Stark, and Gross and Robertson (Tewarson, 1972), Fleishmann & Parkes (1993, 1994) inter-alias have also observed this phenomenon,

Typically the occurrence of such phenomenon occurs under restricted ventilation conditions, where the upper-layer may extend throughout the majority of the compartment and the layer-composition is likely to be fuel-rich, and also low O2 concentrations. Although descriptions of ghosting flames typically suggest burning under the ceiling rather than the layer-interface, there is a certain similarity to the layer-burning studied by Gottuk et al. (1992) and as such there may be a similarity to the onset of flashover, which is sometimes described as the point at which flames may be observed to emerge from ventilation openings. It should be noted that, while the ghosting flame phenomenon is ventilation related, flashover is not.

These dancing flames are thought to cause large thermal instabilities in the compartment increasing the mixing of fuel and air.

Description of Experimental Set up

Fire Rig

The fire Rig, modelled in this project was built by Deansfield Fabrications in 1996, with sponsorship from Leeds University Academic Development Fund. Further modifications were made in the original rig in the later years. For example, in 1998, both the air inlet plenum was modified to allow the used of different air inlet openings and the vertical thermocouple trees were added to allow vertical temperature-profiles to be examined. Another modification was that of the air distribution plenum in 1999, to allow the study of forced ventilation fires and the corresponding air regulation/control valves added in addition to air mass flow metering (Ledger 2003)

The fire-rig shown above, consists of three identical air sealed steel two-level structures, each consisting of a 1.4m x 0.96m x 1.25m high fire compartment and a 1.4m x 0.96m x 0.25m high air-distribution plenum situated below. The fire compartment is internally lined on three sides with 25mm thick Triton Kaowool 1260 insulation board. On the fourth side, the observation window is fitted which is externally lined with an outer shutter of the same material (Ledger 2003).

Air entered into the fire entered the fire compartment via 30mm wide ventilation distribution slots. These ran along each of the four sides of the steel floor. There were no air gaps in the corner regions so as to support the floor. The air distribution plenum was then fed by a variable width inlet port situated on its front, whose area remained less than what was provided by the floor’s ventilation slots. This was done to allow the inlet port to always control the air-flow rate into the compartment and thus be able to provide the various ventilation conditions required for study (Ledger 2003).

Also, a pressure regulated compressed air supply, which went through an Endress & Hauser T-MAS AT533 mass flow meter, was connected to the rig. This was to allow forced-ventilation compartment fires to be examined. A 152mm diameter exhaust-port was situated above a suspended ceiling. This was located centrally on the roof of the fire compartment and housed a multi-hole ‘X’ gas-sample probe with 36 sample holes on centres of equal area, so as to allow sampling of the fire-gases. (Ledger 2003)

After the test, the products of combustion having passed through the sample port are extracted into a dump volume. Here, a large hood and extraction system transported these fire products to the atmosphere through a 10m high, 381mm diameter vertical chimney. Extraction was provided by a 381mm bifurcated fan, manufactured by Halifax Fan Manufacturing Co. Ltd., which provided a free air-flow rate of »0.8917m3.s-1 (Ledger 2003)

Temperature Measurements – Temperatures within the fire compartment are monitored by 26 type K mineral insulated exposed hot junction, 1.5mm bead, 613 stainless steel sheathed thermocouples from TC Ltd. The thermocouples were divided into:

Rear vertical thermocouple tree – which consisted of 10 thermocouples each separated by 0.1m. The lowest thermocouple was 0.15m from the compartments floor and the highest was 1.05m from the floor, with the tree located centrally on the rear compartment wall and the thermocouple beads were located 0.175m from the rear wall.

Central vertical thermocouple tree – This was situated vertically above the test fires. It consisted of 7 thermocouples, each separated by 0.1m, the lowest was 0.45m from the compartments floor and the highest was 1.05m above the compartments floor.

Gas Analysis

In the rig, the combustion products leaving the fire compartment are sampled using a multi-hole ‘X’ sample probe. The gases are then transported to a stack of analysers through a stainless steel filter assembly via a 10m filtered Teflon transmission line. The line is then electrically heated to 180oC, above the dew points of sample gases. This is so as to prevent the condensation of water and hydrocarbons, at time of transmission to the analysers. The sample is first passed into a heated oven containing a pump and filter maintained at 180oC. The it is analysed for Unburnt hydrocarbon (UHC), Oxides of nitrogen (NOx), Carbon dioxide (CO2), Carbon monoxide (CO), Oxygen (O2).  After this, the sample is passed through refrigeration cooler to remove water vapour, before entering the analysers which are responsible for the measurement of CO2, CO and O2  Fig 7 below shows a schematic (Ledger 2006).

Experimental Procedure

The above set up described is taken from a set of experiments done by Jon Ledger, a student of University of Leeds, as a part of a much larger program. His work included the burning of Heptane samples, Kerosene samples, Polyethylene samples, and Pine samples. Using the fire rig described above, these samples where burnt in different condition of fuel quantity and ventilation. The ventilation conditions include a wide range conditions, ranging from forced ventilation to very small natural ventilation. The samples were burnt in classified tests and the temperature profiles and the gas analysis was recorded for every sample burnt. The total number of tests carried out in the original project is well exceeding of 100 tests.

However, in this project the only test numbers used are – 1 & 34 for Kerosene, and 101 & 102 for Heptane. These tests will be explained in detail in this section.

All the test cases used in this work were from the naturally ventilated test. Here, the amount of ventilation is expressed in terms of the air inlet coefficient Kin , the air in leakage area divided by the cross sectional area of the cubic room of equivalent volume of the test room (Andrews et al, 1999). This can be expressed in the following mathematical form:

A 0.6 kg kerosene sample was prepared in a 20 × 20 cm mild steel dish. This dish was placed on a 0.7 by 0.7 m piece of thick triton kowool, which rested on the load cell tray. The ignition was done by heating the surface of fuel using a butane gas torch. After ignition, the compartment door was instantly closed and the recording of data began. This recorded data includes oxygen, carbon monoxide, and carbon dioxide levels. The temperature thermocouples where also recorded every 5 seconds. The main feature of this test was the amount of ventilation, where Kin is 1.09%. This means an air inlet area of 16 × 10-3 m3 with the dimensions 0.095m × 0.169m.

The procedure here is the same as in the Test 1 above. There was one modification; the air inlet coefficient Kin was now 0.55%. This meant an air inlet area of 8.03 × 10-3 m2 with the dimensions 0.047m × 0.169m.

A 0.35 kg Heptane sample was prepared in a 20 cm × 20 cm mild steel dish. The dish was then placed on 0.7 × 0.7 m piece of thick triton kowool, which rested on the load cell tray in the combustion room. The Heptane pool was ignited using a wooden spill. The compartment door was closed instantly and the gases and temperature values were recorded every 5 seconds. The air inlet coefficient Kin used in this test was 1.09%. This meant an inlet area of 16 × 10-3 m3 with the dimensions 0.095m × 0.169m.

The procedure here is the same as that for Test 101. There was one modification done for the ventilation condition, air inlet coefficient Kin was changed to 0.55%. This meant an air inlet area of 8.03 × 10-3 m2 with the dimensions 0.047m × 0.169m.

The Model Design Procedure

The fire rig was constructed using the most appropriate materials provided by Pyrosim, which is described in the background section. However, many modifications were done on some properties to try and simulate the test described above as exactly as possible. The FDS input file was then exported from Pyrosim and the input file was used to run the FDS simulation. The design procedure and the experiment are explained in detail in this section.

Mesh Design

The design and validation of such a complicated design, as described above, needs a very accurate mesh to recognise the details. At the same time the number of cells should be kept as minimum as possible to lower the simulation run time. This is best achieved by using more than one grid. The most accurate grid is used for the combustion room, and grids with lower cell numbers are used for less important parts of the design.

As shown in the figure above, there were four grids defined in this model. The main grid is in the middle and is the most accurate, with dimensions 0.96m width × 1.6m depth × 1.6m height and 24 cells wide × 40 cells deep × 40 cells height. The total number of cells with the dimensions of 4 cm × 4 cm × 4 cm each is 38,400. The three other grids were made of larger cells, which meant less accuracy, to lower the number of unimportant calculations made by FDS. The total number of cells in the design was 49,280 cells.

Base, Rooms, and Extraction Hood

After the mesh was designed the model structure was divided into three main groups to make the design simpler. As in the figures below, the first group defined was named as the Base where tow large chunks of steels were assumed to fill the area under the unused compartments on the left and right of the combustion room. At the same time a void was created under the combustion room to allow the air movement and air entrainment.

The left bulk of the base starts from the point -0.96 to the point .02 and the right bulk started at 0.95 to the point 1.92 on the X axis. The two start from the point 0.1 and end at 1.5 on the Y axis. Also, a 2 cm walls where used to close the void under the combustion room. The base was 0.25m high on the Z axis and all its parts were made of steel.

The second Group, is the group of rooms, where again two large chunks of steel were defined on the right and left. They both were around 0.96 meter wide with the dimensions starting -0.96 to .02 and 0.95 to 1.92 on the X axis. Both started from 0.1 to 1.5 on the Y axis. The height was 1.25m on the Z axis starting from the point 0.25 to the point 1.5. Again all this bulk was made of steel and the space in between was the actual combustion room, which will be described later in this section.

The last group was the extraction hood which is again made of steel and is on two levels, one at the back with extraction fan in the middle of it, and the largest part in the front with a slope in the roof. The hood starts from the point -0.96 to 1.92 on the X axis and 0.10 to 1.5 on the Y axis.  The back level starts from 1.6 to 2.11 on the Z level, while the front level is from 1.6 to 1.8. The slope between the front and back was designed using two large triangles to simulate the 5 cm difference between the front part and the back part. Finally the combustion room door is hidden so that the combustion can be seen in smoke view after the simulation is done.

The Combustion Room

This is the most important part of the design. A lot of time was spent to simulate an accurate replica of the compartment described in the fire rig section. First, the dimensions of the room were made close to the actual compartment, which is described in the report to be 0.96 × 1.4 × 1.25. But when the room is examined, taking into consideration the insulation used, a few cms are lost from each of the measurements.

The base room design is 0.02 to 0.95 on the X axis, 0.1 to 1.5 on the Y axis, and 0.25 to 0.27 on the Z axis, resting above a 25 cm void. This base combustion room floor was also fitted with 3 cm wide air slits. These air slits were 55 cm long and were along the four sides of base. The room walls were made of Gypsum Board, which was the closest material to be found in the materials library, to simulate the isolation board in the fire rig. The floor and the ceiling and the dish were all made of steel. A suspended ceiling was designed at about 6 cm from the roof and was made of Gypsum Board. At the top of the roof a 16 cm × 16 cm hole was made to simulate the exhaust port. Gas quantity thermocouples were fitted in the design just under the hole, to simulate the gas analyser used in the fire rig.

Moreover, the temperature thermocouples trees, explained earlier in the fire rig section, were all designed at the same heights and dimensions in the model. The figure is as shown below:

Fuel Sample & Reaction

The Fuel samples were modelled, as described in the actual experiments. However, as there is no circular shape in FDS, 20 cm × 20 cm× 4 cm square samples were designed. Another dimension problem was that, since the main Grid was made of 4 cm × 4 cm × 4 cm cells, FDS could not recognise the fuel if its height was less than 4 cm. This resulted in the samples becoming higher with fuel quantity, however, this option was chosen to keep the pool features as close to the 20 cm diameter dish as possible.

In FDS, there is a reaction library. For liquid fuels if the reaction is chosen, the samples start burning directly from the start of the simulation. The reactions used were – Kerosene reaction for test 1 and 34 and the Heptane reaction for tests 101 and 102. The only change made to the fuel properties was the maximum burning rate which exceeded the values for a small pool fires. For everything else the standard values provided by the developers were used including the reaction parameters.

The Runs

The runs of the simulation were done using a personal laptop with 1 Giga Ram, and 100 Gigabyte hard disk. Each simulation took 14 hours to complete. For 480 seconds of simulation, the total data obtained was about 5 Giga bytes. This, combined with the trials for this experiment and previous trials on the model, comes out to a total data of more than 11 Giga bytes.

Results & Analysis

In FDS a large amount of data is produced as the result of the simulation. However, in this section few chosen data is presented in graphs with the Actual tests data. This was done to give a clear and easy comparison. The Graphs include Upper lever temperature, Heat Release Rate, O2 concentration, CO2 concentration, and CO concentration.

It can be shown from Graph 1 that the peak temperature is about the same in both, the Actual test and the modeled test. However, the modeled test reaches the peak levels at 500 C in about 30 seconds, while the actual test needs about 300 seconds before it reaches 500 C. This is because, the modeled test fire growth is much faster then the actual test for many reasons, which will be explained in the discussion section.

Graph 2 above, shows the Heat Release Rate in both the modeled and Actual Tests. In the modeled test, the peak HRR is about 95 kW and is achieved in less than 10 seconds. While in the actual test the peak HRR is the same, it is reached after 360 seconds from the start of the combustion.

Graph 3 shows the O2 concentration in both the Actual and modeled tests. It  clearly shows that the O2 Levels drop down to about 5% in less than 60 seconds, while in the actual test it drops down to a similar level gradually and reaches the same level after about 360 seconds.

Graph 4 shows the CO2 levels, which increases from 0 to 10% in less than 60 seconds, while the actual test curve shows that the CO2 level increase gradually and reach 10 % after 300 seconds.

Graph 5, for the Test 36, shows the CO concentration in the compartment. This was less than 0.2% in the modeled test, and reached about 1.8% in the actual test
Test 3 (Kerosene 0.55%)

Graph 6 shows the Upper layer temperature in the actual and modeled designs. It clearly shows a rapid increase in the modeled design in the first 50 seconds to reach 400 degrees, which was followed by a sudden increase to 100 degrees before it vibrates in the range of 100 and 200 degrees. However, the actual test graph shows a gradual increase until it reaches the peak temperature at 400 degrees after 450 seconds.

In addition, the Heat Release Rate shown in Graph 7 indicates a very rapid increase in the modeled test in the first 10 seconds. But soon after when the modeled test reaches its peak of 90 kW, the HRR drops down to around 10kW. At the same time, graph of the actual test shows a gradual increase in the HRR until it reaches its peak about 85 kW after 460 seconds.

Graph 8 shows a rapid decrease in O2 levels in the modeled test to about 0 in less than 50 seconds, while the actual test curve shows a slower decrease over time till it reaches 5% after about 450 seconds.

Graph 9 shows the CO2 concentration in the exhaust; where the modeled test over predict the CO2 and again it peaked at 12 % very early in less than a minute. At the same time, in the graph of the actual test CO2 levels where increasing gradually until it reached its maximum at 10 % after 480 seconds.

Finally, Graph 10 shows the slight increase of CO levels in the modeled test and the higher increase in the actual test. In the modeled test the values stopped increasing at 0.1 %, but the actual test produced a maximum concentration of 1.5 % after 480 seconds.

In test 101, Graph 11 shows the average upper layer temperature jumping from 20 C to 400 C in less than a minute where it becomes almost constant during the experiment. The same graph shows a steady increase in the actual test upper layer temperature where it increases over time and reaches a peak at 500 C.

Graph 12 shows the Hear Release Rate in both the actual and modelled 101 test. It shows the model with a huge and fast increase in the early seconds of the compartment reaching the range of 60-70 kW in few seconds. While the actual test started with a fast increase to 20 kW in the first 20 seconds, but a more smooth increase after that until it reached 60 kW at 420 seconds mark.

Graph 13 shows the O2 levels in the exhaust of the compartment, where it shows a rapid decrease in the O2 levels and stabilise at 10 %. However, the actual test showed slower and smoother decrease overtime and reached a low at 9 %.

Graph 14, shows the concentration of CO2 in the exhaust where the model test curve shows a rapid increase to reach 7 %, but in the actual test this was not the case. In the actual test CO2 levels increased smoothly over the time of the experiment to reach a peak of 9 % after 420 seconds.

Finally, Graph 15 shows the CO levels in the modelled test increasing to a very negligible value of 0.02%, while in the actual experiment it reached a high of 0.7 %.
Test 83 (Heptane 0.55%)

Graph 16 shows the Average upper layer temperature in the actual and modelled kerosene test number 83. It shows a fast increase in the modelled test upper layer temperature which reached 400 C in about one minute and then started decreasing to become steady at 200 C. However the actual test curve showed another pattern where, the temperature curve started a smooth increase from ambient at 20 C to a little less than 500 after 480 seconds.

Graph 17 shows a huge increase in the Heat Release Rate in the modelled test. In the first 10 seconds the HRR of the combustion reached about 60 kW before it starts decreasing and vibrate between 10 and 20 kW. However, the actual test curve shows a small rapid increase from ignition to 15 kW in few seconds, then a steady smooth and slow increase in the HRR until it reaches 50 kW after 480 seconds.

In graph 18, the O2 levels are shown, where in the modelled test oxygen levels fall to zero in about 2 minutes. But, at the same time the actual test results show the Oxygen levels decrease smoothly until it reaches a low at 5%.

Graph 19 shows the CO2 levels. Here modelled test showed a rapid increase in CO2 levels before it stabilise ate 12 %. But, this was different to the actual test result wish indicate a slower increase over time before it reaches a peak at 10 %.

Finally Graph 20 shows the CO levels which were very small in the modelled test. While the actual test was not very different in the beginning, but it increased rapidly after 300 seconds to reach a peak of 0.8 %.

Discussion

This section focuses on the validity of the modelled test and the factors affecting the accuracy in detail. There are many areas of possible errors and faults when the data described in the previous section is considered. Some of the points covered include the input file, the accuracy of the data, and the theoretical basis of FDS etc.

First of all, the input file, as described in the FDS section, contains a large amount of data comprising of the materials used and their properties; information about the mesh; the air entrainment and the reaction properties. This information, which is normally given by the user, is used in the FDS equations and multiple calculations, before the final output files are produced. However, in this project most of the input file data was built using the Pyrosim data Libraries. Here, the materials and reactions libraries were provided by the developer, with some changes done to this data, as will be explained later.

A broad look on the results section (where the output data are compared with the actual data), shows a relatively better performance in the modelling with the higher ventilation conditions. It was clear that the major problem in the existing model was – over-predicting the growth rate in all the tests. Better models will be discussed before moving on to the two bad models at the end of this section.

It is conventional to state the tests with higher ventilation performed relatively better, but with the same major problem of over predicting the growth rate in the early stages. This include tests 36 and 101 where the Kerosene and Heptane were burnt in the Kin=1.09% conditions. Both tests, number 36 and 101, showed a very good prediction in terms of peak temperature and peak heat release rate. They also showed good prediction of O2 and CO2 levels at the end of the experiment. At the same time the CO prediction was much less than expected. However, the development of the pool fire over time was very far from the actual scenario. Usually, the modelled pool fires showed a very fast growth rate before it stabilises at constant level.  This was not the case in the actual test, where the fire continued to grow relatively slower over time until it reached the peak values, and then started to decline. Accordingly, the gases levels in the modelled tests excluding CO were going to the critical values very fast. O2 drops much faster than the actual, while CO2 production increases dramatically in the first few seconds.

An investigative analysis to the input file shows that few points were critical in the model design. First, the CO yield was set to be constant at 0.006 in the Heptane reaction and 0.012 in the Kerosene reaction, which resulted in a very bad CO results when measured in the model. These values were set by the reactions library developed by thunderbird and NIST, and was discovered at the late time of the project timeline when there was no time to run some corrected input files. Another important input in the reaction properties is the soot yield which was 0.012 for Heptane and 0.042 for Kerosene. They were also provided by the reaction library and discovered late, and so there was no chance for performing the corrected runs.

An interesting area in the input file was the materials properties. One of which was the maximum burning rate for both Heptane and Kerosene. The data provided by the materials library was 0.065 kg/(m2s) for Heptane and 0.039 kg/(m2s) for Kerosene. These were wrong data for small pool fires and were corrected using the following equation:

m'” = m'”inf(1 – e-kBD)

This results in value for both Heptane and kerosene as 0.02kg/(m2s). However, for the purpose of the accuracy of the model the maximum burning rate values recorded in the actual test were used in the model as given below:

Heptane: 0.032 kg/(m2s) and Kerosene: 0.039 kg/(m2s)

This resulted in very good results in terms of fire intensity, peak upper layer temperature, and maximum heat release rate.

Another important area to be considered was the thermal properties of the combustion room. The room’s walls were made of non flammable solid Gypsum Board with a wall thermal conductivity of 0.48 W/(m.k). This was to simulate the 25mm thick Triton Kaowool 1260 insulation board, used in the actual test. Also, the top and the base of the room were to be made of steel, the very same material used in the actual test. This can be one cause of the unexpected fast growth in the models, where the heat loss to the walls of the compartment is not as accurate as the actual scenario.

Adding to the above points, the mesh design can be another point of weakness in the design causing, the faster then predicted scenario. The size of cells used in the model was 4cm×4cm×4cm and this is not accurate enough for such an experiment. In the early trials 2cm ×2cm×2cm cells were used, and was run on a super computer for the first 30 seconds of one of the modelled tests The results showed an increase of about 10% in the temperature values when compared with the original mesh used in the project. Also, the relatively large size of cells has caused some problems when designing the fuel sample. The sample was modelled larger than the actual in order to be recognised by FDS. However, this accuracy was the best choice, as the accurate mesh resulted in very large number of cells and needed a much more powerful and faster computer.

One major cause of the unexpected growth rate could be the FDS code itself. According to technical guide the liquid fuel is treated like a thermally thick material and the convection is not computed for simplicity. Moreover, the ignition in the model assumes that the fire starts on the whole surface of the fuel, which is not the case in the actual tests. The guide also states “The rate at which liquid fuel evaporates when burning is a function of the liquid temperature and the concentration of fuel vapor above the pool surface. Equilibrium is reached when the partial pressure of the fuel vapor above the surface equals the Clausius-Clapeyron pressure”, i.e.

Where, hv is the heat of vaporization, Mf is the molecular weight, Ts is the surface temperature, and Tb is the boiling temperature of the fuel

This method can increase the growth rate if the gas temperature was over predicted. Adding to this the fact, the heat feed back to the surface of the liquid was also not well developed. All this resulted in over prediction of heat release rate levels and heat flux produced. (FDS Tech. Guide)

The two other tests had another problem, in addition to the over predicted growth rate. This was the run out of Oxygen state after few minutes of the test start. It is obvious from the graphs that the main cause of the sudden decline in the temperature levels and the heat release rate is the lack of Oxygen in the compartment. This could be linked to the fast growth rate, where – if the fire growth was gradual the Oxygen consumption would be much less and would give time to the air movement from the air inlet to the burning compartment. However, it may also be that the FDS prediction to the air movement was not very good for such a very small amount of ventilation and caused the lack of Oxygen in the compartment.

A very interesting phenomenon was observed in the lower ventilation tests. The presence of ghosting flames was an expected observation. At the peak of the fire the vertical flame started to vanish and some horizontal flames traveling to the air inlets on the sides of the base of the compartment was clearly observed. This continued through the rest of the experiment and matched the ghosting flames. Although this was not observed at all levels of ventilation in the actual tests, and was only observed in lower ventilation conditions.

Conclusion

In conclusion, the results obtained from the modelled test showed a relatively good prediction of peak upper layer temperature and heat release rates in two of the modelled tests. The better tests were the ones with higher ventilation conditions. It also showed a good prediction of the O2 and CO2 levels at the peak of the fires. However, the fire growth design was the main weakness in the modelled tests. The results comparison showed a very obvious over prediction in the models, where the fires grew up very rapidly in the first few seconds. This resulted in the over prediction of CO2 production and over consumption of O2 in the early stages of the tests. This proved vital in the models with lower ventilation where the O2 levels dropped to around zero and the lack of Oxygen caused a sudden decline after a relatively short time.

Consequently, it can be stated that FDS models can give a relatively good prediction of the fire intensity and peak temperatures in small pool fires. But, FDS also over predicts the pool fires growth and development. This is because of many factors including, the soot yield design and liquid fuel model used in FDS. However, this does not affect the proved abilities of FDS in fire protection and fire safety use, unless the conditions simulated includes a restricted ventilation conditions, where the over prediction can result in false simulation of fire self extinguishment.

In addition, the ghosting flames observed in the experiment proved the code to be able to predict unusual phenomena in closed compartment fires. This suggests further studies about these phenomena and others which are not easily observed in the actual studies such as back draught and flash over. Moreover, the smoke yields and gases production prediction is another area of interest for further investigation and modelling.

Finally, the main outcome of this project suggests a better consideration by the FDS developers to the small pool fires fire growth design. This includes the ignition of the pool, the convection in the pool and the heat feed back from the flames and gases to the pool surface. These areas are either simplified or over predicted in the designed model. All these areas are considered to be weakness in the code and the results of this project has proved it.

References

1. Drysdale D, ‘Introduction to Fire Dynamics’, 2nd ed., 1999, Wiley
2. Quintiere J.G, ‘Principles of Fire Behaviour’, 1997, Thomas Delmar Learning
3. McGrattan K, ‘FDS Technical Reference Guide – NIST Special Publication 1018’, March 2006
4. McGrattan K, Forney G, ‘FDS Users Guide – NIST Special Publication 1019’, March 2006
5. ‘PyroSim User Manual’, Thunderhead Engineering, 2006

Thesis

1. Ledge, J, University of Leeds, 2003
2. Novozhilov V, Koseki H, ‘Computational Fluid Dynamics Prediction of self-sustained Pool Fire Combustion’, 2004
3. Smardz P, ‘Validation of Fire Dynamics Simulator (FDS) for forced and natural             convection flows’, University of Ulster, 2006

Websites

1. Ashgriz N, Mostaghimi J, ‘An Introduction to Computational Fluid Dynamics’, Chapter 20 in Fluid Flow Handbook, http://www.mie.utoronto.ca/labs/MUSSL/cfd20.pdf
2. Chitti R, ‘A Survey of back draught – Main report’, May 1994, http://www.communities.gov.uk/pub/310/51994Asurveyofbackdraughtmainreport_id1508310.rtf
3. Gobeau N, Ledin H, Lea C, ‘Guidance for HSE Inspectors: Smoke movement in complex             enclosed spaces – Assessment of Computational Fluid Dynamics’, 2002 http://www.hse.gov.uk/research/hsl_pdf/2002/hsl02-29.pdf
4. Hurley M.J, Society of Fire Protection Engineers, 2006, http://www.fpemag.com/articles/department.asp?i=289
5. Interflam, 11th International Conference on Fire Science and Engineering, 2007             www.nfpa.org/assets/files//Interflam_2007_Brochure.pdf
6. PyroSIm – A Model Construction tool for FDS, Fire Dynamic Simulator, 2006             http://thunderheadeng.com/pyrosim/
7. Xing T, Stern F, ‘Introduction to Computational Fluid Dynamics (CFD)’, University of Iowa, 30th August, 2006, http://css.engineering.uiowa.edu/~me_160/Lab/CFD_grid_postprocessing_2006.ppt

——————–PyroSim-generated Section——————–

1. &BNDF QUANTITY=’WALL_TEMPERATURE’/
2. &BNDF QUANTITY=’HEAT_FLUX’/
3. &GRID IBAR=24 JBAR=40 KBAR=40/ MainGrid
4. &PDIM XBAR0=0.00 YBAR0=0.00 ZBAR0=0.00 XBAR=0.96 YBAR=1.60 ZBAR=1.60/ MainGrid
5. &GRID IBAR=36 JBAR=20 KBAR=8/ TopGrid
6. &PDIM XBAR0=-0.96 YBAR0=0.00 ZBAR0=1.50 XBAR=1.92 YBAR=1.60 ZBAR=2.14/ TopGrid
7. &GRID IBAR=10 JBAR=16 KBAR=16/ LeftGrid
8. &PDIM XBAR0=-0.96 YBAR0=0.00 ZBAR0=0.00 XBAR=0.00 YBAR=1.60 ZBAR=1.60/ LeftGrid
9. &GRID IBAR=10 JBAR=16 KBAR=16/ RightGrid
10. &PDIM XBAR0=0.96 YBAR0=0.00 ZBAR0=0.00 XBAR=1.92 YBAR=1.60 ZBAR=1.60/ RightGrid
11. &HEAD CHID=’kk109′ TITLE=’Kerosene_Leeds’/
12. &HOLE XB=0.0500, 0.0800, 0.55, 1.10, 0.2500, 0.2700 RGB=1.00, 1.00, 0.4000/ Left
13. &HOLE XB=0.87, 0.90, 0.55, 1.10, 0.2500, 0.2700 RGB=1.00, 1.00, 0.4000/ Right
14. &HOLE XB=0.2300, 0.73, 0.2100, 0.2400, 0.2500, 0.2700 RGB=1.00, 1.00, 0.4000/ Front
15. &HOLE XB=0.2300, 0.73, 1.37, 1.40, 0.2500, 0.2700 RGB=1.00, 1.00, 0.4000/ Back
16. &HOLE XB=0.4000, 0.56, 0.72, 0.88, 1.50, 1.54 RGB=1.00, 1.00, 0.2000/ Top Hole
17. &MISC REACTION=’KEROSENE’ DTCORE=5.00 NFRAMES=2800 RENDER_FILE=’kk109.ge1′ RESTART=.TRUE./
18. &OBST XB=-0.96, 1.92, 0.0800, 0.96, 1.82, 1.90 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.FALSE. THICKEN=.FALSE./ Top1 [Segment]
19. &OBST XB=-0.96, -0.95, 0.1000, 1.50, 0.00, 0.2500 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Right
20. &OBST XB=1.91, 1.92, 0.1000, 1.50, 0.00, 0.2500 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Left
21. &OBST XB=-0.96, 1.92, 1.49, 1.50, 0.00, 0.2500 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Back
22. &OBST XB=-0.95, 0.4000, 0.1000, 0.1100, 0.00, 0.2500 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ FrontL
23. &OBST XB=0.56, 1.91, 0.1000, 0.1100, 0.00, 0.2500 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ FrontR
24. &OBST XB=0.4000, 0.56, 0.1000, 0.1100, 0.1000, 0.2500 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ FrontTop
25. &OBST XB=-0.95, 0.0100, 0.1100, 1.49, 0.00, 0.2500 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ R1
26. &OBST XB=0.95, 1.92, 0.1100, 1.49, 0.00, 0.2500 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ L1
27. &OBST XB=-0.96, -0.95, 0.1000, 1.50, 0.2500, 1.50 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Right
28. &OBST XB=1.91, 1.92, 0.1000, 1.50, 0.2500, 1.50 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Left
29. &OBST XB=-0.95, 0.0100, 0.1000, 0.1100, 0.2500, 1.50 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Left F
30. &OBST XB=-0.95, 0.0100, 1.49, 1.50, 0.2500, 1.50 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ BackF
31. &OBST XB=0.95, 1.92, 0.1000, 0.1100, 0.2500, 1.50 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Front R
32. &OBST XB=0.95, 1.92, 1.49, 1.50, 0.2500, 1.50 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Back R
33. &OBST XB=-0.95, 0.0100, 0.1100, 1.49, 0.2500, 1.50 SURF_ID=’INERT’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ R1[1]
34. &OBST XB=0.95, 1.91, 0.1100, 1.49, 0.2500, 1.50 SURF_ID=’INERT’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ L1[1]
35. &OBST XB=0.0100, 0.95, 0.1000, 1.50, 0.2500, 0.2700 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Floor
36. &OBST XB=-0.96, 0.0100, 0.1000, 1.50, 1.50, 1.54 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ TopL
37. &OBST XB=0.95, 1.92, 0.1000, 1.50, 1.50, 1.54 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ TopR
38. &OBST XB=0.0600, 0.90, 0.2100, 1.37, 1.40, 1.43 SURF_ID=’GYPSUM BOARD’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Top 1
39. &OBST XB=0.3600, 0.4000, 0.68, 0.92, 1.54, 1.62 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Hole L
40. &OBST XB=0.56, 0.60, 0.68, 0.92, 1.54, 1.62 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Hole R
41. &OBST XB=0.4000, 0.56, 0.88, 0.92, 1.54, 1.62 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Hole B
42. &OBST XB=0.4000, 0.56, 0.68, 0.72, 1.54, 1.62 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Hole F
43. &OBST XB=0.0100, 0.95, 0.1000, 1.50, 1.50, 1.54 SURF_ID=’STEEL’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ TopMiddle
44. &OBST XB=0.0100, 0.95, 1.47, 1.50, 0.2700, 1.54 SURF_ID=’GYPSUM BOARD’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Back1
45. &OBST XB=0.0100, 0.0300, 0.1000, 1.47, 0.2700, 1.54 SURF_ID=’GYPSUM BOARD’ PERMIT_HOLE=.TRUE. SAWTOOTH=.TRUE. THICKEN=.FALSE./ Left