1 Introduction

Purpose-built structures (safe rooms/shelters) have been widely used as last resort options in many countries to safeguard lives and valuables from natural disasters such as tornadoes, wildfires, and storms, and human-induced hazards, such as wars, etc. [1,2,3,4,5,6]. However, the construction of safe rooms for bushfires/wildfires can be more challenging than other disasters due to their (1) nature and characteristics such as propagation and fire path, and (2) associated complexities such as flames, smoke, toxic emissions, embers and radiant heat [7,8,9]. Such associated complexities with bushfires lead to additional requirements/checklists for construction, especially to reduce the chance of ember collection on surfaces, smoke penetration, etc. Furthermore, human behaviour also impacts the safety of lives (during sheltering) and properties. However, it is hard to accommodate this during construction since human behavioural patterns are unpredictable, and people are unlikely to act rationally during bushfires [10, 11]. These uncertainties and requirements have made sheltering in safe rooms during bushfires a controversial topic.

Early evacuation is favoured in bushfire survival plans over safe-sheltering in most countries [12, 13]. However, bushfire safe rooms can play a vital role, especially, when early evacuation is not possible or too risky due to unpredictability of the fire path and extreme environmental conditions like dry wind and low humidity which can speed up the fire propagation. Furthermore, alternatively, safe rooms enable people to safeguard the valuables and memorabilia, as carrying such items during evacuation can slow down the evacuation process and can ultimately result in life losses [14]. The limitations and inadequacy of past studies can be identified as the main reasons behind the controversies, since the performance of bushfire safe rooms has not been well investigated (due to higher time and cost, and the need for specialised equipment). Thus, there exists ambiguity regarding the safety of such structures [10, 11, 14, 15].

A survey showed that the majority of bushfire fatalities during Black Saturday fires (in 2009) in Australia happened during sheltering in houses [1]. Furthermore, it further stated that people had to move from one shelter to another multiple times due to the destruction of the previous shelters, raising questions on the structural adequacy and integrity of the shelters they used [2]. Recent studies on bushfires in Canada and Portugal showed that people often under-estimate the bushfire risk resulting in life losses highlighting the impact of human behaviour on the sheltering process [16]. Nevertheless, studies based in the United States of America (USA) questioned the practicality of conducting mass evacuation every time there is a bushfire [1, 17, 18], and many complexities in this regard have been reported in Australia as well [19]. Therefore, for either of the above-mentioned reasons, (i.e. inadequate protection provided by the houses, inaccurate perception of fire severity and inability to evacuate), safe rooms can be identified as a last resort option for survival or to protect memorabilia during bushfires. Recent post-bushfire surveys have reported of people surviving in safe rooms (either self-constructed or commercially available) worldwide [20]. However, a detailed analysis cannot be conducted due to the lack of information provided in such literature. Thus, the literature review of this study has been limited to the scientific research studies on safe rooms.

Nguyen et al. [15] conducted experimental and numerical studies on a prefabricated bushfire bunker built with aerated concrete panels. This safe room was a steel-framed one-compartment building with two layers of 78 mm thick concrete-filled steel hollow section panels fixed on each side of the frame without cavity insulation. The temperature readings before and after exposing the safe room to the standard fire curve for 30 min showed that the internal room temperature increased only by 5°C. However, this study was conducted only for 30 min, thus the temperature profiles are available only up to 30 min. Due to the low thermal conductivity of the wall configuration with an air cavity and aerated concrete, the temperature increment in internal layers is delayed. Hence, the Australian standard (AS 1530.8.2) [21] recommends observing the temperatures for a further 60 min for a possible temperature increment, especially ambient side temperature peaks after the fire termination (for buildings in bushfire flame zone). Furthermore, it is required to have a minimum of 60 min fire protection for bushfire shelters. Hendawitharana et al. [14] conducted an experimental study on the bushfire performance of a three-compartment Light-gauge Steel Framed safe room externally lined with Autoclaved Aerated Concrete (AAC) panels under simulated realistic bushfire conditions. It showed that the internal air temperature increment was less than 1°C after 67 min fire exposure [14]. Furthermore, they used two different realistic fire curves, (1) simulating an approaching bushfire with pre-bushfire radiant heat, bushfire flame zone and post-bushfire radiant heat and (2) simulating the flame zone and radiant heat from a nearby building fire (at a distance of 10 m from the safe room). Two opposite sides of the safe room were exposed to the two fire curves. The temperature data across and along the walls and roof were obtained using more than 128 thermocouples.

However, due to the high monetary cost and the time associated with full-scale experiments, extensive experiments for different safe room configurations are difficult to conduct highlighting the importance of numerical modelling to assist with detailed studies. Therefore, this research focuses on developing full-scale numerical models of bushfire safe rooms and studying the impacts of related parameters on their thermal performance. The literature review showed that Computational Fluid Dynamics (CFD) modelling tools such as Fire Dynamics Simulator (FDS), Ansys fluent and FireFOAM have been used in numerous studies to simulate fires [15, 22,23,24,25,26,27,28,29]. All these modelling tools follow similar theoretical concepts in underlying mathematical calculations [30, 31]. Therefore, in this study, we used FDS, specialised for fire modelling applications.

Considering the availability of data and the extended duration of temperature records, the full-scale test of the three-compartment safe room conducted by the authors [14] was selected to develop and validate the numerical models. The validated models were then used to study the influence of the internal compartments, area of fire exposure on the safe room, location of the safe room in bushfire-prone areas and the impact of different environmental conditions. The results of this study enhance the understanding of the performance of above-ground bushfire safe rooms detached from the dwelling under realistic fire conditions. They showed that the safe rooms can be constructed using available materials and provided conditions of usage depending on environmental factors. The modelling approach emphasises the applicability of such methods in evaluating the safe rooms and modular buildings in bushfire or other fire related applications.

2 Experimental Set-Up

The safe room was 3 m (width) × 4 m (length) × 2.4 m (height) in size and had three internal compartments as shown in Fig. 1. The main entrance to the safe room was through Door 1, which opens to Compartment 1. The entrant then needs to close Door 1 before opening Door 2 (to reduce the probability of smoke penetration into the compartment) which leads to Compartment 2. A viewpoint will be installed in this compartment to observe the outside fire conditions before leaving the property after the bushfire passes. Compartment 3 is the main compartment where the occupants are supposed to stay and shelter until the bushfire passes. Having multiple compartments in the safe room limits smoke and toxic gas entrance into Compartment 3, where the people shelter during the bushfire event. Having three separate compartments over one reduces the risk of smoke and toxic gas penetration to the main compartment in the unlikely event of the main door failure.

Figure 1
figure 1

(a) Internal compartments and (b) plan view of the safe room

Full size image

2.1 Materials and Building Components

The safe room was constructed with light steel framed walls lined with AAC panels and was cuboid in shape. The structure was mounted on concrete beams. The self-weight of the safe room was borne by the steel frame which was then transferred to the ground through concrete beams. The configurations of the building components are shown in Fig. 2 [14].

Figure 2
figure 2

Building components (a) external wall, (b) internal wall and (c) roof configurations [14]

Full size image

The light steel framed walls consisted of 90 × 39 × 9 × 1.0 mm lipped channel studs, placed on tracks. Noggings were placed at the mid-height of the studs located at 550 mm spacing. The external walls consisted of 75 mm thick AAC panels fixed to the external side of the steel frame using 25 mm battens. Two layers of 16 mm fire-rated gypsum plasterboards were fixed to the ambient side (internal side) of the steel frame. The stud cavities were insulated using R2.0, 90 mm thick glass fibre insulation. Water, air and vapour barrier sarking was used between the studs and battens to limit the moisture movements into the building. Figure 2a shows the configuration of the external walls. The internal walls consisted of the same stud arrangement but were lined with one layer of 16 mm fire rated gypsum plasterboards on each side. No cavity insulation was used for internal walls (see Fig. 2b).

The roof of the safe room consisted of \(90 \times 39 \times 9 \times 1.0\) mm lipped channel sections arranged back-to-back. Two layers of 75 mm AAC panels were fixed to the external side of the steel frame and one layer of non-fire rated 10 mm thick gypsum plasterboards was fixed to the internal side (Fig. 2c).

The safe room had three 2-h fire rated 37 mm thick doors as shown in Fig. 1. The external side of the fire doors had a 1 mm steel sheet while both sides of the 29 mm insulated core had 3.6 mm thick plywood layers. Figure 3 shows the completed safe room with the labels for the four sides and this notation will be used hereafter.

Figure 3
figure 3

Four sides of the safe room [14]

Full size image

2.2 Burner Arrangement and Test Set-Up

Figure 4 shows the safe room located in an outdoor setting with two gas-fired burners located on Sides A and C of the safe room. The burners were manually controlled based on the readings obtained from the heat flux meters located on the external surface of the walls (Sides A and C). Thermocouples were placed along the wall and roof surfaces and across the thickness to obtain the temperature distributions. Air temperatures inside the compartments were obtained using handheld devices and K-type thermocouples located at different heights.

Figure 4
figure 4

Burner arrangement and the fire curves simulated using each burner (a) Fire Curve 1 (Burner 1) and (b) Fire Curve 2 (Burner 2)

Full size image

2.3 Fire Test Details

The test scenario considered in the experiment was a safe room located by the forest facing Side A. Therefore, Side A of the safe room was exposed to a fire curve (Fire Curve 1) representing a bushfire which approaches and passes by and Side C (the opposite side of Side A) was exposed to the radiant heat emitted by a nearby building fire caused by a bushfire (Fire Curve 2). The fire curves representing these two cases are given in Fig. 4. The total test duration was 67 min [the fire exposure on Side C had additional 5 min exposure during the experiment than the expected fire curve (Fire Curve 2)]. At t = 0, the fire from Side A started and the radiant heat gradually increased to 40 kW/m2 within 30 min. Then flame immersion phase started, where the radiant heat flux reached beyond 100 kW/m2. After 2 min of flame exposure, the radiant heat dropped and the low radiant heat for the given duration represented the radiation from the burning of nearby fuel. The flame from Side C started in parallel with the flame immersion in Side A. After that, a constant heat of 10.6 kW/m2 was simulated until the test termination at 67 min to represent radiation from a nearby building fire at a 10 m distance. The experiment was undertaken in an outdoor setting, and thus the measured heat flux curves and fire side temperature profiles were affected by outdoor wind conditions and manual control of heat flux [14].

3 FDS Models

In this study, FDS was used as the computational modelling tool due to its capabilities in simulating fire conditions in both small and large scales. The main consideration during the modelling procedure was to ensure realistic simulation of heat transfer through surfaces. This is challenging due to the difference in the scale of the building components and the overall building. The thicknesses of the material layers inside the building components were in millimetre scale while the overall safe room dimensions were in meter scale. Therefore, small-scale wall models of the two fire side walls (Sides A and C) were developed first and the models were then extended to the full-scale safe room of dimensions 3 m × 4 m × 2.4 m. The internal layers of the walls were simplified to be conservative and efficient.

3.1 Thermal Properties of Materials

Temperature-dependent thermal conductivity and specific heat values of the materials that were used in this numerical study (i.e. AAC [32], fire-rated gypsum plasterboard [33], non-fire rated gypsum plasterboard, steel [34], plywood [35], glass fibre insulation [36]) are shown in Fig. 5.

Figure 5
figure 5

Specific heat and thermal conductivity values of (a) fire-rated gypsum plasterboard, (b) non-fire-rated gypsum plasterboard, (c) cold-formed steel, (d) auto-claved aerated concrete (AAC), (e) glass fibre insulation and (f) plywood used in the numerical models

Full size image

3.2 FDS Heat Transfer Model—Walls

In the first phase of the model development, two small-scale wall models were developed to simulate the bushfire heat transfer through Sides A and C walls. The entire wall frame dimensions of Sides A and C walls of the safe room were 3 m (width) × 2.4 m (height) with studs used at 550 mm spacing. However, the sizes of the lipped channel sections used in the experiment were \(90 \times 39 \times 9 \times 1.0\) mm, which are in millimetre scale compared to the wall frame which are in metre scale resulting in a trade-off between the model mesh size and the computation complexity. If the mesh size of the model is set to the millimetre range to consider the smallest possible element (without simplification), it will require an excessive amount of computational resources and will take significantly higher time. In contrast, if a larger mesh size is selected for numerical modelling, the model run-time will be reduced, but the original geometry will not be able to be replicated.

A previous study on the numerical modelling of external wall systems using FDS showed that the reduction of height up to 0.2 m did not affect the mid-height temperatures of the walls across the thickness [22]. Therefore, the model height was selected as 0.2 m. The study further emphasised the importance of fixing the model width to the stud spacing. Therefore, the model width used in this study was 550 mm (0.55 m). The mesh sensitivity study showed that the mesh size of 0.01 m × 0.01 m × 0.01 m provided a good geometric representation with acceptable accuracy at an affordable computational cost. Therefore, a model of dimensions 0.2 m (height) × 0.55 m (width) was used with a 0.01 m mesh cell size. The model was divided into 6 meshes which ran in parallel using 6 CPU cores (each mesh was assigned to a separate CPU core). In the model, each material was given as a separate one-cell thick obstruction with the actual material thickness given under surface properties. The backing condition of all the obstructions was set to “EXPOSED” to simulate heat transfer through each layer. A 25 mm gap was present between the AAC panel and the steel frame representing the batten cavity. Steel stud was placed at the centre of the model and the cavity insulation was modelled as multiple layers of obstructions. Figure 6 shows the developed FDS heat transfer model of the wall.

Figure 6
figure 6

FDS model of individual walls

Full size image

The fire was introduced as a heater surface. The top and bottom boundaries of the model were “ADIABATIC” and the sides of the model were considered to be “MIRROR” to simulate the continuity of the cavity. The initial air temperature was 25°C (the average initial internal temperature of the safe room during the experiment). The average surface temperatures of each material at the start of the test were given as the initial temperature. The surface temperatures of each layer of material (AAC, steel studs and plasterboards) were measured at mid-height and compared with the corresponding experimental results.

3.3 FDS Heat Transfer Model—Full-Scale Safe Room

After validating the small-scale wall models for the surface temperatures across the thickness of the individual wall, the study was extended to understand the effect of replacing the detailed small-scale model with a one-cell thick layered obstruction. With the huge difference in the scales of the internal wall layers and the overall safe room size, it is required to simplify the wall representation in the full-scale model. The comparison of the detailed and one-cell thick small-scale models of walls under the standard fire exposure for 4 h showed a good agreement and the one-cell thick model predictions were conservative. In the one-cell thick wall representation, the air cavity between the AAC panel and the stud was omitted. Therefore, the temperatures in the internal layers were slightly higher compared to the detailed model. Furthermore, considering the small area occupied by the studs and as the stud cavities were filled with cavity insulation, the steel studs were not modelled. However, if the stud cavity is not cavity insulated, the presence of steel studs will have a noticeable impact and need to be included in the model. Furthermore, any detailed observations such as the effect of thermal bridging in the presence of studs need to be studied using small-scale detailed models. The four material layers (1 mm steel, 3.6 mm plywood, 29 mm insulation and 3.6 mm plywood) in the fire door (in that order) were simulated as layered surfaces with appropriate thermal conductivity values to simulate the heat transfer through the fire door.

Figure 7
figure 7

FDS model of the full-scale safe room

Full size image

All the elements in the full-scale safe room were modelled as one cell thick elements (Fig. 7). Layered surfaces were used to include the material layers in the thermal conductivity calculation. No gap was introduced between the door and the external wall. The two fire curves were introduced to the model as heater surfaces (time-dependent heat flux) (Fig. 8). The flame zone bushfire curve (Fire Curve 1) was given on Side A boundary and on Side C, the radiation from the building fire (Fire Curve 2) was simulated (Fig. 4). Since the measured heat flux curves showed fluctuations due to the outdoor setting, the originally expected heat curves were used. The other two sides (Sides B and D) and top boundaries were kept open. The final arrangement of the safe room model is shown in Fig. 7.

It was identified that radiation is the main mode of heat transfer for this case and the default convective heat transfer model in FDS was used. The mesh cell size used in the study was 0.1 m. The computational boundary was divided into five meshes which ran in parallel to each other. High-performance computing facilities were used to run multiple models in parallel at a limited time.

4 Model Validation

In this section the results predicted by the small-scale and full-scale heat transfer models are compared with the corresponding experimental results.

Figure 8
figure 8

Difference in heat exposure in the experiment and the FDS models

Full size image

4.1 Individual Walls

Figure 9a and b shows the time–temperature curves across the thickness of Wall A. As the temperature distribution across the wall surface was not uniform, the model predictions were compared with thermocouples in one location across the thickness and the results showed a good agreement. Figure 10a and b shows the agreement of the model predictions with experimental results at the centre of Wall C. The model predictions show a reasonable agreement with the temperatures across the thickness. The main reason for the model predictions to differ slightly from the experimental values is that the area of fire exposure was not uniform across the wall surface. The fire source for this study was a gas-fired burner with two flame outlets which was controlled based on the heat flux recorded at the centre. However, the thermocouple readings showed that the temperatures across the wall varied with the difference between maximum and minimum fire cavity temperatures being 21°C for Wall A and 13°C for Wall C (refer blue lines in Figs. 9a and 10a).

Figure 9
figure 9

Experimental versus model predicted time–temperature curves for Wall A (a) AAC and plasterboard surfaces (FS Fire Side, FC Fire Cavity, AC Ambient Cavity and AS Ambient Side), and (b) steel stud hot flange (HF) and cold flange (CF) (Color figure online)

Full size image
Figure 10
figure 10

Experimental versus model predicted time–temperature curves for Wall C (a) AAC and plasterboard surfaces (FS Fire Side, FC Fire Cavity, AC Ambient Cavity and AS Ambient Side) and (b) steel stud hot flange (HF) and cold flange (CF) (Color figure online)

Full size image

In the numerical model, smaller model dimensions were selected and equal temperature distribution was assumed on the fire side (see Fig. 8). This made the model predictions comparatively higher than the experimental results (i.e difference between the peak fire cavity temperature of the experiment and models was 14°C for Wall A and 30°C for Wall C). Other causes for the deviations include the differences in the moisture contents of the AAC panels used in the safe room and the panels that were considered when proposing the apparent thermal conductivity values. AAC panels are lightweight and highly moisture-absorbent. The absorbed moisture acts positively in reducing the temperature increment on the ambient side. Furthermore, the apparent thermal conductivity values of other materials of the wall system (glass fibre insulation and gypsum plasterboards) have some influence on the temperature predictions. However, when observing the model predictions, the ambient side temperature which is the deciding factor for the performance of the wall was well-predicted by the models (i.e. the peak differences being 3°C for Wall A and 2°C for Wall C) (refer the green lines in Figs. 9a and 10a).

4.2 Full-Scale Safe Room

The external and internal surface temperatures, the air temperatures in the compartments and the heat flux on Walls A and C of the safe room were compared with the predictions of the FDS model of the full-scale safe room. Figure 11b shows the incident radiant heat flux on Walls A and C which shows a reasonable agreement with the experimental results. The experimental heat flux values showed fluctuations due to the outdoor wind conditions and the entire wall was not exposed to the same heat flux level as the middle part, where the heat flux sensor was located. In contrast, the heat flux in the model was introduced as a radiating surface with no wind, and the heat transfer model exposed the safe room to a more uniformly distributed heat flux. This exposed the fire side walls to comparatively higher heat exposure and is a conservative approach. Figure 11b shows a comparison of experimental and numerical model heat flux values.

Figure 11
figure 11

Experimental versus model predicted curves for (a) internal air temperature and (b) radiant heat flux on Sides A and C

Full size image
Figure 12
figure 12

Experimental versus model predicted time–temperature curves for external (FS) and internal (AS) surfaces of (a) Wall A and (b) Wall C

Full size image

Furthermore, outdoor conditions caused non-uniform temperature distribution on the fire sides of the safe room. A comparison of the fire side and ambient side wall temperatures of Walls A and C of the safe room are shown in Fig. 12a and b. The results show a reasonable agreement with the experimental values. The model air temperatures agreed well with experimental temperatures until the test termination (Fig. 11a). The internal surface temperature of the model continued to rise and reached a peak at 39°C (Fig. 12a) and 35°C (Fig. 12b) on Walls A and C, respectively. However, the maximum temperature recorded in the experiment was 29°C for both Walls A and C. This difference is caused by the rapid reduction of fire side temperatures caused by forced convection during the experiment whereas in the model, no forced convection was used and thus, a comparatively slow temperature decrease was observed. Furthermore, the internal air temperatures inside Compartment 3 recorded a maximum value of 27°C in the model and the experiment after 67 min of fire exposure (Fig. 11a). After the termination of the fire, the temperatures continuously increased to 33°C whereas in the experiment no temperature rise was observed. The cause for this 5°C temperature deviation is due to the more evenly distributed fire exposure in the model compared to the experiment. However, these values are well below the limiting internal surface temperature and air temperature values for safe rooms, which are 70°C and 45°C, respectively.

Moreover, it is expected that a 1 h protection is adequate for a severe bushfire exposure and the safe room provided similar thermal performance as the experiment for 67 min and adequate performance until the model termination which was at 2 h. Therefore, it is concluded that the safe room model provided a good agreement with the full-scale safe room experiment and the model can be used to evaluate other scenarios related to bushfires.

5 Parametric Study

The validated safe room model was then modified for use in a parametric study. The scenarios considered in this parametric study are illustrated in Table 1, where all four sides of the safe room were exposed to either of the two fire curves shown in Fig. 4. One of the main concerns regarding the bushfire safe rooms is maintaining the internal surface and air temperatures below appropriate limits. The internal arrangement of the safe rooms has rarely been evaluated even though it acts as an important consideration for human survival. Furthermore, for above-ground safe rooms, maintaining the internal air temperature is a critical factor due to the hot summer temperatures. The internal air temperatures should be maintained at a rate where no serious health impacts are induced. Increasing the internal temperatures beyond the tolerable limits leads to disastrous situations making the safe room a death trap. Therefore, the main causes for internal temperature rise should be identified and evaluated. This parametric study is focused on addressing the above-discussed issues using the selected models listed next.

SR1: SR1 model was similar to the validated model but with different fire exposure conditions. To understand the effect of fire exposed area, Sides A and C of the safe room were exposed to Fire Curves 1 and 2 (Fig. 4), respectively (similar to the full-scale safe room experiment). Fire Curve 1 represents an approaching bushfire including flame immersion for 32 min, whereas Fire Curve 2 represents the bushfire flame immersion followed by a constant radiant heat of 10.6 kW/m2 for 30 min to simulate the radiant heat from a nearby building fire (total 62 min). In Model SR1 (refer Table 1 and Fig. 13), Side A was exposed to Fire Curve 1 and all other sides were exposed to Fire Curve 2.

SR2: When the fire is approaching, the sides of the safe room (Sides B and D) will also be exposed to the increasing radiant heat. During the full-scale experiment, the external surface temperatures at a distance of one fourth of the safe room length from the bushfire exposed side also showed higher temperature values even when using one burner located at the centre. In reality, bushfires are line fires and in that case, the exposure of the sides is inevitable. Therefore, Fire Curve 1 was introduced up to one third of the length (to be conservative) on Sides B and D from Side A surface. Rest of the Sides B and D and Side C were exposed to Fire Curve 2 (Model SR2). Table 1 shows the illustrations of these scenarios.

Table 1 Models Used in the Parametric Study
Full size table
Figure 13
figure 13

The illustrations for (a) the internal compartments and (b) the sides of safe room used in Table 1

Full size image

SR3 and SR4: Historical summer temperature data show that in the event of extreme cases, the environmental temperatures can reach above 40°C during heat waves in bushfire-prone countries [37, 38], whereas the bushfires have occurred with different temperatures at different times of the year [39,40,41,42]. It is crucial to take this factor into consideration for safe rooms. It is a challenging task to maintain the internal air temperature below 45°C for above-ground safe rooms for 1 h when the initial environmental temperatures are already high. In this study, the time–temperature variations of the internal compartment (Compartment 3) of the safe room were predicted for three different initial environmental temperatures (20°C, 30°C and 40°C in SR2, SR3 and SR4, respectively).

SR5, SR6, SR7 and SR8: The internal arrangement of the safe room in the experiment consisted of three compartments (i.e. two small compartments to act as a smoke barrier and one large compartment for people to shelter during the fires). In this numerical study, the effect of having internal compartments on the heat transfer and internal air temperatures was evaluated. The temperature predictions from models SR5, SR6, SR7 and SR8 together with SR2 and SR4 were used to understand this effect.

SR9: A comparative study was also conducted for different maximum fire intensities, i.e., in SR9, the bushfire exposure representing a BAL-40 bushfire attack level was investigated and the results were compared with the flame zone predictions.

6 Results and Discussion

The results obtained from the parametric study are then evaluated in this section. A summary of the temperatures of each compartment at the end of the fire (i.e., 62 min) and the maximum temperatures for 2 h duration in each case are given in Table 2. The thermocouples used in this comparison were located at 200 mm height from the ceiling.

Table 2 Predicted Internal Air Temperatures for the Considered Cases
Full size table

6.1 Effect of Fire Exposed Area

As shown in Table 1, the exposed area of the two fire curves was changed from SR1 to SR2. Table 2 shows the internal air temperature distributions of Compartments 1, 2 and 3 (C1, C2 and C3) when the fire exposed area was changed. It shows that both cases showed similar internal conditions, thus, showing negligible influence due to the change of fire exposed area.

6.2 Effect of Environmental Temperatures

Figure 14a shows the temperatures of Compartment 3 for three different initial environmental conditions [20°C (SR2), 30°C (SR3) and 40°C (SR4)]. It shows that when the environmental temperature was 20°C or 30°C, Compartment 3 temperature was below 45°C for the 2 h duration whereas, the time to reach 45°C was 83 min when the initial temperature was 40°C. Therefore, it shows that Compartment 3 remains safe throughout the fire duration of 62 min and thus could provide tenable conditions with respect to heat transfer during extreme bushfire exposure conditions. Furthermore, as shown in Fig. 14b, the rate of temperature increment was reduced when the initial environmental temperature was high (40°C). However, this difference was negligible for the considered time duration. Therefore, it shows that the initial environmental temperatures have a negligible impact on the rate of temperature increment inside the safe room.

Figure 14
figure 14

(a) The internal air temperatures in Compartment 3 for the initial environmental temperatures of 20°C, 30°C and 40°C and (b) the temperature increment

Full size image

Figure 15 shows the internal air temperatures at 62 min and the maximum temperature after 120 min for the initial environmental temperatures of 20°C, 30°C and 40°C. For ease of visualisation, the temperature scale was divided into 5°C intervals and different colours were given. Each of these divisions was given a number (rating) ranging from 0 to 9, i.e., if the number is 2, the temperatures are within the range of 25°C to 30°C. It should be noted that if the number given is less than or equal to 5, i.e., less than 45°C it is safe for human survival when heat transfer is considered. It showed that Compartments 2 and 3 are safe until the fire ends for all three initial environmental temperatures. However, the temperature of Compartment 1 showed significantly high values even during the 62 min test period when the environmental temperatures are above 30°C.

Figure 15
figure 15

The internal air temperature distributions across the 3-Compartment safe room when the initial environmental temperatures were 20°C, 30°C and 40°C at the end of (a) 62 min and (b) 120 min (Color figure online)

Full size image

The investigation of the causes for this temperature distribution showed that the 2-h fire-rated commercially available fire door was the main contributor to heat transfer into the safe room. Compartment 1 temperatures showed a rapid increment when compared to Compartment 2 temperatures when exposed to the same bushfire curve. Both Compartments 1 and 2 are the same in dimensions and were exposed to the same fire curve, but showed a significant difference in air temperatures. The air temperatures showed high values close to Door 1. Therefore, it was identified that the external door is the main contributor to controlling the internal temperatures in the safe room.

The requirements of fire resistance for building fires and bushfire shelters are significantly different. The FRL (Fire Resistance Level) values given for building components are not sufficient in specialised applications like bushfire shelters where the maximum acceptable surface temperatures (70°C) are significantly lower. Therefore, even though the doors do not fail (based on structural adequacy, integrity or insulation criteria for building fire resistance, they can reach the maximum acceptable temperatures and emit radiation into the safe room, as shown in this study. This will increase the air temperature in compact, airtight spaces in safe rooms leading to unsuitable air temperatures for human sheltering. Therefore, the commercially available fire doors that are designed for the building fire requirements may not always be suitable for bushfire safe room construction, and further studies are required on the bushfire resistance of the doors for use in bushfire safe rooms.

6.3 Heat Transfer Through the External Building Envelope

The above observations were further confirmed when the ambient/internal surface temperatures of the external envelope were analysed. The internal surface temperature contributes directly to the internal air temperature. This numerical study showed that in all the cases, the internal surface temperatures of the walls remained below 60°C, which is acceptable. Figure 16a shows the internal surface temperature profiles from the SR4 model (3 compartments at 40°C initial environmental temperature), which is the critical case. As shown in Fig. 16a and b, the surface temperature of Door 1 showed temperatures above 70°C after 44 min, which directly contributed to higher internal air temperatures in Compartment 1. Therefore, it shows that even though external walls and roof comprise of the largest surface area, the doors play a vital role in ensuring tenable conditions in the safe room. Furthermore, the building element models are not sufficient to determine the performance of the safe rooms, and full-scale models are required.

Figure 16
figure 16

(a) Time–temperature curves for the internal surface temperatures and (b) the internal surface temperature distribution from SR4 model at the end of 62 min

Full size image

6.4 Effect of Internal Compartment Arrangement

The main reason for having three compartments in the experiment was to avoid smoke penetration in to the main compartment (Compartment 3). However, it shows that this was favourable to control the internal temperatures as well. As shown in Fig. 15, the higher temperature increment is limited to Compartment 1 and both Compartments 2 and 3 remained safe during the fires. Therefore, having multiple compartments acted favourably in limiting the internal air temperatures in Compartment 3. However, this confines air into small segments which leads to higher temperatures within them. Therefore, the internal compartment arrangement was changed from three compartments to two compartments (models SR7 and SR8) and one compartment (models SR5 and SR6) and the results are compared with those from three compartment models SR2 and SR4. The overall sizes of the safe room remained the same as shown in Table 1.

When the safe room consisted of one compartment, the internal temperatures started to increase after 40 min [with an increment of 7.5°C at 62 min)]. The inclusion of two or three compartments delayed the start of increasing air temperatures in the internal compartment (Compartments 2 and 3) until 60 min at both 20° and 40°C initial environmental temperatures. However, Compartment 1 temperature in the multiple-compartment safe rooms started to increase after 40 min, similar to one compartment safe room but at a higher rate (see Fig. 17). This is due to the comparatively small volume of air.

A comparison of Compartment 3 temperatures showed that internal air temperatures were the same when the safe room consisted of either two or three compartments. This is further elaborated in Fig. 17a and b. Therefore, with respect to heat transfer, it was identified that two compartments were adequate to reduce the internal temperatures in the main compartment.

Figure 17
figure 17

Internal air temperatures in Compartments 1, 2 and 3 (C1, C2 and C3) when the number of compartments was reduced from 3 to 2 and 1 at the initial environmental temperatures of (a) 20°C and (b) 40°C

Full size image

The reason for the temperatures in Compartment 3 to be higher compared to Compartment 2 is due to the higher duration of heat exposure from Side A compared to Side C. Models using standard fire curve from both sides (Sides A and C) showed similar temperatures in Compartments 2 and 3. Therefore, with respect to heat transfer in the bushfire safe rooms, multiple compartments are recommended than a single-compartment to account for the weaker element failure. However, the size of the first compartment (first entering compartment) must be derived after a thermal analysis as the internal air volume affects the temperature increment. Furthermore, having more than two compartments did not show any improvement to the heat transfer performance.

6.5 Effect of the Location of the Safe Room

The risks of using above-ground safe rooms will also depend on its location in the bushfire-prone areas. Hence further analyses were conducted to evaluate the effects of the location of the safe room by considering both flame zone and BAL-40 exposure conditions. For BAL-40 conditions, the maximum heat flux from the two fire curves was capped at 40 kW/m2. The most critical model (three-compartment safe room at 40°C environmental temperature) was exposed to the fire curves with flame zone intensity of 40 kW/m2 (SR9 Model). This is to simulate the realistic bushfire exposure conditions in the area adjacent to the bushfire flame zone (BAL-40).

Figure 18 compares the results of internal air temperatures from the flame zone (SR4) and the BAL-40 (SR9) models at 62 min. The results show that all the compartments exhibited similar performance with respect to internal air temperatures in Compartments 2 and 3. A maximum 5°C difference in air temperature was observed in Compartment 1 when exposed to the two bushfire exposure conditions. Therefore, it is emphasised that having one weak element in the safe room significantly impacts the overall performance and the risk is still present even if the safe room is not situated in the flame zone.

However, the model exposure conditions were extreme when compared to the realistic bushfire exposure conditions. In reality, bushfires do not expose all four sides of the safe room at similar intensity. Therefore, the door should be placed in a direction away from direct fire exposure or any potential fuel sources. Furthermore, it is recommended that further detailed studies are required on the performance of the doors to be used under bushfire conditions, and additional measurements are required to protect the doors.

Figure 18
figure 18

Internal air temperatures for BAL-40 and BAL-FZ (flame zone) exposures at the end of (a) 62 min and (b) 120 min

Full size image

Overall, this study has shown that construction of the above-ground safe rooms is feasible with respect to heat transfer, provided that the doors are placed in a non-fire exposed direction and improved fire-resistant doors are used. Furthermore, the use of multiple internal compartments reduced the adverse effect of weak door elements on the inner compartments. However, the area of the compartments has to be determined based on the potential temperature increment in the confined air volume.

6.6 Limitations

Since this study is limited to the heat transfer aspects of safe rooms, it is recommended that further detailed studies are required in the following areas: human behaviour, smoke and toxic gas penetration and behaviour of different building materials in the above-ground detached safe rooms.

7 Conclusions

This paper has presented a numerical modelling study undertaken on bushfire safe rooms subjected to realistic bushfire exposure conditions. FDS was the tool used in the numerical study. The results of a full-scale experimental study conducted on a steel framed three-compartment bushfire safe room were used in the validation process. In the first part of the study, small-scale models were developed for the fire-exposed external walls. The time–temperature predictions of the models across the wall thickness (external side, fire side cavity, ambient side cavity, ambient side, steel stud hot flange and cold flange) were compared with the experimental results and they showed a good agreement. Then a model of the full-scale safe room of dimensions 4 m × 3 m × 2.4 m was built and was exposed to bushfire conditions. The external surface heat flux, external and internal surface temperatures and internal air temperatures in the safe room were compared with the experimental results and they agreed reasonably well. The developed models were used to further investigate the complexities associated with safe rooms. The findings of this study are as follows.

  • Safe rooms that satisfy the heat transfer requirements during bushfires can be constructed using available building materials. The internal wall surface temperatures were less than 60°C for a duration of 2 h when the outside initial environmental temperature varied from 20°C to 40°C.

  • The external door is the weakest element in the safe room and the commercially available 2-h building fire-rated doors may not be always suitable for safe rooms. Thus, further studies of fire doors are needed before using them in safe room applications.

  • Bushfires occur at different environmental temperatures and can happen at any time of the year given the other factors contributing to them are present. However, the initial environmental temperature has a significant impact on the ability to shelter in the above-ground safe rooms. The main compartment of the safe room showed only a 2°C increment during bushfire exposure. However, in the regions where the summer temperatures reach beyond 45°C or during heat waves, even very small temperature increments inside the safe room during bushfire exposure can cause negative impacts on human survival due to high internal air temperature. Therefore, under extreme environmental temperature conditions, above-ground safe rooms are not recommended for sheltering purposes. However, they can be used as safe storage rooms during bushfires.

  • Single compartment safe rooms are highly vulnerable especially when heat transfer through the fire door is significant (maximum internal air temperature showed a 7.5°C increment by the end of fire exposure). Having multiple compartments in the safe room acted favourably to regulate the internal air temperature in the main compartment of the safe room (only a 2°C increment). Furthermore, having more than two compartments did not show any reduction to the internal temperatures compared to the two-compartment safe room.

  • The basic rule is to select the area of the safe room based on the number of potential occupants (to retain an adequate amount of air to breathe). However, due to the confined space with no openings to the outside, the internal volume has a significant effect on the internal temperatures, thus it should be considered in determining the size of the safe room.

  • Exposures based on both flame zone and BAL-40 conditions showed similar performance in terms of the main compartment air temperatures.

  • In real-world applications, safe rooms must be located such that the door is not exposed to radiation from a nearby fuel source. Creating a fuel-free zone around the safe room is vital.