Fire Performance of Automated Tall Car Park Structures

: Automated tall car park structures are modern alternatives to conventional parking structures to save space and volume in highly demanded parking regions in urban areas. -The design of such structures has significant knowledge gaps especially in regarding the effects of fire spread between passenger cars. The purpose of this study is to estimate the horizontal and vertical fire spread between passenger cars in automated tall car park structures and provide fire safety design to eliminate fire spread and possible structural collapse. The fire spread between cars is established by estimating irradiance heat flux of each car component. An 8-floor automated tall car park structure is designed in accordance with European standards. The results show that steel car pallets underneath cars reach to 1000 degree Celsius in early phases of fire, which could potentially cause a structural failure. Without any fire protection on the structure, the fire spreads to the neighboring cars in 25 minutes and to the cars above in 20 minutes. Significant fire protection is needed to eliminate fire spread between passenger cars. A more effective sprinkler system is also proposed to suppress the car fire.


Introduction
The purpose of this study is to model  Passenger car fires in parking structures are not common, but such events can develop into large and uncontrollable fires [2]. Some cases of car park fires have involved hundreds of vehicles and caused structural collapses in the last two decades. A car park fire in Schiphol Airport in Amsterdam burned nearly 30 passenger cars and partly damaged around 101 passenger cars [3]. Another vital car park fire event is Kings Dock fire in Liverpool [4]. The fire led to a total loss of 1150 passenger cars. Firefighters have reported rapid lateral fire spread and vertical fire spread through both downward and upward. The most recent open car park fire has occurred in Stavanger, Norway [5]. With the effect of strong wind, flames have engulfed the structure partially, a significant structural collapse has occurred, and around 300 passenger cars were destroyed. A few passenger car fire tests in full-size open car parks are carried out [6][7][8][9]. The general conclusion was that structural fire protection is not necessary for open deck car parking structures.
Automated tall car park structures are modern alternatives to conventional parking structures to save space and volume in high-demanded parking regions in urban areas. They are generally constructed from steel. An 8-floor car park structure from Balikesir, Turkey, is shown in Fig. 1. It is an externally braced steel structure with embedded elevator and car pallets to store and retrieve passenger cars. These structures may be constructed with either open or closed façade. Fire characteristics of passenger cars depend on vehicle size, ignition source and location, environmental conditions and ventilation level. Peak HRR levels vary in a wide spectrum, from 1.9 MW to 10.8 MW [2]. The high level of heat release rate per unit area during car fires increases the collapse risk of car park structures. Most of fire spread tests are based on lateral fire spread scenarios in the literature [10,11]. Weisenpacker et al. [11] focused on temperature levels around and inside of burning cars and did not measure the heat release rates (HRR). On the contrary, Park et al. [ [12]).
The use of total combustion energy can classify passenger car fires during fire. The calorific potential classification that has five different categories is generally based on vehicle size and curb weight. Schleich et al. [12] have presented equivalent HRR curves for the potential calorific classifications of passenger cars in Fig. 2. The characteristics and trends of the HRR curves are identical.
Schleich et al.'s [12] approach is based on the amplification of HRR values. In addition to the overall energy potential, ignition mechanism of individual components in vehicles are also essential to understand car fires. After the 1990s, the use of plastic components in passenger cars has increased from 5.1% in 1970s to 8.8% of car weight in 2018 [2]. In addition to this, the average combustion energy per mass of plastic components within passenger cars has increased. Although the change in the mass of plastic materials is 72%, the total combustion energy has increased by 91% since the 1970s.
The risk of fire spread from one car to another has risen with the increase in the use of plastic materials for exterior parts of passenger cars.
The occurrence of ignition can be forecast with a calculation of surface temperature if the substance is heated by convection and radiation. A set of test results of plastic vehicle components that used on outmost surfaces of passenger cars are given in Table 1 [13]. Ignition times under different irradiance levels and critical irradiance levels of plastic components are determined by tests [13]. All exterior plastic components excluding tires can be ignited by an irradiance level of 20 kW/m 2 in 7.5 minutes, whereas it takes less than 1 minute with an irradiance level of 30 kW/m 2 . On the other hand, the ignition of tires takes much more time, but its critical irradiance level is not high. Building Research Establishment (BRE) conducted a set of fire tests on fire spread modes between passenger cars [13]. All possible fire spread patterns between passenger cars are tested. These patterns are side-by-side fire spread, nose-to-nose fire spread and vertical fire spread at car stacker.
Before and after photographs of tests are shown in Fig. 3 [17]. BRE has conducted tests on the efficiency of sprinklers on passenger car fires for ordinary car parks and stacker systems [18]. The setup represents a closed car park, but ventilation is enough for a fuel-controlled fire. The fire started at the outermost passenger car. The first sprinkler was activated after 4 minutes, then all sprinklers were activated at the early stage of fire. The fire was not extinguished, and its thermal power reached around 7000 kW. However, the fire did not spread to the next vehicle. Cooling and transport effect of water droplets from sprinklers caused the smoke to drag down.

Methodology
In this study, an automated open tall steel car park structure is designed based on an existing example shown in Fig. 1 and modelled in Fire Dynamics Simulator (FDS) [19] and SAP2000 [20]. A widely used design car fire curve is modified to decrease computational demand of FDS and vertical and lateral fire spread criteria between passenger cars is defined via FDS simulations. Thermal response of structural members including the car pallets are obtained from FDS analyses.

Car park structure
An 8-floor open-facade steel car park with 4 units and a capacity of 56 cars is designed according to Turkish Building Earthquake Code 2018 and Eurocode 3 [21,22]. Construction material is chosen as S235 grade carbon steel. The structure is braced for lateral resistance. All column cross sections are TUBO 160x160x10. All beams are HEA100 and connect to columns via shear connections.
The structural design is shown in Fig. 4. Member cross sections are tabulated in Table 2.
(a) Car pallets stay on wheels, which are mounted to short cantilever beams fixed to the columns.
The car pallet is illustrated in Fig. 5. The pallet contains 4 longitudinal beams and 4 cross beams   Table 2. Member sections of car park structure (see Fig. 4b).

Member Section
Columns TUBO160x160x10 Beams HEA100 Bracing TUBO80x80x8 Car pallet HEA100 The thermo-mechanical analysis of the cark park structural system is not conducted since both columns and beams under low utilization ratios have significantly high critical temperatures even without any fire protection. TUBO160x160x10 columns with = 0.129 have = 809 ℃ . On the other hand, HEA100 beams underneath the car pallet carry a significant load and they are likely subjected to extreme temperatures from car fires just below. HEA100 beams with = 0.502 have = 590 ℃. A possible collapse mechanism of the car pallet is illustrated in Fig. 6 where a plastic hinge at the midspan forms during fire. Such failure indicates that the pallet collapse can occur before the fire spreads to other cars. Fig. 6. Car pallet collapse mechanism.

FDS model
The Fire Dynamics Simulator (FDS) model is created in PyroSim [23]. The fire initiates from the car labeled as V13 on the 2 nd floor as illustrated in Fig. 4a. No smoke control system is installed on the model. The heat and radiation transport calculation is performed with polyurethane as the fuel load and with 20 cm mesh size [24]. HEA100 and TUBO160x160x10 structural member surface temperatures are calculated utilizing the 'exposed back condition' in the FDS model, i.e. assuming that the members conduct heat through the cross-sectional thickness.
To estimate this cell size in the FDS model, the characteristic length scale of fire is calculated using Equation 1, where D* is dimensionless diameter in m, Q is peak heat release rate (HRR) in kW, ∞ is ambient air density in kg/m3, ∞ is ambient temperature in °C, heat capacity of air under constant pressure in J/kgK and g is gravitational acceleration in m/s 2 . * = ( The heat release rate curve of Category III car is with a peak HRR of 8.3 MW, hence D* is obtained as 2.235 m. It is suggested that for a reliable large eddy simulation (LES) at least 10 cells shall fit within the dimensionless diameter [25]. Therefore, 20 cm or smaller cell size with a simple chemistry model can be used. Cell size is chosen as 10 cm for all fire simulations in this study. The Prandtl number is taken as 0.7. The Radiative Transfer Equation (RTE) is used with a radiation fraction of 0.35. It is prescribed as a lower bound in order to limit uncertainties in radiation calculation.

Passenger car design fire
To reduce computational effort and shorten the run time on FDS simulations, both the heat release rate (HRR) curve and the vehicle model detail is modified without compromising the accuracy.
For the passenger car design fire, Category III HRR curve with a peak HRR of 8.  Fig. 7. Category III design car fire [12] and Modified Category III fire.
As can be seen in Fig. 7, the cumulative thermal energy is shifted to the left in the Modified Category III fire with respect to the original curve. The decreasing phase is also faster so that the dying out of fire occurs quickly. This allows the flexibility to stop simulation earlier without compromising accuracy. The predefined termination of simulation is when the fire curve drops to 5% of its peak value at 43 rd minute. The total energy release (i.e. area under HRR curve) is not violated by using the modified HRR curve. Differences between maximum gas temperatures are detected on a ceiling just above the fire pool as low as 50 °C, and maximum surface temperature levels are very similar in both cases [26].
Overall, this modification results in 28% reduction of the computational time.  [11]. It is expected that windshield glass will break first followed by side and rear windows. The ground clearance of the passenger car is 20 cm.   [18]. To assess criteria for vertical fire spread; undercover, tires, and front bumper of the car model given in Fig. 8 is altered with combustible polymers. The undercover and bumper are chosen as polypropylene and the tires are selected as rubber. Thermal properties of materials used in the model are given in Table 3.  During the fire simulation, the first ignition occurred at the undercover of car, which is also observed in BRE tests [13]. However, the undercover caught fire at 12 th minute, in contrary of 5 minutes observed in BRE tests. After the ignition of undercover, front tires ignite at 16 th minute. The vertical fire spread is illustrated chronologically in Fig. 9. The result indicates that at approximately 10 kW/m 2 of incident surface heat flux is sufficent to start combustion on undercover. Incident surface heat flux curves for tires are shown in Fig. 10. The incident surface heat flux at ignition time nearly reaches to 20 kW/m 2 on front tires and 10 kW/m 2 on back tires.

Horizontal and vertical fire spread between passenger cars
With the aforementioned analysis, both vertical and lateral fire spread criteria are proposed.
When the incident heat flux level is lower than 8 kW/m 2 , there is no risk of ignition, whereas the level is higher than 16 kW/m 2 , passenger cars start to burn and spread the fire. If the incident heat flux level is between 8 and 16 kW/m 2 , the adiabatic surface temperature [30] should be observed closely. As long as the adiabatic surface temperature is lower than the ignition temperature of the material, the related component cannot catch fire.

Case Studies
In order to provide fire safety to automated tall car park structures, it is imperative to eliminate or slow down the fire spread between passenger cars. In addition, the structural integrity including the steel pallets should be maintained during fire. Fire spread risk levels are shown in Fig. 11. Fire starts at vehicle V13 as shown in Fig. 4a. The neighboring passenger cars are designated as 'R' (right side of the car), 'L' (left side of the car), and 'U' (upper side of the car). Case studies are illustrated in Fig. 12 with passive and active fire safety measures. In all case studies, Modified Category III HRR design curve is utilized. Unprotected car park is the base scenario (Case A). The same structure with partial firewalls and fireproof ceilings is Case B. The structure with fire shutter doors is Case C. Finally, Case D utilizes sprinkler layout by [31] and Case E utilizes an improved sprinkler layout proposal.

Case A: Car park without fire protection
The fire safety level of an open façade unprotected car cark is inspected. Fig. 15 Fig. 15. Passenger cars L0, L1 and L2 are at a relatively far distance from the fire, where the gas temperature remains low. Incident heat flux levels over these cars are under 5 kW/m 2 which is deemed safe as previously stated. Incident heat flux levels for car U1, R0 and R1 are given in Fig. 16. Incident heat flux levels at mid bumpers of upper cars reach over 50 kW/m 2 that causes fire to spread. Fig. 16 also indicates that the fire can spread to all surrounding cars within 25 minutes.

Case B: Car park with firewalls
In order to minimize the fire spread, firewalls between passenger cars and outmost columns are placed throughout the structure. In addition, fire ceilings with 30cm overhang are placed just underneath car pallets. The firewall configuration is seen in Fig. 18. By utilizing firewalls, hot gases are expected to channelize between the firewall and the overhang and exhausted through the façade. The firewalls are 5cm thick with 0.05 W/mK conductivity and 1 kJ/kgK specific heat. It is assumed that the thermal properties are temperature-independent and thereby stay constant throughout the fire. The gas temperatures in Case B show that the firewall and fire ceiling are mostly effective in preventing fire spread. The incident heat flux levels on neighboring passenger cars as given in Fig. 20 are considerably lower compared to Case A. Car R0 and R1 are also totally protected by extended side heat shield as illustrated in Fig. 18. The vehicles on the opposite of the elevator shaft (i.e. L0, L1) are subjected to incident heat flux below 3 kW/m 2 and therefore the fire spread is eliminated to these vehicles. The mid bumper and the front tires of car U1 are exposed to an incident heat flux between 8 kW/m 2 and 16 kW/m 2 . Ignition temperatures for bumpers and tires were previously defined as 388 °C and 350°C, respectively. As seen in Table 4, the temperature levels obtained from the fire simulation are 382°C for mid bumper and 392°C for the front tire at 25 th minute. Thus, the car U1 is assumed to catch fire at front tires. Case B significantly minimizes the fire spread but it cannot prevent it completely.

Case C: Car park with firewalls and shutter doors
In Case C, fire shutter doors are placed between the slots and elevator shaft. In addition, all firewall overhangs are removed as seen in Fig. 21. This design approach aims to convert the fuelcontrolled fire into the ventilation-controlled fire once the fire shutter doors are activated. The activation time or triggering mechanism of the fire shutter is essential. Fire shutters can be triggered not only electronically but also mechanically. If the triggering mechanism fails, and the electric motor is disabled, the fire shutter should be closed manually by security personnel or firefighter. The previous FDS results reveal that the activation time of 15 minutes to close the fire shutter is deemed to be satisfactory. As seen in the gas temperature map Fig. 22, hot gases rising to car U1 are not able to cause an ignition before 15 th minute, i.e. before the fire shutter doors are shut. After the fire shutter door is activated, the combustion reaction rapidly consumes oxygen in the compartment and the fire burns out.
The maximum adiabatic surface temperature on car U1 is lower than 175 °C as seen in Fig. 23a. This

Case D and Case E: Car park with sprinklers
The main purpose of sprinkler water on a fire zone is to create a cooling effect by absorbing heat during phase change from liquid to vapor [32]. FDS is capable of modelling heating up and evaporation of water droplets engulfed by hot gases or over a hot surface. FDS is also adequate to model a reduction in HRR, while water droplets encounter the burning surface with predefined HRR curve.
The main equation that governs the phenomena is given in Eq. 4 [19]. ̇′ ′ ( ) is the predefined heat release rate per unit area in kW/m 2 . The term may be obtained by dividing time dependent HRR to the area of the burning surface. The term k is calculated by Eq. 5, in where ′′ is the local mass of water per unit area in kg/m 2 . a is an empirical constant in m 2 /kg.s. The empirical constant is dependent on the water flux, material properties and global geometric features of the burning substance. Thus, it is strictly case-specific, and there is no study found that defines a coefficient for passenger car fires.
The coefficient a is taken as 0.001 m 2 /kg.s. Same sprinkler nozzle is used in all cases. K factor is chosen as 160 / √ . The activation temperature and operating pressure are 68 °C and 1 atm.
Latitude angles of conical jet stream is defined as 60° and 75° in the sprinkler spray model in FDS. Jet stream velocity is chosen as 5 m/s. Two different sprinkler system layouts are examined. The layouts are given in Fig. 25. The first layout is marked as 'Case D', which is the current sprinkler application in car parks suggested by Australasian Fire and Emergency Service Authorities Council Limited [31].
It contains one sprinkler at corners of each passenger car. The second layout is proposed by [26], which contains four sprinklers per passenger car at corners. The reduction in HRR is directly related to the amount of water penetrating on the fire surface.
The sudden changes in HRR curves are shown in Fig. 27a. Sprinkler water also absorbs an important portion of total convective heat as shown in Fig. 27b. A similar phenomenon on convective heat transfer during a FDS simulation of a water mist spray on a propane burner is observed in another study [32].
The incident heat flux levels on surrounding vehicles for both sprinkler layouts are shown in Fig. 28.
The performance of the AFAC layout is satisfactory (Case D). Fire spread risk is eliminated for car U1 and R0 and, nearly 80% of convective heat is absorbed. However, the incident heat flux on the mid bumper of car R1 is still over 20 kW/m 2 . On the other hand, the proposed sprinkler layout in Case E not only prevents the fire spread but also suppresses it completely. Incident heat flux levels on all surrounding cars remain below 8 kW/m 2 .

Conclusion
In this study, the fire performance of an 8-story automated car park structure is investigated.
The main goal is to find out the characteristics of vertical fire spread between passenger cars and add passive and active fire prevention measures to the structure to minimize the fire spread. Given the HRR of the passenger car, FDS model simulates the ignition times of the components of nearby cars and thereby realistically estimates the fire spread rate in tall car park structures. The following conclusions are drawn: -Without any fire protection on the car park structure, the fire spreads to neighboring cars in 25 minutes, to cars above in 20 minutes. The ignition in the vehicle starts with undercover and tires with incident heat flux levels higher than 16 kW/m 2 . FDS simulations show that car components do not ignite if the incident heat flux levels are below 8 kW/m 2 .
-During a car fire, all columns of the unprotected tall car park structure remain below their critical temperatures. Maximum column temperatures in all cases are under 200 °C. Such temperature levels are not considered as structurally significant for collapse. The beam temperatures on the façade remain below 150°C. The beams next to the elevator shaft reach critical temperatures as high as 800°C if left unprotected.
-Steel car pallets are not robust against a passenger car fire. The member temperature of car pallet beams reaches to 1000°C at around 20 th minute of fire. Plastic hinge mechanism will likely form in the load-bearing beams underneath the car pallet in the very early phase of fire. This means that a car pallet just above a fire may collapse before the fire spreads vertically to a car on the car pallet.
To prevent such collapse, fire protection is necessary underneath each car pallet.
-The proposed sprinkler layout is more efficient. When the sprinkler heads are placed at the backside, frontside and sides of the cars as opposed to placing at the corners of the cars. The most effective way to suppress passenger car fire spread in a car park structure without a sprinkler system is to deprive the fire from oxygen by the use of fire shutter doors.