Critical Review of Wind Effects on Compartment Fire

Abstract

This paper is a review of the effect wind has on compartment fires. It begins with the basics of compartment fires including an explanation of smoldering, flash over and non-flash over fires. It goes on to describe a case study which was done in a seven-story building to determine the wind effects on fire in this structure. This study involved US government agencies recording the effects wind had on fires in a furnished room. It consisted of 14 experiments. Twelve of these experiments included natural or mechanical wind. A number of variables were investigated and it was found that increasing the wind on the fire increased its rate of spread and temperature. It was also found that the fire could be controlled through fire blankets and closing windows or doors. These findings were supported by research studies which are reported in chapter 4. The studies included fires with the wind in tunnels as well as apartments. A number of ventilation systems were used as well as both man-made and natural airflow. The findings of these studies supported those of the case study. Increasing the wind led to more rapid spreading of the fire and increased levels of heat. One of these experiments made the interesting observation that increasing the level of wind with sufficient ventilation hastened the fire extinguishing itself. Chapter 5 is an exploration of numerical investigations regarding fire and wind. These types of investigations involve computational fluid dynamics and complex mathematical algorithms. This type of analysis is done on computers. This type of analysis of fires is especially useful when there are insufficient resources to conduct actual experiments involving wind effects on a compartment fire. They are also useful in explaining the observations of case studies. The final section of this paper involves the observation that little research has been done on the finding that increased wind can cause a fire to extinguish itself more rapidly. Suggestions are made for further research impossible practical uses for this phenomenon.

Chapter One: Introduction

High-rise buildings and other large structures have become common in the 21st century. These types of structures provide additional risks for firefighters and occupants of the building. There are many risks involved in such buildings in the case of a fire. For example, there is increased travel distance to the exit which means a longer exposure time. There may also be a more complicated path to follow toward the exit as well as the potential for larger fires. Additional problems may be caused by the building's ventilation system as well as the presence of an external wind. This is especially common when high-rise buildings are concerned. There tend to be higher winds at increased levels of elevation. These increased winds can accelerate the energy release from a fire. Gases will also spread through the building at a higher rate.

Many fire departments are now using positive pressure ventilation (PPV) on smaller structures. It has been suggested that this same approach could be used to remove heat and smoke from a larger structures. Other tactics which are important for firefighters regarding large structures include using items such as smoke curtains or fire blankets to adjust the flow of air in the fire environment.

This paper includes a review of current research done in the field of wind effects on fires within different types of compartments. These compartments may include apartments, offices, or tunnels through which vehicles travel. The present research in the field will be reviewed and gaps in the literature identified. Chapter 3 involves a case study which was done of a large building which was set ablaze for research purposes. Significant amounts of wind were applied to the structure in order to determine its effects. There were 12 experiments done which involves applying natural or mechanical wind to the compartment which was on fire. A number of variables were investigated such as the effect of wind control devices and positive pressure ventilation fans.

The case study found that wind which was applied to the compartment fire increased its temperature and rate of burning. The wind also served to clear the superheated and noxious gases from the area. This means the wind can provide an increased level of safety for the emergency personnel in a burning building as well as occupants attempting to exit the area. The increased wind was provided by positive pressure ventilation fans. It was also found that fire blankets and smoke curtains could be used to control the airflow in the compartment area. Other methods of controlling the airflow included closing doors or windows. Fires which had clear path of airflow with even small amounts of wind burned at incredibly high temperatures. Some areas near the ceilings exceeded 1000°C. This is a fatal temperature even for fire personnel in full gear.

Chapter 4 will cover experimental research studies done primarily in tunnels or small structures such as apartments or offices. These studies served to reinforce the findings of the case study. The basic findings were that increasing the wind flow in a compartment will increase the rate in temperature of a fire. An interesting finding was that increasing the wind conditions also caused the fire to extinguish itself earlier (Huang et al., 2009).

Chapter 5 consists of numerical investigations of fire including the effects of wind. This type of analysis involves using computational fluid dynamics. Computers are used to do these complex mathematical calculations. These simulations have proven useful for determining the relationships between a variety of factors, including wind, and the rate with which a fire burns in a compartment. This type of analysis is useful when an actual burning of the compartment cannot be done due to limited resources. It can also be useful for understanding the findings of case studies.

The conclusion and recommendation section points out that there were the following findings:

• Increased wind accelerates the rate of fire spread within a compartment.
• Wind increases the temperature of a fire in a compartment as well as the gases released.
• Clear air flow through ventilation must be present for a large fire to proceed.
• Increasing the wind on a fire and a compartment causes it to extinguish itself at a higher rate.

The finding that increased wind causes a fire to extinguish itself more quickly has not been well researched and leads to the recommendations at the end of this paper.

Chapter Two: Basic Knowledge of Compartment Fire

2.1 Chapter Introduction

This chapter is meant as an introduction to compartment fires and their relevance to 21st century buildings. Most modern buildings consist of multiple compartments such as apartments or offices. The nature of fires in these types of settings is discussed. It is explained that there our flash over, non-flash over, and smoldering fires. There is also a discussion regarding the effect that airflow, wind, will have on these fires. While the remainder of the paper is primarily focused on how wind can affect compartment fires, it is important to understand the difference between these types of fires as their presence is frequently determined by the presence of wind or air flow.

2.2 Compartment Fire Basics

In many large modern buildings such as high-rises, these spaces are divided into a number of separate areas, which can be considered compartments. This is also true for structures such as apartment buildings. In these types of buildings, each of the compartments may be divided into rooms, which can be considered as further separation for more compartments. For example, a 20 story apartment building which has an average of 10 apartments on each floor with four rooms each could be conceptualized as having a minimum of 800 compartments. Hallways, stairways, common rooms, utility rooms, and maintenance areas are likely to add additional compartments to this type of building. An apartment building of this type might have a total of 1000 compartments when viewed from the standpoint of understanding its fire dynamics. The majority of fires which occur in large buildings with compartments began within an enclosed space. The initial flaming compartment may be small such as a single apartment unit. It can also be large such as a common room located in an office building.

There are a number of factors, which influence fire development within compartments. In addition to flammability characteristics of burning objects, there are environmental conditions. It should also be remembered that the fire itself changes the environmental conditions which exist within a compartment. This means that the fire in the compartment itself is intrinsically intertwined with regard to fire development. This is particularly true in smaller compartments where fires often develop quickly and can more easily change the environmental conditions. In larger compartments usually takes a longer period of time for the environmental conditions to be changed due to the fire. The actual length of time for environmental changes to occur within a compartment due to a fire depends upon the size of the compartment and the nature of the fire. In larger compartments there will be a significant length of time before a fire can change the local environmental conditions. Figure 1 below shows a typical compartment fire in a room containing furniture:

Figure 1 is meant to display a common situation within a large residential or office building containing compartments with furniture. In this case, the fire has begun in an upholstered chair. This could happen if an individual fell asleep while smoking a cigarette in the chair or dropped a small electronic device which short-circuited and caused substantial heat or sparks to ignite the chair. As the chair Burns it forms a plume of fire, which spreads toxic gases, smoke, and heat toward the ceiling of the compartment. There will be a hot layer of smoke, which forms near the ceiling. The hot layer will increase with regard to temperature and size. Eventually, this smoke layer will reach toward the door opening. This will allow the exhaust gases to spread quickly to other areas of the building. This could pose a substantial risk to other properties in the building. It can also cause a significant health risk to building occupants.

This paper is focused on wind effects with regard to compartment fires and even the isolated example shown in Figure 1 involves this type of interaction. The fresh air will enter through the door as the exhaust gases exit the compartment. Usually, the exhaust gases will exit near the ceiling while the fresh air will enter toward the bottom of the door. The hot layer of smoke and gas will radiate heat to the combustible materials within the lower section of the compartment. There will be a constant supply of new oxygen provided by the air, sometimes considered wind, entering through the door. This will allow the fire to develop with a continual rise in temperature of the compartment. Many times the hot layer will reach a temperature of 600°C or higher, which is sufficient for spontaneously igniting the remainder of combustibles within the compartment. This is often referred to as a flash over fire. Fires which do not produce this level of heat will frequently decay after the initial object has burned sufficiently. This is why firemen entering a compartment with furniture will sometimes find that only one piece of furniture has burned, while other areas may be completely decimated. The critical factor in these cases is the temperature reaching 600°C or higher.

The severity of the compartment fire is often measured by the amount of toxic gases and heat released when it burns. The typical profiles of non-flash over versus flashover fire release rates are displayed in Figure 2 below:

A Flashover versus non-flash over fire has different characteristics. The flash over fire reaches 600°C and will continue burning until most of the combustibles within the compartment have been consumed by flames. The non-flash over fire stays under the threshold temperature of 600°C and reachs a more moderate level of heat. These fires will generally burn the initial object of ignition and cease when it has burned completely. There are also smoldering fires, which do not produce much heat, but release a significant amount of toxic gases and smoke. These fires can often present a significant hazard to occupants of the building due to the toxic nature of many flammable materials used in modern furniture and other items within the compartment.

2.3 Chapter Summary

This chapter was an introduction to compartment fires. The smoldering, non-flash over, and flash over fires have different characteristics. Whether a fire remains as smoldering or non-flash over is frequently determined by the airflow, or when involved. Therefore, some of the findings of the research and investigations in the remainder of this paper are partially due to the smoldering and non-flash over fires developing into flash over fires due to the presence of wind.

Chapter Three: Wind effects in Real Building Fires

3.1 Chapter Introduction

This section of the paper focuses on the wind effects of fire in a structure which were observed in an actual seven-story building. The study was done by the Fire Department of New York City, the Department of Homeland Security, the United States Fire administration, the Federal Emergency Management Agency, and the Polytechnic Institute of New York University. The building was on Governors Island, New York. An in depth 590 page report was prepared which describes the results.

There were 14 experiments done in a seven-story building in which fires were begun in a furnished room. In 12 of the 14 experiments mechanical or natural wind was allowed to affect the fire. The island is naturally prone to consistent winds and this is one reason the area was chosen for the experiment. There were several variables investigated, which included wind control devices (WCD), positive pressure ventilation fans (PPV), and exterior water application through the floor below nozzles (FBN). These types of nozzles are also known as high-rise nozzles and are a common type of sprinkler system used in commercial buildings to suppress fires.

All the experiments were done to investigate the effect of mechanical or natural wind on the fire within a structure. There were different pressures, wind velocities, and temperatures used in the rooms. There were both video and thermal imaging used to record the experiments. This made it possible to witness fire phenomena, which would not be visible to those outside the structure. The experiments involved a post-flash over apartment fire which was wind driven. Both the stairwells and public core doors were also exposed to fires. Each apartment had a door which led onto the core door that was left open during the experiments. Conditions in the stairwells and core door were deemed important; since they are the areas used by individuals in the building that need to escape a fire.

Fires which are within high-rise buildings present unique difficulties for both firefighters and occupants. The heat and smoke which spread through the corridors and stairwells of high-rise buildings can reduce the evacuation of building occupants. During 2002 there were more than 7000 fires reported in high-rise structures, which had at least seven stories. Most of these fires took place in residentially occupied buildings such as apartments. When the fires occurred in apartments more than 90% of the fatalities among civilians occurred when the fire spread beyond the room in which it began.

The 14 fires were ignited within apartment rooms, which were furnished. There was an excess fuel pyrolysis such that the room of origin could not achieve flash over transition until Windows were self-vented. This allowed the necessary oxygen for the fire to burn in transition to flash over. When there was no wind applied to the window, the fire did not spread beyond the original room. However, even when no extra wind was applied, a flow path to the outside created by a vented window would allow the apartment fire to spread into the core door and stairwells. When a bulkhead was opened on the roof the velocities of wind and fire temperature within the stairwell and core doors was increased. When a wind of 9 m/s was applied through the floor of the fire allowing it to exit through the roof bulkhead door, the fire temperature exceeded 400°C. This would create a deadly situation for residents or emergency personnel. Wind velocities of the superheated air exceeded 10 m/s through the stairwells and core doors. These extreme thermal conditions were caused by rooms which contained common household objects. The application of wind within a structure has the potential to turn any fire into a deadly blaze.

3.2 Wind Conditions

The wind driven condition involves flames and hot gases flowing horizontally from the room of origin. This can act in a fashion, which is similar to a blowtorch and ignite objects in its path. The majority of wind driven fires within a structure will consist of gases and fire, which are in a relatively homogeneous mixture between the ceiling and floor. Temperatures will generally equal or exceed 400°C.

The experiment using the seven story building achieved a wind driven condition in 11 of the 14 experiments. In 2 of the experiments, where this did not occur there had been no additional wind applied. The single condition which did not achieve a wind driven fire condition but had additional wind applied, used a WCD that did not allow ventilation when the closed window achieved failure due to the heat. This condition results in their not being sufficient airflow for the full wind driven fire condition to be achieved.

Typical conditions of a wind driven fire were achieved in an apartment which was labeled as "7K." The fire was initiated in a bedroom which was 15 m from the measurement location. The air temperature increased gradually and was present in stratified bands with hotter areas near the ceiling and cooler areas near the floor. The bulkhead door on the roof was opened after the fire had started and provided a flow path for wind. When this was done the temperature increased to 600°C within only a few seconds. This superheated air and gas were moving at 5 m/s. This would be lethal to any unprotected individuals and is likely to be deadly for those in full firefighting gear. Due to the mixture of the superheated gases throughout the area, there is no safe area near the floor. This emphasizes the importance of training firefighters and residents of buildings that they must not enter the downwind flow path of a fire which is wind driven.

3.3 Door Control

The control of doors is the most basic means for interrupting the flow path of air in a building which is been subjected to a wind driven fire.

Figure 3-Temperature versus time for a wind driven fire in a structure

Figure 3 displays the time history and temperatures for an experiment labeled as 5 E. Times zero is when the window failure occurs. The failure of the window creates a flow path between the apartment and the stairwell. This path ends with the open bulkhead door and allows a free flow of air, which is necessary for a wind driven fire within a structure. Around the time of 70 seconds most of the window in the living room had vented, and the room was in a post-flash over condition. Flames were present from the floor to ceiling. In this condition, the temperature in the living room at all levels exceeded 400°C. When the bulkhead door was closed, the condition changed. The fire was no longer being wind driven in the structure. This was due to an incomplete airflow path. However, the living room was still receiving air from Windows, which were open in the apartment. The room continued to burn even when the bulkhead door was closed and the hot gas flow through the stairwell had slowed. Prior to the bulkhead door being shut the speed of hot gas flowing through the bulkhead was 10 m/s. Temperatures were in excess of 1000°C. A few seconds after closing the bulkhead door the velocity of the wind had decreased to 0 m/s and the temperature was approximately 200°C.

The thermal conditions on the floor changed more slowly. This was due to a continual burn within the living room. PPV fans were aimed at lobby doors with the bulkhead door shut. This meant that the wind was still blowing into the apartment. Pressures within the stairwell near the fire door were between 50 and 60 Pa.

The results of this experiment indicate that closing a door will interrupt the flow path of a wind driven fire. This is true even when the door is relatively remote from the fire. This was demonstrated by closing the bulkhead door which was not near the fire being tested. Closing doors to limit airflow will eliminate the wind driven flow conditions if they are the primary source of air transport. It should be remembered that the further the door is from the fire, the larger the area which can still be filled with stored energy and hot gases. The experiment in the seven-story building required water to be applied in order to control the fire completely and dissipate the heat.

Figure 4-temperature versus time with bulkhead door open

Opening the bulkhead door results in a flow path and increased wind driven conditions. Time zero in this figure indicates the moment that the door to the stairwell on the floor with the fire was opened. The opening of this door allowed for a flow path from the living room window to the stairwell and on through the first floor stair door to the outside environment. This resulted in increased temperature in the stairwell of the fire floor. The floors above and below the fire were also increased temperature by around 100°C. The hot gas which flowed from the apartment into the stairwell did so at approximately 2 m/s. This resulted in the stair door opening on the fire floor. This allowed air to flow out of the stairwell at approximately 1 m/s.

3.4 PPV Fans and Smoke Movement

A number of studies have indicated that PPV fans can help control smoke movement. These types of fans can be used in high-rise buildings in order to provide pressure within the stairwells. In the case of a fire, this protects them from combustion products flowing into the area due to pressures created by the fire. This was found to be the case in the experiment done on the seven-story building. A 27 inch PPV fan was used in stairwells subjected to a non-wind driven fire, which was on the 7th floor of the building. The fan was placed on the outside of the structure and forced air into the stairwell. This resulted in a temperature decrease within one minute that achieved near ambient levels.

There have been several researchers which have suggested that even two 27 inch PPV fans will not be effective in overcoming the pressure generated from a wind driven fire. However, the experiment on the seven story building showed that even a single 27 inch PPV fan placed on the first floor and 5th floor of the building could be effective. This was done with the bulkhead door shut and the fire door open from the floor level. These stairwell pressures were reduced to between 15 and 30 Pa. It should be noted that this is significantly lower than other experiments in the same building which did not use PPV fans. In these experiments, the fire door was left open in the bulkhead door closed. This resulted in a stairwell pressure of more than 40 Pa.

The results of the seven-story building experiment using PPV fans suggests that they will mitigate the wind driven fire when proper door control and airflow tactics are used. In this case, the PPV can successfully clear the stairwell and properly pressurized it in order to provide a safe means of escape for residents and firefighters (Kerber & Walton, 2005).

3.5 WCDs and Wind Driven Conditions

If the flow path is interrupted a wind driven fire will be reduced. The WCD devices which are used to reduce wind driven fires inside structures are an effective method for interrupting the flow path. If they are designed well and can withstand fire conditions they will serve as a method to block the necessary air flow for a wind driven fire. The WCDs are used to cover windows completely and interrupt the flow of air when a window fails due to excessive heat. The task then becomes one of dealing with the residual stored energy and fire. Many times there are hot gases. The floor, ceiling, and walls of a structure may also be superheated. The furnishings which remained, smoke, and fuels present in the structure can radiate heat even after a fire has been extinguished. This is when a WCD is deployed on a post flash over wind driven fire the temperatures will decrease first at locations, which are remote to the origin of the fire. Within only a few minutes there are often significant reductions in temperature near the fires point of origin.

Typical examples of wind control devices are wind control blankets, and wind controlled curtains. The WCD blanket generally measures approximately 3 x 3.7 m and is made of a silica fabric which is aluminized using a foil material. They are frequently quite heavy and can weigh more than 30 pounds. Most of these devices have a strap on each of the corners in order to secure the blanket in a desired location. They usually require between two and four people to install properly.

The WCD curtain will generally measure 1.8 m x 2.4 m. Most of these devices are made of fiberglass and weigh less than the blanket at approximately 26 pounds. These curtains are usually reinforced with metal rods. Ropes are used on each corner in order to secure the device properly. These are often easier to secure, they have the blanket and can be handled by a single individual.

3.6 FBN Control

Another room was provided with an adjustable fog tip. This nozzle was set to approximately 30° and also had a flow rate of 160 gallons per minute. This arrangement also resulted in a rapid suppression of the fire within the room of origin. However, the fog stream did not reduce the temperature downstream with the same rate of efficiency as the flowing water. The same procedure was used with the FBN being shut off in the room of origin when the fire had been suppressed in another being activated in a room in which it spread. Both fires were reduced by this method.

The experiment (Kerber & Madrzykowski, 2009) with the wind driven fire in the seven-story building showed that FBN systems are an effective means of suppressing fires. Systems which use free flowing water rather than Faulk will be more effective in reducing the temperature created by a wind driven fire.

3.7 Window Failures

The experiment on the seven story building involved apartments with double pane windows, which had aluminum frames. The focus of the study was not on window construction or types of failure. However, the data regarding this factor was recorded, and it is important when understanding wind driven fires in a structure. Once a window completely fails it opens up another area in which wind can flow and increase the fire. Failure times for the bedroom double windows in the building were recorded since they were near the point of fire ignition.

In one of the experiments, 7G, the flash over and wind driven fire did not develop until the window had completely failed. In another experiment, 7E, the pane of glass was removed. The failure time in this experiment was 163 seconds. There was a delay for the onset of flash over due to the early window failure time. In both instances, the gas temperature within the room at 2.1 m above the floor was more than 500°C.

After the experiments in 7G and 7E, the inner windowpane was left intact. However, the glass was scored using a glass cutter. This led to a higher number of stress points and an earlier failure rate. The failure times in these experiments ranged from 233 to 496 seconds. The average was 342 seconds. The gas temperatures near the ceiling when failure occurred ranged from 300°C to 700°C. In all the testing situations, it was found that the flames had made contact with the window before the failure of the glass occurred. The glass failure was an event which led to a wind driven fire condition due to increased airflow. The increased airflow serves to bring fresh air to the situation and provide more oxygen to the fire. This efficient flow path was especially effective at increasing the fire when other windows or doors were opened.

3.8 Chapter Summary

This chapter provided a case study of wind effects in an actual seven-story building. The study was conducted by several government agencies in the United States. There were 14 experiments done, which investigated the effect of natural or mechanical wind on fires within a structure. A variety of pressures, wind velocities, and temperatures were examined. There were both thermal imaging and video recordings done of the experiments. These experiments were done to provide an actual example of what occurs when a high-rise building becomes engulfed in flames. The study found that smoking heat spread quickly through the stairwells and core doors of high-rise buildings. This serves to reduce the efficiency of the evacuation and increases the danger to occupants as well as fire fighters.

The study found that wind had a significant effect on fires in buildings. The wind-driven fire causes hot gases and flames to flow horizontally from the origin of the fire. This acts as a blowtorch igniting anything in its path. This was previously discussed with regard to the critical temperature of 600°C. Any time the hot gases are at this temperature, they will cause ignition of any flammable objects which they reach. It was found that an interruption in the flow of wind will reduce the fire. This can be done by using wind control devices, which block the airflow. A number of different devices have been developed for this purpose such as fire blankets. The general concept is that covering a fire and preventing wind or air from reaching it will cause it to exhaust the source of oxygen. This will lead to an extinguishing the fire.

An important part of the case study involves observing the effect of wind of failures. The apartment used for the experiment consisted of individual units with double pane windows and aluminum frames. While the study did not focus on window failure, it was found that failure of these structures often had important consequences for the spread of the fire. Once the Windows had failed, they allowed wind to increase its flow through the compartment. This increased the heat and spread of the fire.

Chapter Four: Experimental Research Studies

4.1 Chapter Introduction

This chapter reviews a number of studies involving the effect of wind, or ventilation, on fires within a compartment. The compartments vary and may consist of tunnels or apartments. The experiments involve using different types of ventilation systems as well as natural or man-made airflow. The summary of this chapter will include a comparison of the findings to observations made in the case presented in chapter 3.

4.2 The Studies

A study was recently done to investigate fire growth in a compartment under different wind conditions. This was done using fire tunnels which formed reduce scale compartments. An external when velocity of between 1.5 and 3.0 m/s was applied to the compartment. This was done to simulate the effect of ventilation passing through a compartment. There was a measurement of temperatures done on the wall surfaces, air, and flames which were passing through the opening in the compartment. There was also measure taken of heat flux and fuel mass loss rate. It was found that wind the conditions provided a more rapid rise in temperatures as well as burnout time for the fire. The external wind increased combustion and temperatures. However, they also hastened the fires extinction. It was also noted that the flame plume was inclined in the direction of the wind.

An investigation was done of measures taken by fire and rescue services following fire fighting operations. These operations were conducted in a 3 room apartment during 15 tests. A 0.5 m diameter heptane pool was used as the fire source. The pressure and temperature in the apartment, as well as the fire flow passing through openings and the weight of the fire source were continuously measured. The test revealed that the burning rate increased with positive pressure ventilation. This type of ventilation also increased the temperature present in the rooms on the downwind side of the fire and reduced pressures on the upwind side. The study concluded that working conditions were made safer for emergency personnel using positive pressure ventilation on compartment fires. However, this tactic could increase the danger for individuals who might be trapped on the downwind side.

A series of fire tests were done in a house to determine the effect of positive pressure ventilation on compartments which were located downwind of the fire. 40 thermocouples were used to measure temperatures in the hallway and rooms. There were also recordings made with infrared and video cameras. There were 11 tests which were reported. Each of these studies had different fan configurations. There were also 2 venting strategies used. One condition consisted of events being placed in the room with the fire and the other with events in the room which was downwind of the fire. The rooms with fan applied air and room venting had higher temperatures. There were also higher temperatures in the rooms which were downwind of the fire.

There has recently been new research evidence combined with information from 3 previous research projects to investigate the effect of longitudinal ventilation on fire spread and size in tunnels The study was primarily concerned with fires within a tunnel such as those used for vehicle traffic. The researchers concluded that increasing the velocity of ventilation tended to enhance the heat release rate of a fire which was occurring in a tunnel. It was assumed that the flames would be coming from a vehicle as a fire source. Despite the increasing heat release rate, the ventilation reduced the chance of fire being spread to nearby vehicles as long as there was no flame impingement present. This might be a unique situation regarding wind and fire since these types of tunnels are made of non-flammable material such as steel and concrete.

The external venting of full-scale flashover fires was studied in two ventilation cases. The study was done to determine the effect that burn room ventilation had on the environmental conditions affecting venting plumes. This was done through examining full-scale flashover fires within and experimental building fire facility at the Victoria University of Technology. The first class of ventilation studied was flow through while the second was a no draft condition. The conditions were used to represent open doors or windows in a burn room as well as rooms in which the door was closed, but a window was open resulting in less airflow. The temperature contours outside of the fire room were measured in 3 dimensions in order to understand the fire plume dispersion. The severity and overall reach of the venting flames were combined with the heat flux measurements in order to understand the effect of flames emerging from a burning room. These flames are often the cause of secondary fires in upper levels. The study found that any swirling motion present within the venting plume was caused by a crosswind being present. This resulted in a higher likelihood of the flames spreading to other compartments. Thus, wind acted as a factor in spreading the flames from one compartment to another.

A study was recently done to investigate the effect of point extraction ventilation on model scale tunnel fires. A series of tests were done using a model scale tunnel. There were both 2 point and single extraction ventilation systems used. The point extraction system was investigated under varying fire loads and situations of natural or forced longitudinal ventilation. The fire source consisted of wood crib piles. This source was used to represent a vehicle carrying heavy goods. The studies examined fires under the conditions of different numbers of wood cribs as well as varying ventilation velocities. There were also different types of ventilation and exhaust used. The measurements taken indicated heat fluxes, maximum temperatures beneath the ceiling, fire growth rates, and heat release rates. The study concluded that the flow of smoke and fire is controlled if the single point ventilation velocity upstream is above 2.9 m/s. It is controlled downstream when it is above 3.8 m/s. In a two point extraction ventilation 2.9 m/s rate will control the fire and smoke flow.

A model for describing the behavior of gases in a tunnel fire has been developed. This model was developed with the intention of predicting the behavior of fire gases which are spreading due to the effect of fresh air current provided by tunnel ventilation. An analytic formula was developed that estimates the critical ventilation velocity. Once this velocity is reached it prevents back layering of the fire. In this situation the gases from the fire remained close to the source. It was found that the critical velocity of the ventilation depended on the heat release rate as well as the height of the tunnel. The formula created by this study provides a reasonably good estimate of the critical ventilation velocity applicable for tunnel fires.

The concepts of critical velocity and back layering in tunnel fires were investigated in relation to longitudinal ventilation. Experiments were done in 2 tunnels which were longitudinal events related. The results of the experimental conditions found that the critical velocity complied with expected results in both tunnels. The length of back layering was associated with the ratio of critical velocity to longitudinal ventilation velocity. The data collected in the experiment indicated that there was a ratio of the critical velocity to the ventilation velocity and dimensionless back layering length is represented by an exponential relationship. A correlation is proposed which is meant to predict the length of back layering. The experiment indicated that a decrease in critical velocity is obtained when an obstruction is present which is higher than the ratio of the cross-sectional area of a vehicle in relation to the cross-sectional area of the tunnel.

CHAPTER SUMMARY

The experiments supported the findings of the case presented in chapter 3. One study investigated the fire growth in a compartment under differing wind conditions confirmed the case study presented in chapter 3. The study found that when additional wind was applied within a compartment, there was an increase in both temperatures and combustion. The finding that fire extinction could be hastened by additional air was not presented in the case study. The findings by the study involving measures taken by rescue workers also supported the case study from chapter 3. These investigators found that the ventilation increased fire temperature in rooms, which were downwind of the fire and served to reduce pressures on the upwind side. The conclusion was that strategically placed PPV systems could enhance fire safety of those fleeing the building and rescue personnel. In fact, these findings are identical to those from the case study in chapter 3. Another experiment which supports the findings of the case study in chapter 3 is the series of fire tests done within the house in order to determine the effect positive pressure ventilation would have on compartments located downwind of a fire. This study found that vented rooms with a fan that applied air had higher temperatures on the downwind side. No mention was made of the fire extinction time in this study.

There were several additional experiments presented in this chapter, which supported the case study in chapter 3. For example, the study involving the venting of full-scale flash over fires also supported the case study. This study found that a swirling motion present in venting plumes was caused by crosswinds. This resulted in spreading the flames from the compartment of origin to other secondary areas. The increased fire spread secondary to wind was noted in the case study as well. The study involving point extraction ventilation on model scale tunnel fires also supported the case study. It was found that the flow of smoke and fire could be controlled by either single point or 2 point ventilation systems as long as a sufficient ventilation velocity was used. The case study suggested that ventilation systems be used to control smoke and fire in order to allow for safer evacuation of flaming buildings.

Chapter Five: Numerical Investigations

5.1 Chapter Introduction

The majority of studies in this chapter will involve the use of computational fluid dynamics (CFD). This is a field of fluid mechanics, which uses algorithms and other numerical methods in order to solve and analyze the flow of fluids. In the 21st century, the application of this science uses computers in order to accomplish the complex mathematical calculations which are necessary. This allows the modeling and simulation of liquids and gases with regard to their interaction and boundary surfaces under a variety of conditions. Until very recently, the majority of these calculations were done with high-speed supercomputers. However, the recent advances in personal computers have allowed many of these calculations to be done with more common computing devices. There has recently been a number of software programs developed for personal computers, which take advantage of more efficient methods of computation. This software allows smaller computers to have the accuracy and speed necessary to do these complex computations of fluid dynamics. A number of scenarios involve the analysis of transonic or turbulent flows. Software which is used for this purpose is often validated using a wind tunnel. This allows for observation of physical phenomena in order to ensure that the modeling calculations are correct.

Underlying nearly all the computational fluid dynamics models are the Navier-Stokes equations. These formulas are used in order to establish the nature of a single phase fluid flow. The calculations are often simplified by describing viscosity, which yields the Euler equations. A further method of simplification is to describe vorticity, which allows for full potential equations. The resulting equations are linearized and known as the linearized potential equations.

5.2 Progress in Computational Fluid Dynamics

During the 1700s, there was significant progress made toward describing the motion of fluids. There were several pioneers of the science which are responsible for this progress. Some of these individuals included Bernoulli, Reynolds, Euler, and Poisson. Euler developed equations, which described the conservation of momentum and mass, which occurred in an inviscid fluid.

There was additional progress, which was made in the 1800s by George Stokes and Claude Navier. These 2 individuals used the Euler equations in order to explain the phenomenon of viscous transport. They were then able to develop equations, which are now commonly known as the Navier Stokes equations. These are differential equations, which can be used to describe momentum, pressure, conservation of mass, and turbulence. They are one of the cornerstones of modern computational fluid dynamics. During the early part of the 20th century, the theories were refined with regard to the boundary layers in the turbulence present in fluids.

The Navier Stokes equations interrelate with each other and are complex. Early attempts to solve the equations were unsuccessful (Cheung et al.). It was not until the age of modern computers in the 1960s and 1970s that some of these equations could be solved. Computer analysis allowed for the resolution of these equations with regard to the actual flow of fluids.

The software which is used for computational fluid dynamics became available commercially during the 1980s. This software allowed computational fluid dynamic modeling to be applied in a wide range of science and engineering applications. It made it possible to understand the flow of liquids and gases. This type of modeling is now used in a broad range of fields, including aerospace, automotive, environmental design, the Marine industry, chemical process engineering, and the medical sciences.

One important use of the computational fluid dynamics software is in the field of fire safety science. One of the early practical uses of computational fluid dynamics for fire safety involved modeling, which was done in 1988. This was part of an investigation into the King's Cross Station fire which occurred in the United Kingdom. The National Bureau of Standards in the United States developed software, which made use of computational fluid dynamic models in order to assist with the development of models that could be used for fire safety applications. There were several models developed and these were eventually consolidated into what is known as the Fire Dynamics Simulator (FDS) which is now widely used by fire safety engineers.

5.3 Necessary Equations

Computational fluid dynamics takes advantage of specific equations to describe processes, which involve heat transfer in the flow of fluids within a fire. The equations which govern these processes are basic to the physical sciences. These are the laws of conservation, which indicate how there is a consistency in nature over periods of time. Differential equations are used in order to represent the laws of conservation.

The dependent variable for each of the differential equations used is a physical quantity. These are generally expressed in terms of a specific mass basis such as velocity or heat. While temperature is not a specific property, it is often used as an independent variable and arises from equations, which are more basic and related to internal energy. An arbitrary domain such as that presented in figure 1 is helpful in understanding the nature of these equations:

Figure 1-Grid Representation of a Computational Domain

5.3.1 Conservation of Mass

The first conservation equation to be considered is that of mass. This equation mathematically represents the physical wall that matter is neither destroyed nor created and the total mass within a system which is isolated does not change regardless of the chemical or physical state of substances, which are within the system. While it could be argued that according to Einstein's theory of relativity this is not entirely true, for the type of analysis being done with regard to fire safety these effects are so slight that they can be ignored. Figure 2 below shows mass which is traveling in the X direction. The mass increase in control volume will be equal to the inflow rate across its faces:

Figure 2-Conservation of Mass

The mathematical formula for representing the conservation of mass is as follows:

5.3.2 Conservation of Momentum

Another important physical law for computational fluid dynamics is the conservation of momentum. The momentum of an object is equal to its mass multiplied by the velocity. This is a vector quantity. Newton's 2nd law of motion can be applied in this situation it indicates that the sum of all forces which are acting upon an object will be equivalent to the time rate of change. This is represented by the following equation:

In a system which is isolated the motion does not change, and the total momentum will be constant with regard to time (Carlsson, 1999). And equation can be derived for this situation, which represents momentum in the X direction. This equation is represented below:

The equations which represent the momentum in the Y and Z directions have a similar structure and can be represented using Cartesian tensors (Carlsson). The equation which represents both the Y and Z axes with regard to momentum is as follows:

A B C D E

A represents the local change in momentum for the control volume.
B accounts for change in momentum due to the motion of an unsteady flow field.
C is the pressure which is acting on the fluid.
D stands for the viscous shearing forces.
E represents the forces on the body such as gravity.

5.3.3 Conservation of Energy

The law which described the conservation of energy is similar to the first law of thermodynamics. Specifically, energy can neither be destroyed nor created. It merely changes forms. For example, energy can be transferred from chemical energy to kinetic. Within a reacting system which has a number of different components several mechanisms can contribute to heat flux. Some of these factors are radiation, convection, and conduction. The energy equation which is used to represent these factors depends upon the dependent variable used. When the dependent variable is the total enthalpy the equation is represented as:

A B C D

A indicates the rate with which the storage of enthalpy within the fluid changes.
B is determined by the influx of enthalpy which can be attributed to convection.
C represents work done upon the fluid resulting from pressure.
D Is the net flux of heat which is a result of thermal radiation and conduction.

5.3.4 Conservation of Chemical Species

When modeling the behavior of a fire combustion must be taken into account in the form of the equations representing the conservation of chemical species. This is actually in addition to the mass continuity equation. The idea is that species which are in a system are conserved. A chemical reaction in this type of modeling can be represented by the following equation:

Indicates the stoichiometric coefficient.
Is the product stoichiometric coefficient.
Represents the chemical symbol for the species.
N is the total number of species which are present within the reaction.

Constraints are placed upon this reaction when the value of has been established. The equation can then be written in differential form as follows:

A B C D

A represents the rate of change which occurs in.
B is equivalent to the influx of species resulting from convection.
C describes the net rate of change as a result of molecular diffusion.
D is the rate that different species change due to varying sources of the control volume.

5.4 Combustion and Turbulence

The development of an accurate computational fluid dynamic model of fire requires the consideration of combustion, thermal radiation, and turbulence. A common way of accounting for these factors is to use discrete transfer. This is often included within a flux model. The proper analysis of combustion is frequently the most difficult part of the analysis related to an accurate computational fluid dynamic modeling with regard to fires. One of the more simplistic solutions is to avoid modeling combustion by representing it as a volumetric source for the fire. A more accurate method is to develop a model which consists of a one-step reaction. In this type of analysis the fuel and oxygen react in an instant way to create the combustion products. This is the way in which the mixture fraction model is developed. This has become a popular method for applying computational fluid dynamic modeling to the analysis of fires. A more sophisticated approach of analyzing combustion is the Eddie breakup model. This type of modeling accounts for the effect that turbulence has on the chemical reaction.

Another important part of an accurate computational fluid dynamic model of fire is turbulence. One method of accomplishing this analysis is to split it into the general groups of Reynolds Averaged Navier Stokes models (RANS) and the Large Eddie Simulation (LES). The more popular of these two methods is the RANS approach. This type of modeling uses Navier Stokes equations, which have been averaged for all the time scales, which are being considered. In the LES approach, the flow of turbulence is averaged at scales that are smaller than the mesh size (Cheung et al., 2004). This means there is a direct solving of the turbulence equations when using the LES approach. This means that the LES approach is more computationally demanding than the RANS approach. This is because a finer mesh must be used in the LES computation. The coarser mesh of the RANS approach allows for more rapid computations to be done).

5.5 Fire Safety Engineering and the use of CFD Modeling

The continuing improvement of computational fluid dynamic modeling has allowed the approach to become more widely used in safety engineering, which is related to fires. The advances in both hardware and software for personal computers have helped with this transition. In the 21st century, it is possible to do complex computational fluid dynamics simulations with a personal computer. This means that using computational fluid dynamics modeling is now more cost effective due to the removal of the necessity in using supercomputers. There are a number of general purpose codes for computational fluid dynamics, which have been developed. These could include Fluent, Phoenix, as well as CFX. There are also codes, which have been designed especially for application to fire engineering. The software codes include Jasmine, Smartfire, Sofie, KOBRA-3-D, and the popular FDS. Some fire engineers prefer general codes because they can be used in a more versatile fashion. Many of the general codes have sub models, which can be used to account for combustion, turbulence, and thermal radiation. Unfortunately, they can be difficult to use. Many times using general codes for computational fluid dynamic modeling involves the preparation of subroutines to account for problems, which are specific to fires. This can be a complex task.

The codes which have been specifically designed for computational fluid dynamic modeling of fires are frequently easier to use than the more general codes. The drawback to these software solutions is that they are less versatile than the general codes. Adapting them to unique problems can be very difficult. Most of the computational fluid dynamic codes which are specifically designed for analyzing fires use the RANS approach to deal with turbulence. However, the FDS software is the exception to this rule and uses the LDS model.

5.6 CFD Validation

The validation of computational fluid dynamic models when used for fire safety engineering is essential in order to ensure that proper data is provided (American Society for Testing). This type of validation includes determining how accurate the models are for representing fires in the real world. This means it is important to determine the appropriateness of using certain mathematical models and any related equations. It must be determined that they are being used in an accurate method which actually describes the phenomenon which is being modeled. One way which this is accomplished is to compare real-world experimental findings with the prediction of these models.

It can be difficult to validate computational fluid dynamic models of fire due to a lack of appropriate data. There is a great deal of experimental data available regarding large-scale test fires. However, most of these tests were not done with the goal of assessing the accuracy of computational fluid dynamic fire models. The problem is particularly problematic for models of computational fluid dynamics due to their high levels of sophistication. It is often necessary that there be vast amounts of data input into these models. The data requirements can be exponentially increased over those required for two zone models of fire. Despite the additional data necessary, the computational fluid dynamic models have the advantage of providing detailed predictions regarding flow velocities, concentration of species, and the spatial variation of temperatures within the fire.

A greater number of parameters must be used in order to validate the more sophisticated types of computational fluid dynamic models, which are now being used. This means that proper validation experiments must collect extra data. In order to do accurate validation experiments for determining the accuracy of computational fluid dynamic models a number of inputs must be accurately recorded. These inputs include quantities predicted, initial conditions, and factors such as boundary conditions. An accurate geometric description regarding the area is also necessary. With regard to the materials, physical properties will need to be recorded. A particularly important aspect of these models is the thermal properties of the area and the compounds involved. Information needs to be collected regarding the heat release rate, soot yield, and fire size. The boundary conditions must also be recorded. Both temperature and flow velocities must be input into the model. Another essential type of input is the addition of active elements, which may be present in the system. This would include factors such as sprinkler systems or other automatic fire depressants.

The location in which a computational fluid dynamic model is being applied is also important and requires a number of accurate measurements. These measures include things such as visibilities, heat flux, flow velocities, and concentration of species. It also involves the accurate measurement of fire behavior, visual indications of smoke, and temperature. There can also be complex parameters, which are vital to the accurate development of a simulation. However, these factors may not be required as inputs because they are already provided by the output of the computational fluid dynamic model an example of one of these factors is burning rate.

5.7 CFD Standards

Computational fluid dynamic models have become routinely used by fire engineers to improve the safety of buildings with regard to unexpected fires. This has made the necessity of standards regarding the use of this type of modeling essential. This must be done in order to ensure that accurate models are being used, and buildings are actually as safe as predicted. This can be a life or death situation with the survival of building occupants being dependent on accurate fire modeling. There must be a reliable assessment of the computational fluid dynamic models in order to ensure predictive accuracy. It has been suggested that there should be a single system of standards for the use of fire field models involving computational fluid dynamics. In the United Kingdom, the Fire Research Division for the Office of Deputy Prime Minister has developed these types of programs and standards. There have also been efforts within the United States to develop standards for evaluating the accuracy of fire models. The United States is now using an ASTM standard to provide guidance for methods of evaluating the accuracy and appropriateness of fire models.

5.8 Fire Dynamics Simulator Modeling and Software

Computational fluid dynamics has been used to develop software, which is now known as the Fire Dynamics Simulator (FDS) which allows the modeling to be done that predicts the fluid flows which are driven by fires. The original language used to create the software for the FDS was Fortran 90. This was introduced to the public during February of 2000. There have been a number of improvements since that original release and many useful features have been added. The FDS software has proven valuable for modeling many of the problems faced by fire protection engineers. It has also been used for experiments related to firing, combustion, and computational fluid dynamics.

There are a variety of phenomena related to fire, which can be modeled with the FDS software in its present form. Accurate analyses of pyrolysis, fire growth, flame spread, secondary systems of fire suppression, and activation of heat detectors can all be accurately predicted with the FDS modeling. It is also effective for understanding low-speed transport of heat and combustion products as well as heat transfer which occurs between the solid and gas surfaces. It is now common for fire safety professionals to use the FDS software in order to develop more effective smoke control systems and fire suppression using sprinklers. The FDS can also be used for forensic analysis of fire scenes in which arson is suspected. An example of this which is relatively well-known to the public is the investigation of the tragic events which occurred on September 11, 2001 at the World Trade Center.

5.8.1 Geometries and the use of Multiple Mesh Sizes

The FDS software approximates the governing equations which determine a rectilinear grid. The operator of the FDS must approximate the shape of the area being analyzed, including its obstructions, in order to conform to an underlying grid which is based on a rectangular shape. This can be a limitation of the FDS modeling when the area has geometric features, which are not rectangular. A particular problem with regard to this modeling is the accounting for sloping angles. The FDS software does use a number of techniques in order to account for the effect of these nonrectangular objects. The majority of studies have found that the overall effect of using these adjustments for the model makes the results relatively accurate. Unfortunately, the adjustments cannot be completely relied upon to be accurate in the case of an important boundary layer which consists of a nonrectangular object. In this case, the FDS will make allowance for the problem by allowing a changing grid size. The grids can be stretched in two coordinate directions.

Another method of dealing with geometries, which are nonrectangular, is to use more than one size mesh within the calculations. The smaller mesh allows for a more accurate computation of nonrectangular shapes in the underlying grid. When the multiple mesh size approach is used the engineer using the FDS can use a finer mesh in critical areas. In most cases, this will only slightly increase the computational power which is necessary to do the appropriate modeling. This is preferred to changing the entire area to one, which uses a smaller mesh size. This would substantially increase the computational power and slow the mathematical analysis and modeling.

5.8.2 Radiative Heat Transfer and Combustion

The mixture fraction model is used by the FDS to determine combustion. The mixture fraction is a scalar quantity which is conserved in this type of analysis. It is generally defined as the amount of gas present near a given point that is within the flow field that has originated from a fuel for the fire. The assumption is made that the combustion is occurring as a mixed control process. It is also assumed that there is an instantaneous reaction occurring between the oxygen and fuel. The major reactants have mass fractions, which are determined according to the mixture fraction. This is established using state relations. These relations can be determined by using the empirical expressions which are a combination of simplified analyses and measurements.

A radiation transport equation is used by the FDS to account for the radiative heat transfer. This equation was established by using the properties of a non-scattering gray gas and its heat transport properties. In some types of limited cases a wide band model is used instead. In the majority of cases, the technique used is similar to the finite volume method which was originally used to solve the equation. This is quite similar to the application of the finite volume method for application to convective transport. This is the Genesis of the name "finite volume method." (Xin et al.).

5.8.3 FDS Accommodation of Hydrodynamics

The FDS software takes advantage of the Navier Stokes equations in order to determine thermally driven flows, which occur at low speeds. The modeling focuses on smoke and heat transport, which occurs with a fire. The software uses the same equations which are the basis of computational fluid dynamics. These consist of mathematical representations, which determine the conservation of momentum, energy, and mass. The Navier Stokes equations which account for fully turbulent systems do not have a single analytical solution. Numerical methods are used, which use a three-dimensional grid that divides components into small cubes. These cues are frequently referred to as grid cells. The FDS software uses models that calculate the physical conditions in each of the grid cells for given periods of time.

An explicit predictor-corrector scheme is used by the FDS in the form of an algorithm which accounts for the hydrodynamics in a system. This model is accurate up to the second order in space and time. The Smargorinsky form of large Eddie simulation is used to account for the turbulence in this model. Alternatively, some simulations use the direct numerical simulation approach which is also offered in the FDS software system. However, the direct numerical simulation requires the use of finer numerical grids. This substantially increases the computational power necessary in order to solve these types of equations for practical applications. For this reason, most engineers and researchers prefer to use the LES approach which requires less computational power.

5.9 Wind

5.9.1 Wind Aided Flame

Before computers were being used to do computational fluid dynamic modeling investigations were being done to determine the effect that wind had on the spread of flame (Mekki et al., 1990). A detailed investigation was done on how wind would affect the way flame spread over would as well as Poly (methyl methacrylate) which is commonly known as PMMA and marketed as Plexiglas. The experiment involved laminar forced flow wind aided flame. The researchers measured the spread of the flame and pyrolysis front as well as the production rate of chemical species over given periods of time. The goal of the study was to determine the dependence of the pyrolysis and flame front speed on the oxygen mass fraction and free stream velocity. A secondary goal was to determine the local rate of fuel pyrolysis as measured by the production rate of the primary chemical species.

The study had a number of important conclusions (Mekki et al). The flame and pyrolysis front for wood and Plexiglas was close to one another. In fact, they were closer than would be predicted by the theoretical models used at the time. This similarity was also true for the flame and pyrolysis front speeds. This was true regardless of the oxygen mass fraction or free stream velocity. The flame front speed and pyrolysis front for Plexiglas and wood varied in a linear fashion according to the free stream velocity. These researchers concluded that the Emmons' solution was a good predictor of the burning zone. The flame spread rates were dependent on the local heating by the flame tip which was in the adjacent preheat zone.

5.9.2 Wind and Slope

A recent study was done to investigate the effect that slope and wind had on the scaling of forest fire rates (Perez et al., 2010). The scaling laws used were from Froude Modeling. This was done in order to account for the variations which would be present in the wind dated fires, which traveled up slope. This scale model size was changed from that of a forest fire to the smaller scale condition of a laboratory experiment. The research indicated that the spread scaling law with regard to rate of the fire was no longer valid when using the laboratory scale. This lack of validity pertained to high-speed wind conditions with steep slopes.

These researchers concluded that when scaling laws are being used to reduce the scale of fires in order to examine them in an experimental setting, a number of important factors must be kept in consideration. The scaling laws have a limit with regard to their applicability. Neither the rate of spread nor fuel bed can increase indefinitely as is suggested by scaling laws. In other words, the way in which the fire acts in a forest will be different from the limited model used in the laboratory. There are also undesirable scaling effects, which happen in the laboratory with regard to the wind profile. The laboratory results in the yield inaccurate effects regarding the fuel bed depth within actual forest. This is especially true when wind tunnels are being used since these may produce wind velocities, which are not present in most forest fires.

5.9.3 Wind and Complex Terrain

Understanding the characteristics of wind over complex terrain can have implications for managing bush fires.

Understanding the spatial distribution of a wind field over this type of terrain provides the opportunity for more accurate prediction regarding a number of practical applications. These applications include bush fire risk management, general fire spread modeling, and pollutant dispersion modeling. Directional changes which occur in the surface winds can be particularly important within the practical application of fire management.

A study was recently done, which analyzed terrain-modified winds by considering the joint probability distributions arrived at by collecting both directional data and wind speed. This was done over a rugged terrain in the southwest and west of the Australian Capital Territory. There were two landform elements, which were of interest. These were a moderately steep valley as well is a steep slope. The distributions created showed themselves to be useful tools for characterizing and identifying the primary states within wind terrain systems.

Several flame processes were explained by this type of analysis. These processes included dynamic channeling; Lee-slope eddies, as well as thermally driven winds, which are frequently present in complex terrain flame scenarios. The experiment also revealed that there was a stochastic nature to the wind terrain systems. The finding of a stochastic nature means that some of the more deterministic approaches used in modeling of this type of terrain modified surface wind may not be completely accurate.

5.9.4 Wind Driven Transport of Firebrands

The distribution of downwind firebrands created by a wild land fire has been examined. This study was done using an analytical approach. The process which was studied involved the associated spot ignition created by wind driven transport as well as the omission of firebrands. Firebrands arising from a fire front were examined using a stochastic process. It was hypothesized that this would reflect the interaction of the burning fuel debris formed and the gas flow plume with a number of other factors. These factors included firebrand size, emission, and rate of fuel consumption. This can be conceptualized as being analogous to the random distribution present in non-burning windborne particles.

The study found that the downwind firebrands could be understood as a statistical pattern with a Rayleigh form. The mass and number of firebrands which eventually landed downwind determined the maximum travel distance. This was analyzed over the full impact, including the fire spread and the eventual burning out process. These researchers found that the model they were using was applicable to the Bush fire which occurred in Canberra, Australia during 2003. They believe this indicates that the model can be used as a method for predicting the distribution of firebrands downwind in natural wildfires. There was also an accurate determination of the effect that the ember attack of this type of fire had on homes, which were near the urban interface. The conclusion of these researchers is that modern fire modeling can be an asset to predicting the effect of wildfires at the urban interface.

5.9.5 Fire Whirls

A fire swirl resembles a small tornado within a fire. There is a vertical vorticity which creates a swirling plume of fire. While this is a relatively rare phenomenon, it is potentially catastrophic. A numerical investigation of fire whirls was done in order to understand how the swirling motion can alter the combustion and plume dynamics of a fire. This type of plume fire is buoyancy driven and can entrain ambient fluid due to the rise of heated gases. The vorticity which is associated with this type of phenomenon has many mechanisms, which include wind shear, which can serve to concentrate the fire. There can also be a vortex core which is created along the plume axis. This whirling fire can be superheated and especially destructive.

A recent study used an approach which considered the relationship which exists between buoyancy and the whirl. This type of model was based on the consideration of the circulation as well as the heat release rate of the fire. The numerical analysis of the model used large Eddie methodologies. The results indicated that the fire plume creates a structure which is significantly altered from the associated fire with regard to angular momentum. This effect is imparted to the ambient fluids. The buoyancy causes vertical acceleration, which generates fields of strain that stretch out beyond the flames and wrap themselves around the center line of the plume. This causes the fire to constrict upon itself radially and rise in the vertical direction. It is hoped that understanding the way in which these plumes form and progress will lead to later studies that help predict their formation and methods of reducing their devastating effect.

5.9.6 Heat Transfer with Wind Aided Flame

Another experiment which was done prior to the common use of computational fluid dynamic modeling with computers investigated heat transfer of wind aided flame near a ceiling. In order to investigate this phenomenon the ambient oxygen mass fractions and wind speed were recorded with regard to their effect on heat transfer at the surface underneath a flame which was aided by wind. The experiment was done as a ceiling configuration. There were high temperature ceramics used on the surface in order to record the temperatures accurately. There were also in depth thermocouples that were used downstream of a Plexiglas sample which was subjected to flame spread. There were simultaneous measurements made of the flame tip, gas phase temperatures, ceramic temperature, and surface temperature of the Plexiglas. There was only a slight amount of excess pyrolyzate produced in this experiment so the flame length was nearly equal to the length of pyrolysis. This allowed for heat transfer measurements below the flame to be conducted using a steady-state methane diffusion flame. This was present in the boundary layer on top of the ceramic detectors which were mounted in the ceiling.

The results of this experiment demonstrated that the heat transfer measurements which were downstream from the flame tip agreed with the flame spread experiments. This means there is a partially steady gas phase behavior which occurs with the wind aided flame. This implies that transient behavior in this reaction is the result of the solid phase. This allowed for accurate models to be created, which represented surface temperatures downstream of a pyrolysis front.

5.9.7 Cross Wind Effects

A finite volume procedure can be used to numerically study the effect that cross wind conditions have on the behavior of a turbulent diffusion flame. With this type of study a turbulence model of the form and ß-probability density function can be used to describe the turbulent combustion process. A gas-soot radiation model can be used to analyze the soot formation. This requires the use of a 2-equations sub model based on a PI-differential approximation method. This type of modeling is effective for studying methane/air flames, which have been subjected to cross wind velocities of between 0.5 m/s and 2 m/s. When there is a moderate wind of 1 m/s the flame associated with the developing large structures can be examined concerning its transient behavior.

Under varying wind conditions the numerical results indicated that the instabilities which occur at the thermal plume are dependent upon the orientation of the density gradient in relation to gravity. The effect that winds have on the flame trajectories can be examined by noting the angle of deflection between the wind direction and hot gas steam. This results in a numerical consideration that is somewhat correlated with a pool fire. The soot and temperature concentration profiles showed that there are aerodynamic interactions that result in significant changes in the thermal flux. This was especially true for the Leeward surfaces.

5.9.8 Wind and Large Cylinders

A complete understanding of fire phenomenology requires that objects within the fire environment be considered and their effect understood. A study was done to investigate the effect that large cylinders would have on fires that were subjected to wind. The experiment consisted of using cylindrical calorimeters, which were 3.66 m in diameter. They were engulfed in flames of an 18.9 m diameter pool fire of JP-8 fuel. This environment was subjected to a variety of wind speeds. There were measurements taken and analyzed. This was done in order to determine the heat flux distribution which was present in the presence of varying wind speeds.

The results of this study showed that the magnitude and distribution of the heat flux as well as the location of the continuous flame was influenced by the variations in the speed of the wind. When there was high wind present at a rate of 10.9 m/s there was a significant increase by an order of 2 in the incident heat flux. On the leeward side, there was a heat flux of up to 300kW/m2. This was a significant increase relative to the heat flux which is typical of hydrocarbon fires, which engulf objects. These investigators concluded that fires which occur in windy conditions with large objects present can act significantly different than fires which occur when no large objects are near. The wind can significantly increase the difference these objects have on the nature and progression of the fire.

5.9.9 Wind and Atria Fires

The increasing popularity of the computational fluid dynamic-based modeling programs such as FDS require that these models be tested as to their applicability in a variety of different circumstances. With regard to fire safety engineering of buildings this includes fires, which occur in atria. A study was recently done, which investigated calculations regarding unsteady atria fires. A finite volume CFD program was used, which considered both 3 and two-dimensional buildings, which had been immersed in atmospheric boundary layers. These models showed that external wins are able to modify the exfiltration and infiltration of air through a number of apertures into atria. These apertures include windows and external doors as well as any other structures, which allow airflow. This can result in a distortion of the smoke and thermal columns which rise above a fire within an atria. This can result in fire plumes, which have an impact on the walls of the atria. It can also modify the way in which the fire interacts with the atria space.

The researcher for this study points out that it may be limited because a number of two-dimensional impressions were projected onto three-dimensional spaces. It has been previously shown that the flow field in a two-dimensional space can be significantly different from that in a three-dimensional configuration. The calculations done for the experiment were preliminary but provided significant clues into the possible effect of external wins on a fire within an atria of the building.

5.9.10 Wind Velocity and Slope

It has been known since the 1990s that slope and wind velocity can affect flame properties. A study was done, which investigated the effect that percent slope and when velocity would have on the link the flames in an open topped wind tunnel which could be tilted. Vertical birch sticks were used in the fuel bed as well as aspen excelsior. The average flame lengths were between 0.08 and 1.69 m. The majority of the backing fires had a flame length which was no longer than 0.25 m. The angle of the flames ranged from -46° to 50°. The flame length and angle were compared with predictions from a number of different models, which could be applied to fires present on a horizontal surface.

The results of this study showed that equations based on the Froude number tended to underestimate the flame angle for most of the slope and wind conditions. However, it was found that the data collected supported the theory that the flame angle can be predicted by using the square root of the Froude number. The differences between the predictions and data collected were hypothesized to result from slope effects and measurement difficulties. An equation which made use of Byram's convection number predicted nearly 50% of the observed variation within the flame angle. The original equation used by Byram related flame lengths to fire line intensity and tended to overestimate the length of the flame. These researchers recommended that fires be observed under a wider range of conditions and at the full field scale rather than the reduced laboratory conditions.

5.9.11 Non Parallel Wind and Slope

An experiment was recently done to investigate the physical modeling of surface fires, which occurred under conditions of nonparallel wins and of varying slopes. A previously developed physical model was used in order to predict fire behavior. This allowed for computation times, which were more rapid than real-time. However, this approach introduced in empirical law, which determines flame height according to the heat release rate. This was not believed to be completely accurate. The investigators sought to improve these predictions by incorporating a triangular flame hypothesis. They also included a modification to the formula which accounted for the air cooling which can occur at the rear fire front. In order to test the variation of this model, they compared it to experimental results of fire. The experiments were done in homogeneous fuel beds made of dead pinaster needles, which were subjected to high wind and slope conditions.

The new model which was created by using these modifications was analytical and did not require a mesh. The simulations were carried out using a single set of parameters that allow for an emphasis of the robust nature of the proposed model. This resulted in good predictions of the fire parameters. This was especially true with regard to rate of the spread and direction of the fire. This was true for both the rear fire sections and the fire front head. This was true even in conditions of high wind and slope. The authors conclude that the changes in the model made it more accurate.

5.9.12 Wind Advection

The advection of wind can have an influence on how I fire spreads across the fuel bed. An experiment was done, which did modeling using a semi-physical approach and traditional experiments to investigate this influence. This is seen as being an advance toward elaborating a simplified model of the way fire spread which could be used in a simulator. In order to accomplish this goal a wind velocity profile of the burning zone was presented, which included in an effective term as well as a thermal balance. The general procedure used consisted of a multiphase approach. The semi-physical model was elaborated on by using momentum equations that represented the multiphase model in the specific wind profile. The predictions made by the model were compared to observations of the experimental data, which were recorded. This was done with fire, which was spread through the litter of pine needles. Observations and predictions were made at varying wind velocities and slope values.

The results of this experiment showed that gas velocity distribution was an important part of the model in order to accurately determine fire spread rates. The addition of a wind profile allowed for additional improvements of the predictive model and the use of a single value in relation to the advection parameter.

5.9.13 Wind and Vegetation Fire Measuring Devices

Despite the recent advances in computational fluid dynamics and the modeling of fires, there is still a lack of reliable data regarding the thermodynamic specifics of forest fires. A study was recently done which proposed a method for establishing measuring devices, which could improve the knowledge base regarding the fundamental physical properties involved in the propagation and control of wild land fires. A number of devices were developed, which would determine the rate of spread, fire front shape, energy which spread ahead of the fire, as well as the vertical distribution of temperature. These devices were also able to determine the temperatures which occur inside the fire plume as well as the velocity and direction of wind. These measurement devices were applied to a fire which was spreading across the test site through Corsican shrub. Data was recorded, which allowed reconstruction of the fire behavior and determination of its properties.

The results of the experiment demonstrated that simple devices could be used, which determined the dynamics involved in a wildfire which was spreading through vegetation and driven by wind. The fire plume pattern could also be determined and described by these devices. This type of research is necessary in order to develop instruments, which can be used to predict and understand the behavior of wildfires. The results were seen as adding to the understanding of Byram's fire line intensity and the thermodynamics of a fire in general.

5.9.14 Numerical Study of Wind Effects

Many vegetation fires involve wind as a factor of their properties. The study used a variation of a three-dimensional two-phase model in order to describe how wind can affect the properties of fire spreading through vegetation. A numerical modeling procedure was used. This allowed for the spatial distribution and pyrolysis rate of the burning vegetation to be represented. It was found that it was closely related to the heat release rate associated with a single tree or groups of trees. The wind speed in this experiment was varied between 0.5 and 8 m/s-1.

The results of the study indicated that approximately a third of the heat which is produced through combustion is released in the fuel bed. This occurs relatively independently of the vegetation distribution and wind speed. The results showed that the tilt angle, flame length, and flame height were consistent with data in already available literature. The model indicated that the flame height and length would be normalized by its depth. It further predicted that the flame would be normalized by a modified Froude number represented by as well as a heat release rate which was dimensionless. The velocity flow streamlines, which were predicted by this model, were relatively straight and present in both the inert plume and flaming region. This supports the importance of far field analysis in determining the tilt angle of the flame. The tangent of the flame tilt angle can be expressed by the Froude number and is dependent upon the plume length scale. This numerical model can be used to quantify the relationships. Scaling parameters are important as evidenced by results in both laboratory and field experiments. Any study done regarding the effect of wind on fire must be careful to consider the tilt angle of the flame as well as the scaling parameters.

5.9.15 Steady State for Slope and Wind Driven Fires

Slope and wind driven fires can be a challenge to understand and predict. This is a frequent situation with regard to forest fires that occur outside in the presence of wind amid hills, which provide slopes for the fire. Many of the models which attempt to predict and analyze fire behavior assume that there is a steady-state for the propagation of the fire. This steady-state makes for a consistent rate of fire spread (Viegas, 2004). However, this is often not true in the case of wind driven fires, which occur on land that is not flat. In fact, point and linear ignited fires indicated that even when there are relatively uniform conditions present the fire spread may not remain constant. When there is wind or a slope present there are radiation and joint convection effects, which further disrupt a steady spread of the fire.

Experimental results and accurate models have both indicated that wind or a slope will increase the rate of fire spread for a point ignition fire. However, the opposite is true with regard to a linear ignition fire. In this case, the slope or wind will decrease the rate of fire spread. The convection effects which occur when there is a steep slope can create a blowup effect even when there is no wind present. This means that models which do not take into account both slope and wind may not provide accurate results other than in the laboratory.

5.10 Miscellaneous Concerns

5.10.1 Visualization of Simulated Room Fires

Models using computational fluid dynamics and software such as the FDS allow for the visualization of simulated room fires. This can now be done with computers, which are readily available and allow engineers to predict the spreading growth of fire through structures. A number of different strategies have been used for doing a visual simulation of analysis regarding fires within a room. Many of these analyses include fire suppression systems.

Some researchers have suggested that this type of modeling would be aided using fuzzy data types. These types of data are created by superimposing the predictions of a number of different models into a single visualization. This type of research allows for the strengths of each particular model to build upon the findings of other analyses. This provides an interpretation which is often superior to the single models. It has also been found that different applications and target audiences will benefit from alternative interpretations of the data. Present work is being done to expand these types of those realizations to include entire buildings and multiple room fires. The addition of historical data can also be superimposed upon the predictive models to add accuracy.

5.10.2 Physically Based Modeling

Physical-based models can be important for animating and analyzing fires. This is a useful method for both turbulent flames and smooth laminar conditions. It can be useful for the animation of fires, which are created by the burning of gas fuels or solids. Incompressible Navier Stokes equations can be used to independently create models, which represent the behavior of hot gaseous products as well as vaporized fuel (Nguyen, Fedkiw & Jensen). This type of physical modeling accounts for the expansion which occurs when hot gaseous products are formed by vaporized fuel. It can also be used to develop models for use with expansion, which occurs when solid fuels are vaporized into their gaseous states. Physical models have also proven useful for understanding the blue core which forms when radicals are formed through chemical reactions. This occurs when a fuel is converted into byproducts. Physical-based modeling is also helpful to understand how smoke and fire interact with objects in order to expand the fire.

5.10.3 FDS for Crime Scene Investigation

One of the expanded practical uses of computational fluid dynamics and the FDS software is in the reconstruction of fire scenes in which arson is suspected. This type of computer simulation can provide evidence for fire investigators on how a fire was initiated and propagated. These types of computer simulations can often reproduce the fire scene explain how the fire developed and reconstruct the fire process. They can also demonstrate the effects of the building design, ventilation systems, and configurations of the fuel and how smoke moved through the environment during the fire. Furthermore, consideration must be given to the effects that an automatic or manual fire suppression system can have in this type of situation. Parallel processing of data has proven to be an especially effective method for accurate simulations of fire scenes.

A recent investigation into using the FDS software with parallel processing was done to simulate a hotel fire occurring by arson in Taiwan (Shen, Huang & Chien). The results of this experiment demonstrated that this type of simulation provided a good prediction of the smoke movement and fire development within the hotel. This was compared with combustion evidence, which had already been collected at the scene and the description given by individuals that were evacuated from the hotel. The application of this type of fire simulation to actual fire investigations can be an important part of a forensic analysis of a possible arson scene.

5.10.4 FDS and Performance based Design

Computational fluid dynamics in the form of the FDS system has been shown to be useful for fire modeling with regard to a performance-based design environment. With this application, the software can be used to quantify fires, which can occur in commercial buildings. The procedure involved uses medium and full-scale experiments as well as building surveys and computer modeling. A study was recently done using this approach. There was a survey conducted of commercial premises in order to determine the types of combustibles, which were present and the fire loads within the buildings. Statistical data was gathered from a literature review and analyze in order to determine the likely frequency of ignition sources and fires in these types of structures. There was also an analysis of the locations of various materials within the premises (Zalok & Hadjisophocleous).

The results of this study indicated that the data could be used to provide a model which would predict the behavior of fuel packages. Combustible materials and fire loads present in commercial buildings could be properly analyzed. These combine to make fuel packages, which allowed full-scale performance analysis of the commercial buildings. It was also possible to do post flash over testing and the analysis of heat release rate. A determination was made of the concentration of toxic gases within the building as well as compartment temperatures.

These researchers conclude that the FDS simulator is highly effective for determining the effect of fires within a building structure. This type of analysis is especially enhanced when there is data available regarding the fuel packages present in the building. Detailed schematics regarding the building design also proved helpful (Zalok & Hadjisophocleous).

5.11 Chapter Summary

This chapter focused on numerical investigations regarding the nature of fires. It began with an introduction to the field of computational fluid dynamics. There was a brief explanation of some of the many different types of mathematics involved including the Navier-Stokes equations. This was followed by sections which explain the important concepts regarding conservation of mass, momentum, chemical species, and energy.

Several other important topics were discussed prior to focusing on how these equations assist with understanding the effect of wind on fire (Hamins et al.). These subjects included combustion and turbulence, CFD modeling for fire safety engineering, CFD standards and validation, as well as the geometries and uses of multiple mesh sizes. The effects of combustion and radiative heat transfer were explored as well. A final section concentrated on the FDS accommodation regarding hydrodynamics (Kang & Wen).

After these initial topics the chapter progressed to wind. The first concern was how wind could aid flame (Mekki et al.). The relationship between slope and wind regarding fires was also discussed (Perez et al.). A number of other topics were covered such as wind in complex terrain, wind driven transport of firebrands, fire whirls, heat transfer associated with wind aided flame, crosswind effects, wins interacting with large cylinders, and how winds affect atria fires (Maroney). Final discussion topics regarding wind and fire included the relationship of wind velocity, slope, and wind advection. The final section of the chapter concentrated on using fluid dynamics to visualize room fires (Govindarajan, Ward & Barnett). The FDS computer simulations have proven valuable in understanding these types of problems. This shows the value of this numerical approach to the understanding of wind effects on compartment fires. There are times when it is impossible or too resource intensive to conduct actual experiments involving wind effects on compartment fires. When these situations arise, the computer simulations can serve as an alternative method of investigating wind and fire in these situations. They can also serve as a method to help explain conditions which have been observed in actual wind and compartment fire situations.

Chapter Six: Conclusions and Recommendations

This paper began by examining the basic knowledge which exists on compartment fires. It was noted that there are smouldering, non-flash over, and flashover fires (Yung & Benichou). The type of fire is frequently determined by airflow and wind. Most large, modern buildings consist of multiple compartments. These usually are in the form of individual residences or offices. Frequently the residences will have additional divisions for compartments which make up individual rooms of a living space. With regard to a fire, all of these arrangements act as separate compartments. The hot gas layer forms near the ceiling and eventually fills the room then passes out of the compartment as exhaust gas. If there is a door or window fresh air will enter below the hot gases. If the gases reach above 600°C a flash fire will occur (Yung). This will result in all flammable objects within the compartment igniting. This is the situation when emergency personnel arrived and find all flammable materials inside a compartment have burned. The wind effects in a building were described as a case study (Kerber & Madrzykowski) in chapter 3. The 14 experiments of this study demonstrated that wind has a significant effect on compartment fires in buildings. Both natural and mechanical winds were observed at different temperatures, velocities, and pressures. As was expected, the application of significant airflow to a compartment resulted in the temperature near the ceiling increasing to 600°C within a few seconds. This superheated gas was moving at 5 m/s and rolling out of any openings in the compartment. The temperature of this gas would have been lethal to any person in the area who was unprotected. It might even be deadly to those in full fire fighting gear (Kerber).

The case study (Kerber & Madrzykowski) found that there were multiple ways to control fires which occur within the compartments of the building. The first of these is door control. Closing the door resulted in temperatures within a room which were at above 400°C reducing significantly and rapidly. Additionally, it was found that providing wind to the fire and opening a door allow the hot gases to travel at approximately 10 m/s. This resulted in temperatures which exceeded 1000°C. This temperature with the lethal for anyone present even if they were in full firefighter gear with protective breathing. The study also found that window failures could increase airflow and the resulting temperature of the fire.

The experimental research studies reported in chapter 4 involved compartments which consisted of tunnels (Carvel et al.) and apartments (Svensson,). All of these studies supported the findings of the case study. For example, the study done involving fire growth under different wind conditions (Huang et al.) found that an external wind would increase temperature and combustion rates of a fire within a compartment. It is interesting to note that this particular study also found that additional external wind caused a higher rate of fire extinction.

The numerical investigations presented in chapter 5 further supported the findings of a case study presented in chapter 3. The main theories used for numerical investigations involve computational fluid dynamics (Cheung et al.). This is a field which consists of algorithms used to solve complex equations involving the flow of fluids. Despite this field originating for the study of fluids it has become useful for studying gases as well.

There has been substantial progress in computational fluid dynamics during the 21st century (Shen, Huang & Chien). This has been primarily due to increased computing power and the availability of additional memory space in computers. These advances have led to applicability of computational fluid dynamics for many studies involving the dynamics of fire. The studies include those of wind effects on compartment fires. This type of numerical investigation of room fires has proven useful for understanding compartment fires in buildings (Govindarajan, Ward & Barnett). This is especially valuable when resources do not allow an actual study of a fire in a compartment or a case study needs further examination through numerical analysis. The field of wind effects on compartment fires has been well researched, however there does exist areas in which further research might prove helpful. The study (Huang et al.) involving fire growth in a compartment with different wind conditions found that increasing the wind potentiated the fire resulting in a higher temperature. However the extinction of the fire was also accelerated. There has been very little research done on this effect. It is possible that an extreme amount of wind applied to a fire might cause it to extinguish so rapidly as to reduce its damage potential. Emergency personnel are already prepared to carry fans into fire situations in order to provide positive pressure ventilation. An extension of this practice might prove useful but would require further research.

Research being done on the extinction rate of a fire due to high wind levels could begin with wind tunnels. This would involve placing a compartment in the tunnel which simulates a typical room of an apartment or office. Wind would then be applied in recordings made of the fire. The wind could be accelerated rapidly and the rate with which the fire is extinguished recorded. It is possible that there is a critical wind speed beyond which a fire is extinguished so rapidly that minimal damage is done. If this wind speed is within reasonable limits, it could be that high-speed air devices could be made portable and used by emergency personnel to both extinguish a fire quickly and clear an area of dangerous gases which may be superheated. The practicality of this approach depend upon the critical wind speed necessary for fire extinguishing as well as the technology to provide portable units for providing wind at this speed. This approach would have the advantage of clearing the area of gases as well as doing less than water to compartments below the origin of the fire. The possibility that 21st-century firefighting could work on the same principle as blowing out a candle is intriguing in may prove fruitful.

References

American Society for Testing and Materials (2004), ASTM E 1355-04, Standard Guide forEvaluating the Predictive Capabilities of Deterministic Fire Models, 2004.

Andrews, P.L., Bevins, C.D., Seli, R.C., (2008) BehavePlus fire modeling system, version 4.0: User's Guide. USDA, Forest Service, Rocky Mountain Research Station, General Technical Report.

Balbi, J. H., Rossi, J. L., Marcelli, T., & Chatelon, F. J. (2010). Physical modeling of surface fire under nonparallel wind and slope conditions.Combustion Science and Technology, 182, 922-939.

Battaglia, F., McGratten, K. BE., Rehn, R. G., & Baum, H. R. (2000). Simulating fire whirls. Combustion Theory Modeling, 4, 123-138. Carlsson, J. (1999). Fire modeling using CFD. Lund, Sweden: Department of Fire Safety Engineering, Lund Institute of Technology.

Carvel, R. O., Beard, A. N., Jowitt, P. W., & Drysdale, D. D. (2005). Fire size and fire spread in tunnels with longitudinal ventilation systems. Journal of Fire Sciences, 23, 485-518.

Cheung, A.L., Lee, E.W., Yuen, R.K., Yeoh, G.H. & Cheung, S.C. (2008), "Capturing the pulsation frequency of a buoyant pool ?re using the large eddy simulation approach", Num. Heat Transf. Part A 53, pp. 561-576.

Cheung, S.C., Yuen, R.K., Yeoh, G.H & Cheng, G.W. (2004), "Contribution of soot particles on global radiative heat transfer in a two-compartment fire", Fire Safety Journal, 39, pp. 412-428.

Cox, G. & Kumar, S. (2002), "Modeling Enclosure Fires Using CFD", The SFPE Handbook of Fire Protection Engineering, Section 3 / Chapter 8, SFPE, Maryland.

Ezekoye, O. A., & Nicks, R. (2005). Effects of PPV attack on thermal conditions in a compartment downstream of a fire. Fire Technology, 41, 193-208.

Govindarajan, J., Ward, M., & Barnett, J. (1999). Visualizing simulated room fires. IEEE, 475-478.

Hall, J. R..(2005) High-rise Building Fires. National Fire Protection Association, Fire Analysis and Research Division, 1 Batterymarch Park, Quincy, MA.

Hamins, A., Maranghides, K, McGrattan T, Ohlemiller R., & Anleitner A., ( 2005) "Experiments and Modeling of Multiple Workstations Burning in a Compartment", NIST NCSTAR 1-5E, National Institute of Standards and Technology, Gaithersburg, Maryland.

Hostikka, S., McGrattan, K.B., & Hamins, A., (2003) in: Proc. Seventh Intl. Symp - IAFSS, "Numerical Modeling of Pool Fires Using LES an Finite Volume Method for Radiation", Intl. Assoc. for Fire Safety Science, Worcester, MA.

Huang, H., Ooka, R., Liu, N., Zhang, L., Deng, Z., & Kato, S. (2009). Experimental study of fire growth in a reduced-scale compartment under different approaching external wind conditions. Fire Safety Journal, 44, 311-321.

Hume, B. (2003) "Development of Standards for Fire Field Models", Fire Research Report No. 85, ODPM, London.

Ingason, H., & Li, Y. Z. (2011). Model scale tunnel fire tests with point extraction ventilation. Journal of Fire Protection Engineering, 21(1), 5-36.

Jensen, K.A., Ripoll, J.F., Wray, A.A., Mouna, D.J. (2007). "On various modeling approaches to radiative heat transfer in pool ?re." Combust. Flame 148, pp. 263-279.

Kang, Y. & Wen, J.X, (2004). "Large eddy simulation of a small pool fire." Combust. Sci. Technol.,176, pp. 2193-2223.

Kerber, S. & Walton, W. D. (2006) Full Scale Evaluation of Positive Pressure Ventilation in a Fire Fighter Training Building. National Institute of Standards and Technology, Gaithersburg, MD.

Kerber, S. &Walton, W. D.(2005) Effect of Positive Pressure Ventilation on a Room Fire. National Institute of Standards and Technology, Gaithersburg, MD.

Kerber, S. (2006) Evaluation of the Ability of Fire Dynamic Simulator to Simulate Positive Pressure Ventilation in the Laboratory and Practical Scenarios." National Institute of Standards and Technology, Gaithersburg, MD.

Kerber, S.& Madrzykowski, D.(2007) Stroup, D. Evaluating Positive Pressure Ventilation In Large Structures: High-Rise Pressure Experiments. National Institute of Standards and Technology, Gaithersburg, MD.

Kerber, S., & Madrzykowski, D. (2009). Firefighting tactics under wind driven fire conditions: 7-story building experiments. US Department of Commerce Building and Fire Research Laboratory, National Institute of Standards and Technology, Gaithersburg, Maryland.

Klopvac, S., & Turan, O. F. (1998). Flames venting externally during full-scale flashover fires: 2 sample ventilation cases. Fire Safety Journal, 31, 117-142.

Kunsch, J. P. (2002). Simple model for controlled fire gases in a ventilated tunnel. Fire Safety Journal, 37, 67-81.

Li, Y. Z., Lei, B., & Ingason, H. (2010). Study of critical velocity and back layering length and longitudinally ventilated tunnel fires. Fire Safety Journal, 45, 361-370.

Madrzykowski, D. & Walton, W.D. (2004) Cook County Administration Building Fire, 69 West Washington, Chicago, Illinois, October 17, 2003: Heat Release Rate Experiments and FDS simulations. National Institute of Standards and Technology, Gaithersburg, MD.

McGrattan, K. & Forney, G., (2004). Fire Dynamics Simulator (Version 4), User's Guide. NIST Special Publication 1019, National Institute of Standards and Technology, Gaithersburg, Maryland.

McGrattan, K., (2004). Fire Dynamics Simulator (Version 4) Technical Reference Guide. NIST Special Publication 1018, National Institute of Standards and Technology, Gaithersburg, Maryland.

McGrattan, K., Floyd, J.E., Forney, G.P., Baum, H.R. & Hostikka, S., (2003). "Improved radiation and combustion routines for a large eddy simulation model."Fire Safety Science, Proceedings of Seventh International Symposium,pp. 827-838.

Mekki, K., Atreya, A., Agrawal, S., & Wichman, I. (1990). Wind-aided flame spread over charring and non-charring solids: an experimental investigation. 23rd Symposium on Combustion, The Combustion Institute, 1701-1707.

Meroney, R. N. (2010). Wind effects on atria fires. The 5th International Symposium on Computational Wind Engineering, 1-7.

Morvan, D., Porterie, B., Loraud, J. C., & Larini, M. (2001a). A numerical investigation of cross wind effects on a turbulent buoyant diffusion flame. Combustion Science and Technology, 164, 1-35.

Morvan, D., Porterie, B., Loraud, J. C., & Larini, M. (2001b). Effects on a turbulent buoyant diffusion flame. Combustion Science and Technology, 164, 1-35.

Nguyen, D. Q., Fedkiw, R., & Jensen, H. W. (2002). Physically based modeling and animation of fire. Association for Computing Machinery, 721-728.

Nmira, F., Consalvi, J. L., Boulet, P., & Porterie, B. (2010). Numerical study of wind effects on the characteristics of flames from non-propagating vegetation fires. Fire Safety Journal, 45, 129-141.

Perez, Y., Pastor, E., Agueda, A., & Planas, E. (2010). Effect of wind and slope when scaling the forest fires rate of spread of laboratory experiments. Fire Technology.

Pope, N. & Bailey, C. (2006) "Quantitative comparison of FDS and parametric fire curves with post-flashover compartment fire test data." Fire Safety Journal , 41, pp. 13-22.

Santoni, P. A., Simeoni, A., Rossi, J. L., Bosseur, F. , Morandini, F. , & Silvani, X. , et al. (2006). Instrumentation of wildland fire: Characterization of a fire spreading through a Mediterranean shrub. Fire Safety Journal, 41, 171-184.

Sharples, J. J., McRae, R. H., & Weber, R. O. (2010). Wind characteristics over complex terrain with implications for Bush fire risk management. Environmental Modeling & Software, 25, 1099-1120.

Shen, T., Huang, Y., & Chien, S. (2008). Using fire dynamics simulation (FDS) to reconstruct an arson fire scene. Building and Environment, 43, 1036-1045.

Simeoni, A., Santoni, P. A., Larini, M., & Balbi, J. H. (2001). On the wind advection influence on the fire spread across a fuel bed: modeling by semi-physical approach and testing with experiments. Fire Safety Journal, 36, 491-513.

Simiu, E. & Scanlan, R. H.(1996) Wind Effects on Structures, Fundamentals and Applications to Design, 3rd ed.. John Wiley & Sons, New York, NY.

Suo-Antilla, J., & Gritzo, L. (2009). The effects of wind on fire environments containing large cylinders. Combustion Science and Technology, 181, 68-77.

Tieszen, S.R. , O'hern, T.J., Schefer, R.W. , Weckman, E.J. & Blanchat, T.K. (2002) "Experimental study around a one meter diameter methane." Comb. Flame, 129, pp 378-391.

Verification and Validation of Selected Fire Models for Nuclear Power Plant Applications, Volume 1 (2006): Main Report, U.S. Nuclear Regulatory Commission, Office of Nuclear Regulatory Research, Rockville and Electric Power Research Institute, Palo Alto, CA.

Verification and Validation of Selected Fire Models for Nuclear Power Plant Applications, Volume 7 (2006): Fire Dynamics Simulator, U.S. Nuclear Regulatory Commission, Office of Nuclear Regulatory Research, Rockville and Electric Power Research Institute, Palo Alto, NUREG-1824 and EPRI 1011999 (Draft).

Viegas, D. X. (2004). On the existence of a steady-state regime for slope and wind driven fires. International Journal of Wildland Fire, 13, 101-117.

Wakatsuki, K. (2005) High Temperature Radiation Absorption of Fuel Molecules and an Evaluation of Its Influence on Pool Fire Modeling, Ph.D. thesis, University of Maryland: College Park, MD.

Wang, H. H. (2009). Analysis on downwind distribution of firebrands sourced from a wildland fire. Fire Technology.

Weise, D. R., & Biging, G. S. (1996). Effects of wind velocity and slope on flame properties. Can. J. For. Res., 26, 1849-1858.

Xin,Y. & Gore,J.P. (2005) "Two dimensional soot distributions in buoyant turbulent ?res. Proceedings of 30th Symposium (International) on Combustion, The Combustion Institute, pp. 719-726.

Xin,Y., Gore, J.P., Mcgrattan, K.B., Rehm, R.G., & Baum, H.R, (2005) "Fire dynamic simulation of a turbulent buoyant ?ame using a mixture fraction-based combustion model", Comb. Flame, 141, pp. 329-335.

Yeoh, G.H. Yuen, R.K. Cheung, S.C. & Kwok, W.K. (2003) "On modeling combustion, radiation and soot processes in compartment ?res." Build. Environ., 38, pp. 771-785.

Yeoh, GH, Yuen, RK & Chen, DH (2003) "On numerical comparison of enclosure ?re in a multi-compartment building." Fire Safety J., 38, pp. 85-94 Yung, D. T. (2008). Principles of fire risk assessment in buildings. West Sussex, United Kingdom: Wiley.

Yung, D.T., & Benichou (1999). Design fires for fire risk assessment and fire safety designs. Proceedings of the Proceedings of the fire protection research foundations 1999 fire risk and hazard research applications symposium, San Diego, California(pp. 101-112). Ottawa, Ontario, Canada: NRC-IRC.

Zalok, E., & Hadjisophocleous, V. (2009). Assessment of the use of fire dynamics simulator in performance-based design. Fire Technology.

Zou, G., & Chow, W. (2005) "Evaluation of the field model, Fire Dynamics Simulator, for a specific experimental scenario." Journal of Fire Protection Engineering, 15 (2) pp. 45-56.