Experimental Study on the Fire-Spreading Characteristics and Heat Release Rates of Burning Vehicles Using a Large-Scale Calorimeter

In this article, large-scale experimental studies were conducted to figure out the fire characteristics, such as fire-spreading, toxic gases, and heat release rates, using large-scale calorimeter for oneand two-vehicle fires. The initial ignition position was the passenger seat, and thermocouples were attached to each compartment in the vehicles to determine the temperature distribution as a function of time. For the analysis, the time was divided into sections for the various fire-spreading periods and major changes, e.g., the fire spreading from the first vehicle to the second vehicle. The maximum temperature of 1400 °C occurred in the seats because they contained combustible materials. The maximum heat release rates were 3.5 MW and 6 MW for one and two vehicles, respectively. Since the time to reach 1 MW was about 240 s (4 min) before and after, the beginning of the car fire appears to be a medium-fast growth type. It shows the effect on the human body depending on the concentration of toxic substances such as carbon monoxide or carbon dioxide.


Introduction
Many vehicle fire studies are being carried out since vehicle accidents result in catastrophe due to various reasons such as drivers' negligence, electrical faults, deliberately lit and arson in parking lots [1], buildings, and tunnels, and so forth. To prevent vehicle disasters and suggest a guideline for evacuees in case of emergency, figuring out the fire-spreading mechanism and measuring other key parameters such as heat release rates and temperatures are necessary to install sprinklers, smoke control systems, and setting up fire extinguishers.
In fire researches, calorimeters, which can estimate a heat release rate, smoke generation, carbon dioxide, carbon monoxide, and so on, have been widely used for experimental study [2][3][4][5]. In the case of heat release rate, it has been regarded as one of the most important factors during experimental studied since it can be used to calculate the size of the fire and be more easily used in performancebased design (PBD). Large-scale tests are considered the most accurate way to secure fire-spreading effect and heat release rate, however, there is limited research, as it cost too much [6][7][8]. Katsuhiro et al. [6] found that the temperature distribution and maximum HRR reached 3 MW by changing the initial ignition location of the fire. In addition, they found that the fire spread radically after the windows were broken. Throughout related large-scale experiments, a single vehicle represents the heat release rate of 2.5-5 MW. However, most of the experimental studies were conducted only on one vehicle. Vehicle fires normally occur between two vehicles because of accidents. Li et al. [9] conducted large-scale experiments using two sedan vehicles to determine the change of the temperature distribution; however, in this study, the heat release rate was not considered, although it is an imperative part not only for analyzing the fire phenomena, but also suggesting guidelines for rescue as a function of time. Therefore, we investigated fire-spreading phenomena considering the effect of two vehicles fire scenario to figure out an important parameter such as heat release rates, fire-spreading time, and so forth.
A secondary factor that causes catastrophe in vehicle fires is the influence of toxic gases, e.g., carbon monoxide (CO) and carbon dioxide (CO2), on human breathing and other functions [10][11][12]. Truchot et al. [13] investigated toxic-gas emissions from vehicle fires in tunnels. They did experimental studies using large-scale facilities, where the air flow in the tunnel could be controlled. They found and measured the toxic substances through a flame ionization detector/analyzer. It showed the toxic substances emitted in the tunnel shaft and analyzed the effects of various heat release rates. Smoke and toxic gases that are produced in the open are released into the air, due to their buoyancy. However, when a vehicle burns in an enclosed space, e.g., a tunnel or an indoor parking lot, the gas cannot escape to the atmosphere. Therefore, it is one of the main factors that should be considered in fire research to understand the concentration of the toxic substances generated when a vehicle fire occurs. However, there is limited information regarding the toxic gases that appear in the event of the fire.
Furthermore, many experimental studies are being actively carried out with pool-fire tests to represent fire accidents in tunnels or high-rise buildings [14][15][16][17][18][19][20][21][22][23]. Beak et al. [23] conducted tunnelfire experiments using small-scale tunnels with pool fires, because it is challenging to experiment in real tunnels with real vehicles. He provided detailed information on the fire's heat release rate and its hole-plugging effect on tunnels; however, the results were obtained from pool-fire experiments, not real vehicles. It is clear that the results obtained from real products and pool-fire experiments will be different.
Therefore, it is necessary to analyze the fire-spreading characteristics of an actual vehicle and analyze the heat release rate and temperature distribution in order to understand the fire phenomena and its applications. Figure 1 represents a schematic diagram of the basic concept for this study. In this study, the fire-spreading characteristics, unsteady heat release rate, and the toxic substances in vehicle fires are measured using a large-scale calorimeter. All information, such as the change of heat release rate, temperature distributions and toxic gases as a function of time, gained throughout this research can be applied to other fire-related researches regarding vehicles and tunnels, parking lots, and so forth.

Experimental Setup and Conditions
Experiments were conducted using a large-scale calorimeter (LSC) from the Korea Institute of Construction Technology (KICT), applicable up to 10 MW, to determine the unsteady heat release rate and toxic gases, e.g., CO and CO2, generated by the fire. The schematic diagram of LSC is represented in Figure 2 to figure out the values regarding change of the heat release rate and toxic substances as a function of time, and specifications of the experimental apparatus are represented in Table 1.  To investigate the fire-spreading characteristics and temperature distributions for one and two vehicles, K-type thermocouples (OMEGA, measuring range: −200-1260 °C) were attached to the engine room, bumpers, seats, and fuel tank. Temperature data were transmitted to a data-acquisition unit (DAQ) on a PC every second, and the experiments were recorded by a video recorder. The initial ignition location was assumed to be the passenger seat. Figure 3 represents the experimental setup and conditions. In the two-vehicle experiment, the distance between vehicles was set to 50 cm, and the thermocouples were attached in the same positions. The fuel was almost eliminated to prevent explosions during the experimental study. Four-door sedan vehicles were used in the experiments. The vehicle size was 4.7 m (length) × 1.8 m (width) × 1.4 m (height), and was Hyundai's EF sonata released in 1998. These vehicles have not been used for about 10 years, and the experiment was carried out after the pressure of tires were removed with opening full front windows. The location and number of thermocouples in each compartment are represented in Figure 4. Thermocouples were attached to the engine room, bumpers, seats, and fuel tank. Sixteen thermocouples were attached to the engine room, four each at the top and bottom, and two on each side. Ten thermocouples were attached to the bumpers; five each on the front and rear. In addition, 14 thermocouples were attached to each seat in the vehicle interior, representing the head, waist, legs, and feet. Twelve thermocouples were attached to the fuel tank, one on each side, and four each on the top and bottom. Thus, 42 thermocouples were used to figure out the change of the temperature distribution as a function of time.

Fire-Spreading Characteristics
Fire-spreading over one and two vehicles as a function of time while on experiments are presented in Figure 5. Also, a flow-chart about fire-spreading inside a vehicle and two vehicles as a function of time are represented in Figure 6. These represent the major fire-spreading points as a function of time. As represented in the Figure 5 and the flow-chart, in the one-vehicle experiments, the fire spread to the driver seat and rear seat almost simultaneously, followed by the fuel tank, engine room, and finally the front and back bumpers. In the two-vehicle experiments, the fire followed the same sequence until it spread to the next vehicle, due to radiation. It then spread simultaneously to the rear seat and driver seat of the second vehicle, then the engine room, fuel tank, and bumpers in a regular sequence. Detailed information on the fire-spreading characteristics and special events as a function of time are represented in Table 2.   Table 2. Detailed information on the fire-spreading characteristics and special events as a function of time.

Fire Reaches Maximum Temperature and Goes Out First Vehicle
Second Vehicle 0 Ignition in passenger seat 150 Fires active in passenger seat 300 Fire spreads to driver and rear seats almost simultaneously 500 -Fire spreads to the second vehicle 700 Fire spreads to fuel tank Fire dramatically spreads to seats 900 -Fire spreads to fuel tank 1000 Fire spreads to engine room -1100 -Fire spreads to engine room 1250 -Fire spreads to bumpers 1500 Fire spreads to bumpers -1500-3600 Fire goes out Fire goes out Therefore, the main temperature changes in the seats, engine room, fuel tank, and bumpers were represented. The temperature distribution on seats are represented in Figure 7. The passenger seat actively caught fire around 300 s (5 min) after ignition, and the highest temperature was observed in the interior, due to its combustible materials compared with the other compartments. The fire spread to the driver and rear seats simultaneously; however, the temperature differed slightly, as can be seen in Figure 8, which represents the temperature distribution of the driver and rear seats for the two sets of vehicles. The temperature increase in the driver seat and the rear seat can be observed almost simultaneously. The beginning of the temperature increase and the highest temperature are represented as the green and red lines, respectively. This means that the fire in the driver seat was larger than the rear seat, even though the fire-spreading time was similar because the driver seat was located closer to the fire than the rear seat.
To investigate the effect of temperature propagation on the number of vehicles, the driver seat and the rear seat inside the first vehicle are represented in Figure 8a,b. In the case of Figure 8c,d, it showed the driver seat and the rear seat inside the first vehicle when the two-vehicle experiment. After 500 s (8 min 20 s), the fire spread to the next vehicle, and the interior's temperature rapidly increased around 700 s (11 min 40 s), as shown in Figure 8c,d. Since the second passenger seat was located next to the first vehicle, the fire reached the thermocouple's temperature limit of 1370 °C.
The analysis was begun by dividing the time into sections for the various fire-spreading periods and major changes. Section 1 represented the period of the fire spreading from the passenger seat to the driver and rear seats after 300 s (5 min). Section 2 was 500 s (8 min 20 s) after the ignition, in which the fire spread to the next vehicle. Section 3 was when both the passenger seat and driver seat of the second vehicle caught fire. Section 4 was the period when the HRR reached maximum in the case of two-vehicle fire. These same sections were used for representing the heat release rates and the toxic substances in Sections 3.2 and 3.3, respectively.  The temperature distribution in engine room and fuel tank were represented in Figure 9. The engine room was comprehensively burning within about 1200 s (20 min) of the fire ignition. In addition, the fire in the initial vehicle did not propagate directly to the engine room, but first spread to the seats. After about 1300 s (21 min 40 s), the fire spread to the adjacent vehicle. The fire went out 2500 s (41 min 20 s) after reaching the maximum temperature.
In the fuel tank, the temperature change was relatively small compared to the other combustibles. This was because most of the fuel was removed; however, in a real vehicle fire, it is the most vulnerable and dangerous compartment for the fire to reach.  The temperature distributions at the bumpers are represented in Figure 10. For the front bumper, the temperature rose from the place nearest the initial fire vehicle. After about 1400 s (23 min 30 s), the temperature rose in the next vehicle. The temperature rose later because it was located farthest from the first vehicle. In addition, in the rear bumper, the fire spread more quickly than the front bumper because it was relatively close to the seat.

Heat Release Rate
The heat release rates considering the unsteady fire phenomenon are represented in Figure 11. As can be seen in Figure 6, the heat release rate increased rapidly around 180 s (3 min) after ignition because the fire spread into the vehicle interior. Further, when the windshield broke around 300 s (5 min), the heat release rate increased sharply to about 2.3 MW. However, it continued to increase because the fire spread from the passenger seat to the driver seat. The passenger-seat cushion spontaneously ignited and then temporarily decreased. Because the cushion was composed of a composite material, the heat release rate increased and decreased repeatedly until it reached its maximum.
For the engine room and the fuel tank, the increasing and decreasing heat release rate was not apparent because there was little fuel in the engine room and the engine oil was low. After reaching the maximum heat release rate of 3.5 MW, the fire slowly went out.
In the case of the two-vehicle experiments, the heat release rate followed a similar pattern until the initial window breakage. After that, the fire spread to the entire passenger seat and then the next vehicle at about 600 s (10 min), due to radiation. After the ignition, the fire spread rapidly through the open passenger window. The spacing between the vehicles was about 50 cm, and there was a 2 MW fire around the vehicle. When both vehicles were burning, the fire spread dramatically to all of the seats. The maximum heat release rate was about 6 MW, and the fire went out gradually. Since the time to reach 1 MW was about 240 s (4 min), the fire growth in a vehicle is considered to be a mediumfast fire. Figure 11. Heat release rate for one and two vehicles as a function of time.
The heat release rate is one of the most important parameters to calculate the flame height [24,25]. McCaffrey [24] presented several formulas under various conditions, but, one of the functions under the open condition can be calculated the flame height using heat release rate. McCaffrey's formula is as follows: Furthermore, the height of the flame in normal atmospheric conditions was indicated by nondimensional analysis [25], and it is as follows:  Figure 12 and Table 3, respectively. However, measuring the change of flame heights was affected significantly under various conditions. The fire growth can be evaluated using the following generic fire-growth curve: where a is the fire-growth coefficient (kW/s 2 ), t means time (s), and ti means the time of ignition (s).
In this study, the time of ignition can be taken as zero. The fire-growth coefficients for vehicle fires up to 1 MW are represented in the Appendix A. Reducing the flame height and fire growth were decisive parameters for suppressing a fire; a fire extinguisher should be considered, especially in enclosed spaces, e.g., tunnels and indoor parking lots.
It is generally well acknowledged that fire experiments are really difficult to do repeated experiments, due to time consumption and expensive time. Thus, various fire articles did not contain a detailed uncertainty analysis. Melcher et al. [26] suggested the impact of random deviations that may occur in a single experiment. Mass loss rates and heat release rates for one and two vehicles were represented with error range in Figure 13. Yellow and lavender indicate the error ranges that occurred in one and two vehicles, respectively.
Applying the actual heat release rates obtained from this study to the numerical analysis study, Park et al. published a study on the effect of a vehicle accident on the evacuation in various tunnel aspect ratio [27].  Table 4 and compared with the results obtained through this experiment. As a rule of thumb, the maximum heat release rates were normally represented 2.5 MW to 5 MW. Okamoto et al. [8] found the heat release rates as a function of time in real-scale experiments, and the vehicles used in this experiment were similar to those used in this study. The experimental conditions were quite similar, such as conducting the vehicle fire test in the absence of fuel, but only with different experimental measuring equipment. In addition, the maximum heat release rate was 3.5 MW, the same as this study. However, in the case of research done by Shipp et al. [28], the maximum heat release rate was significantly higher than other studies since the vehicle tests involved fuel spill from the petrol tank. Furthermore, Ingason [29] conducted the vehicle fire experiments in a tunnel so that the maximum heat release rates were different because of ceiling temperatures, ventilation system on maximum heat release rate, and fire growth rates.

Toxic Substances
The changes in the amounts of carbon dioxide and carbon monoxide as a function of time are represented in Figure 14. These values represent the toxic gases obtained through the LSC from the vehicle as it burned. The main changes on carbon monoxide and carbon dioxide as a function of time are represented in Tables 5 and 6, respectively. Furthermore, it showed similar trends to the distribution of change of heat release rate. However, judging from the influence on the human body, it is obvious that these values might be not correct, because of the many differing conditions in fire phenomena, e.g., enclosed or open spaces, scale of space, ventilation systems, sprinklers, and so on.
With fires in open spaces, the smoke and toxic gases will be released into the air. In enclosed spaces, e.g., tunnels or indoor parking lots, the smoke will accumulate continuously, and the concentration will greatly increase. However, the effects on the human body can be analyzed based on concentrations of CO and CO2 from the vehicle itself [30].
Based on the values obtained from the LSC, the CO2 generated from one vehicle will give a person a slight headache. However, in a two-vehicle fire, the slight influence occurs before 1000 s (16 min 40 s); however, after that, it will affect the respiratory system and nervous system, and cause cerebral impairment. Furthermore, only a slight headache can be felt about 500 s (8 min 20 s) after ignition. However, after that, the area should be evacuated. In the worst case, life is threatened, and evacuation may be difficult because of the high concentration of CO over 1200 s (20 min). The higher the heat release rate is, the higher the concentration of CO and CO2. Therefore, the larger the fire is, the faster the area should be evacuated. Based on the CO and CO2 concentrations absorbed through the LSC, the impact on human health applies when people were directly exposed. It is apparent that the values obtained through the LSC will vary depending on the location and conditions of the fire inside of buildings, but it can provide indirect guidance on ventilation and evacuation in compartment space while presenting the toxic concentration occurring in the vehicle itself.

Conclusions
The aim of this study was to investigate how fire spreads through a vehicle using an actual vehicle-fire test, and to analyze the heat release rate for two vehicles using a large-scale calorimeter. In addition, the influence of the toxic gases generated from the vehicle fires was analyzed. An analysis of the experimental results provided the following conclusions.
(1) In actual vehicle tests, the fire spread from the initial ignition location to the rear seat, engine room, fuel tank, and bumper in regular sequence. In the two-vehicle situation, a similar tendency was observed, and the fire spread to the next vehicle after about 500 s (8 min 20 s).

Conflicts of Interest:
The authors declare no conflict of interest.