- Email: [email protected]
Fire whirls in forest fires: An experimental analysis
Fire Safety Journal 87 (2017) 37–48
Contents lists available at ScienceDirect
Fire Safety Journal journal homepage: www.elsevier.com/locate/firesaf
Fire whirls in forest fires: An experimental analysis ⁎
Cláudia Pinto , Domingos Viegas, Miguel Almeida, Jorge Raposo
crossmark
Association for the Development of Industrial Aerodynamics, University of Coimbra, Rua Pedro Hispano 12, 3030-289 Coimbra, Portugal
A R T I C L E I N F O
A BS T RAC T
Keywords: Fire whirl Fire tornado Forest fires Wildfire Extreme fire behaviour Fire safety Forest fuels
This work presents a study on the formation of fire whirls with vertical axis on wildfires at laboratory scale. A particularity of the study is the use of typical forest fuels instead of fossil fuels as seen in some of previous studies on this topic. The forest fuels tested in the experiments were dead needles of Pinus pinaster, straw of Avena sativa, dead leaves of Eucalyptus globulus and a mix of shrubs mainly composed by heather (Erica australis) and gorse (Pterospartum tridentatum). The experimental results of the tests with and without forced flow inside a fire whirl generator were compared with tests in similar conditions out of the generator. It was possible to evaluate the effects of fuel bed size, bulk density and external vorticity on several parameters like flame height and diameter, mass decay and heat release rate. The results show that forced flow increases dramatically the burning rate and reduces the time needed to achieve a high rate of energy release. Comparison with results of other sources show that the flames that are generated in the present fire whirl generator are in a transition from fire whirl to pool fire regime and that it is possible to scale up some flow and thermal properties of field scale fire whirls and to derive predictive models on the basis of laboratory scale experiments.
1. Introduction
understand and predict this phenomenon [4]. For example, Alexander and Thomas [5] studied the conditions leading to the formation of a fire whirl that occurred during a fire near Santa Barbara, California, in March 1964. These authors stated that unstable atmosphere, high temperature at the ground level, a very low relative humidity, moderate winds and a large fuel load were defined as potential conditions for this phenomenon occurrence. In a general view, one of the main factors that contributes for the formation of fire whirls is the buoyancy force. Other studies with similar statements have been reported by several authors (e.g., [6–8]). The thermodynamic and kinematic characterization of fire behaviour regarding the formation of fire whirls has been the subject of several studies on analytical [9–11], numerical [12,13] and laboratory experiments [14–25] over the past few decades. The earliest reported experimental studies were conducted by Emmons and Ying [14] in 1967 using a rotating screen to generate vorticity on a liquid (acetone) pool fire with only 10 cm diameter. The fire whirl created consisted of a rotating cylinder fuel rich inside and lean outside flame and the authors concluded that the mass loss rate increased with the increase of ambient air circulation. Other experimental studies have been performed, most of them using liquid and gas fuels. One of the few reported experimental studies of fire whirls formation using solid fuels was carried out by Martin [15] who employed cross-piled wood sticks with different dimensions and moisture content values to measure the burning rates in a cylindrical device in which the tangential inlet of air
Forest fires are a natural disaster with great social, economic and environmental impact, making the study and understanding of fire behaviour essential for their successful management. In some circumstances forest fires can cause great devastation and endanger citizens and those who are in charge of suppressing them. It is recognized that situations in which fires spread under extreme weather and topography conditions are more likely to cause wide destruction and produce accidents [1,2]. Fire whirls are a phenomenon in which the fire under certain conditions acquires a vertical vorticity and forms a swirl or a flame column with vertical orientation that is also designated as a fire tornado [3]. They are characterized by a very high rate of energy release, unexpected formation and erratic movement, making them one of the most extraordinary and dangerous phenomena associated with fire behaviour [3]. In their erratic movement fire whirls can transport fire to areas far from the place where they were formed and they can also contribute to spread the fire by spotting due to embers that are lifted in their core to great heights and transported far from the fire perimeter by wind. Several studies have been conducted in order to understand better the conditions required for fire whirls to form and their role on the fire whirl properties. Reports describing the role of topography, fuels and weather on the formation of fire whirls, are important to better ⁎
Corresponding author. E-mail address: [email protected] (C. Pinto).
http://dx.doi.org/10.1016/j.firesaf.2016.11.004 Received 9 August 2016; Received in revised form 9 November 2016; Accepted 23 November 2016 0379-7112/ © 2016 Elsevier Ltd. All rights reserved.
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
density (kg m−3)
Nomenclature
ρ
d h m t v F H L M V ṁ Q̇
Subscripts
diameter (m) height of container (m) mass loss (kg) time (s) velocity (m s−1) fan frequency (Hz) low calorific power (kJ kg−1) length of the flame (m) dry mass (kg) volume (m3) mass loss rate (g s−1) heat release rate (kW)
o b c D f fu m M i in out s
Greek symbols β
initial bulk container chaparral dead fuel flame fuel medium maximum inlet inside outside solid particles
packing ratio (-)
lateral wind velocity and developed and validated a scaling law that predicts the critical wind velocity at which the most intense fire whirl was generated. It was observed too that in actual mass fires the occurrence of fire whirls is time-dependent, in contrast to wellcontrolled small-scale laboratory experiments, but this finding was not explored. In [22] it was verified that an appropriate scale model can successfully reconstruct the occurrence of fire whirls moving along a line fire and that a critical velocity was found to be proportional to the vertical buoyant velocity, which depends on the burning rate and length scale of the burning area. One point that was not discussed in that paper was the possibility of correlating the ratio between the critical velocity and the vertical buoyant force to the existence of a critical Froude number below which no fire whirls can be observed. It was stated in this study, but not confirmed, that a curved-shape geometry is necessary to generate moving type fire whirls. An interesting study was conducted by Chuah et al. [23] on the formation of inclined fire whirls, i.e., fire whirls for which buoyancy cannot be the source of support. Fire whirls with this configuration are observed frequently in forest fires. In this study an adjustable support for producing inclined fire whirls in the laboratory was placed on the base of the fire whirl generator. It was shown that the flame length was independent of the inclination angle at a given burning rate. The upper section of the apparatus was modified to observe the influence of flame height for a certain burning rate and it was concluded that for the same burning rate the flame height can be changed. So, if the structures of fire whirls are sufficiently strong, they will be dominated by the rotation plus entrainment, rather than by buoyancy. So, the several mechanisms that can explain the formation of fire whirls have been a focus point to understand and predict the occurrence of this phenomenon, as shown by the studies of [20–23] mentioned above or, for example, the study of [25] in which it is shown experimentally that the counter-rotating vortex pair is a possible origin of fire whirls that occur downwind of a fire area and are shed downwind. Although important research has been done to better understand the mechanisms and behaviour of fire whirls generation through theoretical or/and experimental studies, it is difficult to get results that are in accordance from different studies with similar analysis conditions (theoretical vs experimental for example), as shown by Hayashi et al. [24] that try to validate, reproducing in laboratory fire whirls at small scale with methanol, the relation between the flame height and the vortex structure that was studied by Chuah et al. [26]. In this work the Burke-Schumann-Burgers (BSB) model for small-scale fire whirls based on the assumption that vortex structures of fire whirls are able to be approximated by the Burgers vortex was developed. The results indicate that the vortex around the fire whirl cannot be
at the bottom could be controlled. They found that burning rates in cribs with fire whirls were 1.4–4.2 times greater than in cribs burning without whirl formation, although they reported that fuel particle density, bulk density, surface area to volume ratio, packing ratio and moisture content influenced the results, they did not study other parameters, such as flame height, that characterize and that are necessary to better understand this phenomenon. A remark that is made on the study of fire whirls by various authors is the lack of data in this area through experimental studies at laboratory scale. In recent years great importance was given to develop experimental equipment capable of further study on this topic. As an example we mention the extensive experimental work performed recently by Lei et al. [16–19]. The experimental apparatus used in their studies was a square enclosure with 2 m×2 m×15 m made of tempered glass and open on the top. The vertical channel had 20 cm wide vertical gaps at the corners to induce tangential air entrainment by the flame produced by liquid fuel. Different sizes of fuel pans (diameters 10–55 cm) were used and the liquid fuel was n-heptane (97%). In [16] the dynamic differences between fire whirls and pool fires were shown, and correlations of the burning rate, flame height and temperature of fire whirls were established. It was concluded that the pool diameter has a great influence on the burning rates of fire whirls, as in general pool fires, and that the transition from laminar to turbulent burning occurs when the pool diameter increases. The study of quasi-steady burning rate of laminar and turbulent fire whirls was done in [17] establishing correlations based on the boundary layer theory, film theory and Chilton-Colburn analogy. Due to the flame instability observed in the previous studies, Lei et al. [18] explored the characteristics of the flame of fire whirls distinguishing the flame revolution and the flame precession. Experimental observations have shown that in the formation of a fire whirl, the entire fire is tilted and revolves around its axis of symmetry with increasing angular velocity. Other example is the work done by Kuwana et al. [20–22] which gave a great contribution to the understanding of this phenomenon by reproducing experimentally at laboratory scale, fire whirls that occurred in real fires to study the mechanisms of formation of these fire whirls in different configurations. Examples are the fire whirl that occurred in Hifukushoato [20] in an open space and the case of fire whirls generated by the interaction between a line fire and the background wind as occurred in Brazil [22]. In [20] the Hifukushoato fire whirls were reconstructed at the 1/1000th scale using a large-scale wind tunnel and where lateral wind velocity was varied from 0.5 to 2 m/s to study its effect on fire-whirl formation. Different types of fire whirls formation in open areas where there was no combustible material were observed. The authors studied the influence of the 38
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
(Fig. 1a) the base of the FWG had a height of 1.3 m while in configuration C2 (Fig. 1b) the base height was raised to 1.8 m and a set of four axial fans was installed to create a forced tangential flow at the base of the combustion chamber. The FWG was installed in the large hall of the Forest Fire Research Laboratory that has a height of 11 m and an area of 72×16 m2. The vertical distance from the FWG exit to the ceiling was of 4 m and possibly a blocking effect exists in the vertical flow in the channel.
2.2. Fuels and containers In the tests reported in this paper four different fuels were used: (1) dead needles of Pinus pinaster (one test), (2) straw of Avena sativa (two tests), (3) dead leaves of Eucalyptus globulus (one test) and (4) a mix of shrubs mainly composed by heather (Erica australis) and gorse (Pterospartum tridentatum) (sixteen tests). These fuels were chosen as they are quite common in forested areas in Central Portugal and in several Mediterranean climate regions. They are also relatively easy to acquire in order to assure repeatability of the tests. The fuel was placed in a container at the base of the FWG in each test. The containers were made of a metallic grid welded to form a cylinder open on the top. Five different containers designated as A, B, C, D and E whose shapes are shown in Fig. 2 were used. The values of the diameter dc and of height hc of each container are given in Table 1. In each test the pre-defined amount of fuel was placed inside the container to form a solid porous fuel with a bulk density ρb given by:
Fig. 1. A schematic illustration of the different configurations used on present study: a) configuration C1; b) configuration C2.
approximated by the Burgers vortex, probably because of the complexity of the flow associated with density and temperature changes near the flame. It is found that sometimes theoretical models cannot be applied correctly to the experimental data showing the importance of having laboratory experiments to explain and validate fire whirl behaviour. It is therefore important to reproduce these phenomena at a scale closer to the real scale. Using image analysis, it is important to determine parameters from real fire whirls such as shape and size of their base and flame height to develop models to predict fire whirl properties at various scales. The present work aims contributing to a better understanding on the formation of fire whirls and characterization of this phenomenon through the operation of a novel equipment the Fire Whirl Generator (FWG), and the development of a test methodology to analyse several fire whirl parameters, including the rate of fuel mass burning, the energy release rate, the flame height, the flame diameter and the influence of fuel bed dimension, bulk density and forced air inlet velocities at the base of the FWG. A particularity of this study is the use of typical forest fuels instead of fossil fuels as was done in some of the previous studies on this topic. It is important to analyse and to understand the differences, advantages and limitations of using solid fuels like those that are found in forest fires in relation to the use of gaseous or liquid fuels. We also intend to characterize differences associated to the fuel bed properties namely its dimension and porosity and of the boundary conditions namely the confinement of the flame with or without an external tangential flow. Comparison with results from other laboratory scale experiments and full size fire whirls and pool fires showed that it is possible to derive predictive models for this type of fire whirls in a wide range of scales.
Mfu
ρb =
Vfu
(1)
In this equation Mfu is the dry mass of the fuel that was estimated by the previous measurement of the moisture content of the fuel (FMC). The value of FMC was measured for one 1.5 g sample of fuel using a moisture analyser A & D MX-50 (0.01 g) immediately before the experiment. The value of Mfu was changed in some experiments but in all other tests reported here its value was kept equal to 4.0 kg. The volume of the fuel Vfu is given by the following equation:
Vfu =
π . dc 2. hfu 4
(2)
In Eq. (2), the height hfu of the fuel inside the container did not always coincide with the height of the container. For example, when a load of 4.0 kg of shrub was placed in container B the value of hfu was around 0.55 m. Typical values of bulk density for tests with shrub fuel with a load of 4.0 kg for each container are given in Table 1. We consider the packing ratio β defined as usual [28] by:
2. Experimental methodology
ρb
β=
2.1. Fire whirl generator
ρs
(3)
In this work we considered that the volumetric mass density ρs of the solid material had an average value of 700 kg/m3. The bulk density can be defined also as the ratio between the volume Vs occupied by the solid material and the total volume Vfu occupied by the fuel:
Following the work of Lei et al. [16–19] and Satoh et al. [27] we designed a Fire Whirl Generator (FWG) apparatus consisting of a vertical channel with a quadrangular section of 1×1 m2 with a height of 6 m with two sides made of tempered glass and the other two made of steel sheet. The base of the FWG has a section of 2×2 m2 and 1.8 m height tapering to the 1×1 m2 section of the main channel section that is open at the top. Each corner of the channel has a vertical opening 10 cm wide to induce tangential air entrainment. Inside the base of FWG there is a plate of 1×1 m2 covered by ceramic tiles to support the fuel container. The FWG is a new equipment that in its development was subject to several modifications that involved some changes of its configuration during the research program. In this paper we refer to two configurations that are shown schematically in Fig. 1. In configuration C1
β=
Vs Vfu
(4)
The value of Vs is equal to:
Vs =
39
Mf ρs
(5)
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 2. Containers in which the fuel was placed.
2.3. Forced flow system
Table 1 Designation and dimensions of the containers used. Container
hc (m)
dc (m)
ρb (kg m−3)
A B C D E
0.760 0.760 1.160 0.760 0.305
0.350 0.500 0.500 0.800 1.120
54.70 26.81 17.56 14.47 13.31
As was described above a set of four axial fans Rosenberg DR 560-4 of 0.81 kW was attached to the base of the FWG to induce a tangential flow to create a forced vortex inside the combustion chamber. Each fan had a duct that made the transition from its circular section (∅ 560 mm) to a rectangular section (200×300 mm2). The fans were controlled by a frequency driver Omron Sysdrive 3G3HV Inverter 3G3HV-A4037-CE, in a range between 0 Hz and 50 Hz. The average flow velocity vi at the inlet of the combustion chamber was measured with a hot wire anemometer Multifunction KIMO AMI 300. A linear relationship between vi (m/s) and the electrical frequency F (Hz) of the fans given by Eq. (6) was found (R2=0.997).
vi=0. 0995F −0. 240
(6)
In the present experiments the following frequency values were used (Hz) 0, 5, 10, 20, 30, 40 and 50, corresponding to inlet velocities of (m/s) 0, 0.31, 0.78, 1.73, 2.74, 3.86 and 4.83, respectively. 2.4. Sensors and data acquisition Fig. 3. A schematic illustration of the system used to measure the evolution of the fuel mass during the combustion process.
2.4.1. Mass loss and heat release rates The mass of the fuel in the container was measured continuously during each test using the system shown in Fig. 3. It consists of a platform on which the container is placed with a lever to allow the
Table 2 Parameters of the experiments reported in this work. Test
FMC (%)
Fuel
Cont.
In/Out test
vi (m s−1)
ṁM (g s−1)
̇ (kW) QM
LM (m)
dfm (m)
GV15 GV16 GV19 GV20 GV23 GV25 GV26 GV27 GV29 GV31 GV32 GVV6 GVV15 GVV17 GVV20 GVV23 GVV25 GVV30 GVV36 GVV37 GVV38 GVV50 GVV52 GVV54
18.2 18.2 15.1 18.5 18.0 16.0 16.0 16.0 16.0 14.4 14.4 11.4 13.6 13.6 11.3 11.3 12.5 10.9 8.4 10.2 10.2 10.5 10.5 10.5
Shrub Shrub P. needles Straw Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Eucalyptus Shrub Shrub Shrub Shrub Shrub
B A B B D D C C C E E B B B B B B B B A D B A D
In In In In In Out In Out Out In Out In In In In In In In In In In Out Out Out
– – – – – – – – – – – 0.78 2.74 3.86 1.73 4.83 – 0.31 – 2.74 2.74 – – –
32.13 22.47 28.59 29.37 47.97 67.87 50.13 38.37 39.92 85.60 79.80 43.96 67.39 58.41 46.88 63.48 50.06 41.67 19.56 46.32 78.31 – 15.68 76.97
540.3 356.4 386.0 472.6 811.6 1180.0 871.8 667.3 694.2 1521.0 1419.0 916.6 1400.9 1214.1 977.8 1323.9 1042.2 869.5 489.6 869.0 1635.9 – 350.4 1720.0
4.90 4.13 4.53 2.60 6.13 4.32 5.00 4.65 4.31 4.58 4.13 5.26 5.56 5.46 5.08 5.08 4.62 4.92 3.52 3.10 3.51 3.84 – –
0.45 0.38 0.45 0.46 0.59 0.64 0.55 0.53 0.61 0.66 0.69 0.48 0.38 0.39 0.44 0.46 0.68 0.59 0.52 0.41 0.36 0.33 0.26 0.41
40
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 4. Typical images taken by the photographic camera for five different phases of GV27 test made outside of FWG.
container C in three different conditions. In Fig. 4 five photos taken during test GV27 in which the container was placed out of the FWG are shown. The time elapsed from ignition is shown in each photo. In Fig. 5 photos of test GV 26 made with the same basket and fuel conditions but inside the FWG in its configuration C1 (Fig. 1a) are shown. In test GVV30 that is shown in Fig. 6 forced ventilation vi=0.31 m s−1 was used. The results of these tests are detailed below.
balance to be far from the heat released during the combustion process. A digital balance A & D HW-100KGL with a resolution of 10 g with a frequency of 1 Hz, by RSKey v.1.40 software, connected to a laptop was used for this purpose. This system was duplicated in order to perform tests with similar fuel containers in a quiescent ambient out of the FWG to check the effect of the boundary conditions. 2.4.2. Flame geometry analysis In each test at least two cameras were used to record images: a video camera Sony DCR-SR87 and a digital camera Canon EOS 550D. Both cameras were located at fixed places close to each other to record images in order to measure flame diameter and flame height. The flame height L and flame diameter df of the fire whirl were obtained through image processing of frames selected with a time interval of five seconds using Photoshop CC software. The flame height L from the base of the container to its visual tip was measured manually in each image using a reference length as a vertical scale. The diameter df of the fire whirl was measured in each time step with the same method at three selected heights and an average value was taken to characterize it. The highest values of the flame height LM in the series of photos for each test and the average value of the fire whirl diameter dfm were two of the parameters analysed for each test as it is summarised in Table 2. In this Table each test has a designation that is composed by a set of letters (GV or GVV) related to the test configuration and a number corresponding to chronological order. The designation GV corresponds to tests made with configuration C1 (Fig. 1a) and also to tests with the fuel outside the FWG and GVV corresponds to tests made inside the FWG with configuration C2 (Fig. 1b).
2.5.2. Mass loss analysis The mass of the fuel in the container was measured during each test with a frequency of 1 Hz and the values were corrected to a dry basis adjustment. In order to smooth the data average values at five seconds intervals were calculated sequentially along the test. The mass loss rate, ṁ, is defined here as:
ṁ =
dm dt
(7)
Its value was determined over the duration of the test at five-second intervals. In the evaluation of the maximum value ṁM of the mass loss rate it was found that in some tests this value was well defined while in others there was not a single maximum value but a series of similar values. In the latter cases the maximum rate of mass loss ṁM was taken as the average of the five higher values of ṁ in the test. The values of heat release rate Q̇ (kilowatt) were calculated using the following equation:
Q̇ = ṁ × Hfu
(8)
In this equation the mass loss rate is expressed in kg/s and Hfu is the low calorific power (kJ/kg) of the fuel used. The value of Hfu was computed using CEN EN 14918:2009 [29] and data from Lopes [30]. In Table 3 the calculated values of Hfu for each fuel are given. A typical value of 20 MJ/kg was considered in most calculations. When using Eq. (8) we are assuming that the combustion efficiency was 100% for each test, although this is not valid as some particles are released without burning completely. As the same value of Hfu was considered in each test the mass and energy release rates are proportional. Therefore, we will refer to either of them to characterize the intensity of the fire whirl. For each test the ̇ of the heat release rate was evaluated in the same maximum value QM form as was done for ṁM. The mass loss and mass loss rate values for a typical test are shown
2.5. Experimental procedure and data analysis 2.5.1. Test procedure In each test the predefined amount of fuel was collected taking into account its moisture content and packed in the container. The container was placed on the platform inside the FWG, the ignition was initiated along the perimeter at the basis of the container using a gas burner and the fans were turned on and adjusted to achieve the desired flow velocity. When the ignition was carried out, the video and photo cameras were started and the measurements initiated. Typical images of three tests GV27, GV26 and GVV30 are shown respectively in Figs. 4, 5 and 6. These tests were performed using
Fig. 5. Typical photo images for five different phases of GV26 test.
41
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 6. Typical photo images for five different phases of GVV30 test.
around the central axis due to the interaction of the ascendant flow and the entrained air from the four openings of the FWG, which indicates the formation of a fire whirl.
Table 3 Characteristic parameters of different fuels. Test
Fuel
Moisture content (%)
Test duration (s)
Flame duration (s)
Hfu (MJ/kg)
GV15 GV19 GV20 GVV36
Shrubs P. needles Straw Eucalyptus
18.2 15.1 18.5 8.4
230 300 260 230
210 300 70 230
16.9 13.5 11.7 22.5
2.5.3. Flame height and flame diameter The flame height or length L and flame diameter df of fire whirls are important parameters when studying fire whirls. They have a very unique visual aspect in comparison to ordinary flames, characterized by a relatively large height and small flame diameter. These features are relatively easy to obtain in the field during experimental and real fires and are very often registered in still and video pictures. It is therefore of great interest to be able to establish relationships between fire whirl height and diameter with other combustion process parameters that support the fire whirl like for example the fuel mass consumption rate or the equivalent heat release rate. A very common relationship that is used in the literature (e.g., [31,34]) is the following:
L = a × Q̇
b
(9)
In this equation L is the flame height and a and b are two model parameters. In the present study, according to Eq. (8) Q̇ is considered to be proportional to ṁ , therefore we will use as well the parameter ṁ to analyse the relationship between the flame height and the mass loss rate. In each experiment the flame diameter varies between a relatively narrow range as it is shown in Fig. 8 for test GV16. For this reason, we will characterize the flame diameter by its average value dm. Given the transient nature of the present experiments in the results and discussion section we shall analyse the time evolution of some parameters when they are correlated with each other in the course of time. But for other parameters like the mass loss, heat release rate or flame height, average or extreme values will be used to characterize the role of some factors and to produce models.
Fig. 7. Instantaneous results for mass loss and mass loss rate for GV31 test.
3. Results 3.1. Role of fuels Fig. 8. Evolution of the flame diameter for GV16 as a function of time and its average value dm.
In order to determine which fuel would be more appropriate to perform the test program preliminary experiments were carried out using various types of surface fuels namely shrubs, dead pine needles, eucalyptus leaves and straw, that are commonly found in forested areas. The tests were made using container B and assuring the same conditions of ambient temperature and preparation methodology, and their parameters are summarised in Table 3. In these tests the fuel load was 4 kg for all fuels. In Fig. 9a and b the corresponding values of flame length and mass loss rate are plotted. In Fig. 9a it is possible to observe that the flames of the tests with straw are very short and last for less than one minute. The other three fuels have similar values of flame length and flame duration but shrubs have higher values of maximum mass loss rate. For this reason and due to the fact that shrubs are possibly more representative of the type of fuels in which fire whirls are more likely to occur we decided to select shrubs as the basic fuel for our test program.
in Fig. 7 in which the initial mass loading was mo=4.0 kg. The instantaneous values of mass at one second intervals are shown as well as the smoothed curve with 5 s average values. The mass loss curve corresponds to average values at 5 s intervals. As reported by Sun et al. [31] the combustion process has three phases. In the initial phase that lasts for 25 s in this test, there was a slight loss in mass corresponding to the loss of moisture and start of combustion; in the second phase (lasting from 25 to 75 s there is a marked decrease of mass corresponding to flaming combustion of the fuel; the third phase corresponds to combustion of thicker particles of the fuel partially in non-flaming combustion. During the second phase in which the maximum value of mass loss rate is observed the flame initially looks like a general fire burning without vorticity but after some time the flame starts to rotate 42
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
a5
b
40
Shrubs Dead pine needles 4
Shrubs
35
Dead pine needles
Straw
30
Eucalyptus leaves 3
Straw Eucalipto
L (m)
ṁ (g.s-1)
25
2
20
15 10
1
5 0 0
50
100
150
200
250
300
0 0
t (s)
50
100
150
200
250
300
t (s)
Fig. 9. Results for different types of fuels: a) flame height of fire whirl; b) mass loss rates.
be limited by the oxygen supply. It is interesting to notice that Rothermel [28] indicates that the potential reaction velocity of excelsior combustion decreases for packing ratios greater than approximately 0.015. The same trend is observed in our work as maximum mass loss rate decreases for packing ratios greater than 0.015. In order to separate these two effects tests with constant values of β in different containers should be performed to better assess the effect of the size of the fuel bed. The average values of flame diameter for tests made inside and outside of FWG are shown in Fig. 12 as a function of container diameter. The results follow the trends shown in Eqs. (10) and (11) with a correlation coefficient equal to 0.97 and 0.89, respectively for tests inside and outside the FWG:
Table 4 Description of the fuel inside the containers. Test
Container
Packing ratio β
GV16 GV15 GV23 GV31
A B D E
0.078 0.038 0.021 0.019
3.2. Role of containers The parameters of tests GV16, GV15, GV23 and GV31 using the different types of containers are given in Table 4. The corresponding results are shown in Figs. 10 and 11. As can be seen in Figs. 10a and 11a, larger diameter containers, dc, correspond to higher values of ṁ and lower values of burning duration. These results are in agreement with other author's studies [15,16]. In Fig. 10b the values are in accordance with Fig. 10a, except for container E (dc=1.12 m), which corresponds to a lower value of flame height in relation to containers B (dc=0.50 m) and D (dc=0.80 m). This lower value could be related to the small height of the container and to its large diameter in relation to the dimensions of table base (1×1 m2). We recognize that in these tests two different factors are affecting our results. One is the container diameter and the other is the packing ration β, that is a measure of the fraction of the fuel array volume occupied by the fuel, and expresses the relationship between the fuel particles and air existing between them. In Table 4 the fuel packing ratio values for each container are given. The maximum mass loss rate values ṁM for these tests are shown in Fig. 11a and b as a function respectively of the container diameter and the packing ratio. It can be seen that ṁM increases with dc and decreases with β. This results can be related to the oxygen supply, which means that for higher values of β and for smaller dc, the ṁM can
d f =0. 42.dc 0.52
(10)
d f =0. 41.dc 0.39
(11)
3.3. Role of forced flow A set of tests (GVV25, GVV30, GVV6, GVV20, GVV15, GVV17 and GVV23) were performed inside the FWG with configuration C1 and container B to analyse the influence of forced flow. In these tests the value zero for the inlet velocity of the forced flow means that the fans were turned off, i.e., the fire whirl was formed due to the natural flow. As can be seen in Fig. 13a and b the maximum value of mass loss rate ṁM increases generally with the forced flow velocity, at the same time the average flame height increases a little (+11%) and the average flame diameter decreases significantly (−33%). The consequence of this is that the surface of the external cylinder enveloping the flame decreases. The apparent contradiction between these two observations,
Fig. 10. Instantaneous results for different diameter containers: a) mass loss rate variation; b) flame height variation.
43
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 11. Results for maximum mass loss rate as a function of: a) diameters of containers; b) fuelbed compactness.
Fig. 12. Flame diameter as a function of container diameter for tests inside and outside the FWG.
Fig. 14. The evolution of mass loss rate and flame height in the course of time for GV16 test.
could be explained by the fact that the flame surface is wrinkled because of the rotational flow, increasing in the same manner the flame surface. By retroaction the heat transfer by radiation (proportional to the flame surface) between the flame and the solid fuel increases contributing to an acceleration of the degradation of the solid fuel. Two exceptions are observed for vi=0 m s−1 and for vin=2.74 m s−1. There is a decrease of the mass loss rate when vi varies between 0 and 0.31 m s−1; the value of ṁ for vi=2.74 m s−1 is actually an absolute maximum and it decreases for vi > 2.74 m s−1. The same effect is shown in Fig. 13b in which we can see that the flame length and diameter modify their trend of variation for vi > 2.74 m s−1. This effect must be studied more carefully as it may be an indication of the existence of either a change of regime in the fire whirls development or the presence of a stagnant flow near the top of the FWG due to the presence of the ceiling constraining the development of the fire whirl.
3.4. Flame height prediction In Fig. 14 the evolution of flame height and mass loss rate (that is proportional to Q̇ ) as a function of time for test GV16 is shown. As can be seen in this figure there are two different phases in the process: (1) the growth phase in which the value of ṁ and of L increase practically at the same time; (2) the decay phase in which ṁ decreases but L decreases less rapidly. From this observation we can conclude that there is not a bi-univocal relationship between L and ṁ and two separate laws can be established for phases 1 and 2, respectively. This is illustrated in Fig. 15a and b in which the data points corresponding to phases 1 and 2 of this test are shown. The corresponding values of ai and bi are: a1=1.09 and b1=0.34; a2=1.61 and b2=0.21. Similar results were found in the other tests,
)
Fig. 13. Results for different air forced flow: a) maximum values of mass loss rate; b) maximum values of height flame and diameter flame of fire whirl.
44
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 15. A schematic illustration of the different phases that correlate the flame height with mass loss rate related to GV16: a) the growth phase (phase 1); b) the decay phase (phase 2).
distinguishing phases 1 and 2. The corresponding result for test GV16 is shown in Fig. 16. The power law fitting yielded the following results: ao=1.29 and bo=0.29. Unless stated otherwise in the following discussion the results are presented in the form of ao and bo. In Table 5 the values of ao, a1, a2 and of bo, b1, b2 for the relevant tests of the present study are shown. 3.5. Role of boundary conditions Flame height and heat release rate are two variables that are strongly dependent on each other, Zukoski [33], cited by Sun et al.[31]. Sun et al.[31] correlated the flame height with heat release rate for dead chaparral fuels using the following equation: 2/5 LD=0. 20QḊ
Fig. 16. A schematic illustration of the relation between flame height and mass loss rate for GV16 test.
They compared it with correlations from earlier studies for ovendried fuels and concluded that the two-fifths power fits well for dead or dry fuels. Dupuy [34] studied combustion dynamics and flames properties through experimental measurements to test models of a forest fire to analyse the variations of the maximum flame height with maximum heat release rate for forest fuels (Pinus pinaster needles and excelsior), these values are shown in Fig. 17a. In the present study we performed tests with the fuel outside the FWG and others inside the FWG in configuration C1 without forced flow. In the second case we had fire whirls due to the boundary conditions while in the first case a fire whirl was not formed. The results of the flame height of these tests are shown in Fig. 17a and b, respectively, and the following correlations were found for each set of tests:
Table 5 Coefficients of Eq. (9) relating Flame Height with mass loss rate. Test
a0
b0
R02
a1
b1
R12
a2
b2
R22
GV15 GV16 GV19 GV20 GV23 GV25 GV26 GV27 GV29 GV31 GV32 GVV6 GVV15 GVV17 GVV20 GVV23 GVV25 GVV30 GVV36
1.73 1.30 1.53 1.24 1.26 0.67 1.09 0.72 0.65 0.21 0.18 1.66 1.15 1.15 1.81 1.92 1.06 0.58 2.10
0.21 0.29 0.28 0.03 0.31 0.44 0.33 0.48 0.47 0.66 0.72 0.21 0.35 0.37 0.20 0.17 0.36 0.58 0.12
0.61 0.72 0.72 0.003 0.68 0.90 0.68 0.86 0.88 0.82 0.85 0.42 0.53 0.86 0.23 0.40 0.83 0.81 0.45
0.70 1.09 0.19 1.26 1.11 1.43 0.60 0.58 0.69 0.03 1.04 1.82 1.84 1.50 2.64 1.64 1.26 0.40 0.67
0.39 0.34 0.81 0.25 0.35 0.23 0.48 0.52 0.44 1.17 0.28 0.23 0.26 0.31 0.15 0.30 0.33 0.66 0.44
0.73 0.48 0.25 0.90 0.39 0.70 0.90 0.98 0.96 0.91 0.68 0.47 0.52 0.92 0.23 0.75 0.87 0.86 0.76
1.95 1.61 1.46 0.08 1.31 0.68 1.17 1.09 0.61 0.27 0.22 1.55 0.63 0.84 0.96 2.03 0.97 2.73 1.37
0.21 0.21 0.31 0.86 0.27 0.46 0.33 0.34 0.49 0.59 0.63 0.20 0.52 0.46 0.38 0.12 0.38 0.12 0.31
0.60 0.49 0.77 0.90 0.88 0.91 0.95 0.93 0.76 0.89 0.96 0.78 0.91 0.96 0.76 0.34 0.84 0.54 0.70
(12)
0.46 Lout =0. 18Q̇out
(13)
̇ 0.38 Lin=0. 35Qin
(14)
Eq. (13) corresponds to the flame height outside the FWG as a function of the heat release rate while Eq. (14) corresponds to the tests inside the FWG in which fire whirls were formed. In Fig. 17a it is shown that our correlation fits well with the correlation of Dupuy [34] and that both are greater than the results obtained by Sun [31]. This difference may be due to the combustion regime; both ours and Dupuy experiments are not steady combustion regimes while the study of Sun [31] correspond to steady state. As can be seen in this figure our experiments cover a much wider range of heat release rates in comparison to the previous two studies. In Fig. 17b it can be seen that the results for fire whirls yield higher values of L for the same value of Q̇ in relation to test performed out of the FWG. The relative confinement of the FWG can contribute to facilitate the development of the vertical flow and also the vertical openings induce tangential flow that enhances the fire whirl effect putting in evidence the role of boundary conditions in the flame development.
with b1 close to 0.4 and b2close to 0.3. Weise et al. [32] made tests with different shrub fuels and found that in the general case the growth of flame height had a time lag in relation to the ṁ growth. This time lag decreased with the dryness of the fuel; in the case of dead fuels the time lag was practically equal to zero as we found in the present experiments. When we observe an image of a fire whirl it is not easy to assess if it is in phase 1 or 2, therefore for practical purposes we will develop an overall relationship between L and ṁ for the entire test without 45
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 17. Flame height versus heat release rate: a) for tests made outside the FWG b) for tests made inside the FWG without ventilation (configuration C1).
3.6. Comparison with full scale fire whirls
Table 6 Parameters of the tests shown in Fig. 18a and b.
Using the present research, a relationship between the maximum heat release rate and the diameter of the fuel container is proposed in the form of Eq. (15). A complementary relationship between the maximum values of the flame height and the heat release rate is proposed in Eq. (16) that is similar to Eqs. (12)–(14).
̇ = a3dcb3 QM
(15)
b
̇4 LM = a4QM
(16)
Our results together with those found in the literature are shown in Fig. 18a and b for these two sets of parameters. The numbering used in graphs 18a and 18b is described in Table 3. Results from tests GV27, GV29, GV25 e GV32 performed inside the FWG are fitted by curve 1 in both figures; results from tests GV27, GV29, GV25 e GV32 performed outside the FWG correspond to curve 2 and results from tests GVV38, GVV15 e GVV37 with a forced flow of vi=2.74 m s−1 correspond to curve 3. Results from fire whirls with liquid fuel (heptane) obtained by other authors are shown in this figure and fitted by curves 4 and 5; results from pool fires also with heptane are fitted by curves 6, 7 and 8 in those figures. As can be seen in these figures the reported tests cover a wide range of scales of fuel size, flame height and heat release rate. Our results are in the middle of this range and seem to be in a transition between small laboratory scale tests and large scale field tests. ̇ and L As can be seen in these figures the results for both QM corresponding to pool fires are usually higher than those for fire whirls. The results of our tests outside the FWG are similar to pool fires while those inside the FWG are in a transition from fire whirls to pool fires. The tests with forced flow are more similar to fire whirls showing the
Ref.
Description
a3
b3
R32
a4
b4
R42
[1] [2] [3] [4] [5] [6] [7] [8]
Our study (inside FWG) Our study (outside FWG) Our study (forced flow) Lei et al. [16] Zhou et al. [35] Klassen and Gore (1994)a Zhou et al. [35]b Kosekia
1213 1339 2047 3307 3224 2295 1211 2401
1.19 0.95 0.74 2.25 2.20 2.41 2.52 2.13
0.98 0.97 0.87 0.99 0.99 0.99 0.99 0.99
2.82 8.61 0.17 0.75 1.49 0.22 0.21 –
0.08 −0.10 0.44 0.28 0.18 0.35 0.36 –
0.10 0.60 0.22 0.98 0.97 0.99 0.97 –
a b
Values from Table 2 of Lei et al.[16]. Values from Table 2 of Zhou et al.[35].
Fig. 19. Formation of a fire whirl during field experiments at Gestosa.
8
3 7
1 5
2
4
6
Fig. 18. Results of different studies: a) maximum heat release for different diameters of containers b) maximum heat flame versus maximum heat release. The sources of data that are indicated by a number are given in Table 6.
46
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
result indicates that it is possible to determine predictive models to estimate the properties of fire whirls on the basis of laboratory scale tests. The values of coefficients a5 and b5 for the series of data points that are shown in Fig. 20 are given in Table 7. We see also that the flames that are formed in our FWG are in the transition from a fire whirl to a free flame regime. In order to overcome this difficulty, we intend to increase the forced flow velocity or to create an opening on the ceiling of the Laboratory to eliminate the stagnation effect that the presence of the ceiling imposes on the flow. 4. Conclusions A laboratory scale Fire Whirl Generator using forest type fuels was developed and the proposed experimental methodology to analyse the properties of fire whirls was described. It is possible to study systematically the role of various factors like the fuel type, the size and porosity of the fuel bed, the presence of enclosing walls and the effect of forced flow on various parameters that can be used to characterize fire whirls like for example the mass loss or heat release rates, the flame length and diameter. The fuel used in these experiments was dried shrub and the maximum heat release power was of the order of 1 MW which is higher than the reported values for similar laboratory tests. Given the transient nature of the experiments relationships between correlated parameters or with average or extreme values were established. Comparison with similar experiments showed that the properties of the flames generated with the present configuration of the FWG are similar to fire whirls produced in other laboratory studies and with full scale fire whirls. It was found nevertheless that further changes must be explored in further work to possibly minimise the effect of the flow stagnation that is produced by the presence of the ceiling. The effect of fuel bulk density must be studied more systematically as well. Further work will be performed to explore the potential of the FWG extending the range of parameters tested so far using other sensors to analyse the flow and temperature field inside the flames which will provide us data to better analyse and understand the physical mechanisms associated to the formation and to the complex dynamics of fire whirls.
Fig. 20. Results of different studies of maximum height flame versus diameter of containers or the diameter of the base of the fire whirls. The sources of data that are indicated by a number are given in Table 7. Table 7 Parameters of the tests shown in Fig. 20. Ref.
Description
a5
b5
R52
[1] [2] [3] [4] [5] [6] [7] [8]
Our study (inside FWG) Our study (outside FWG) Our study (forced flow) Lei et al. [16] Zhou et al. [35] Klassen and Gore (1994)a Zhou et al. [35] [Table 2]b Real fire whirls
5.11 4.19 4.86 7.48 7.48 3.48 2.72 1.09
0.10 −0.10 0.33 0.64 0.44 0.86 0.90 1.06
0.20 0.60 0.02 0.96 0.97 0.99 0.96 0.98
a b
Values from Table 2 of Lei et al.[16]. Values from Table 2 of Zhou et al.[35].
importance of the boundary conditions namely the external vorticity and the dimension of the fuel bed. These results show that in our experiments there is a transition from fire whirl to pool fire regime. The values of coefficients a3, b3, a4 and b4 of Eqs. (15) and (16) are given in Table 6. The above results allow us to propose a relationship between the maximum flame height and the size of the fuel bed given by:
LM =
a5dcb5
Acknowledgements
(17)
The collaboration given to the experimental program that supports this paper by several collaborators – namely, by Mr. N. Luís, Mr. R. Oliveira, Mr. A. Cardoso, Mr. J. Lourenço, Mr. D. Lopes and Miss Ana Rosário – is gratefully acknowledged. The first author is thankful to IEFP and to ADAI for the professional internship. The authors are grateful for the referees' helpful comments.
It is easy to see that:
a5 = a4 × a3b4
(18)
b5 = b3 × b4
(19)
Photos or video images of fire whirls can be used to obtain data on the two parameters that are used in Eq. (17). Provided that there is some scale factor to assess vertical and horizontal dimensions and assuming that the flaming area at the base of the fire whirl – that is usually a circle – can be assimilated to the fuel bed or fuel container in the laboratory experiments we can check the validity of Eq. (17) to determine the properties of real scale fire whirls based on laboratory scale experiments. We analysed the images of nine full scale fire whirls (e.g., [36–38]) for which both the flame height and the diameter of the base of the fire whirl could be measured. As an example, Fig. 19 shows one of the images analysed from a test performed by our team in Gestosa (Portugal). The dimensions of the base and the height of the fire whirl are obtained through a relationship with the known dimensions of the fire truck. These data points are shown in Fig. 20 together with the results from laboratory experiments. The data numbering used in graph 20 is described in Table 7. It is interesting to notice that in spite of the fact that the dimensions of full size fire whirls extend the range of scales in at least one order of magnitude they follow the same trend of the laboratory experiments but with lower values of L. This
References [1] D. Viegas, Extreme fire behaviour, in: A.C. Cruz, R.E. Correa (Eds.), For Management Technology Practice Impact, Nova Science Publishers Inc., New York, 2012, pp. 1–56. [2] D.X. Viegas, A.R. Figueiredo, M.A. Almeida, V. Reva, L.M. Ribeiro, M.T. Viegas, et al., Relatório do Incêndio Florestal de Tavira/São Brás de Alportel, 2012 [3] J.M. Forthofer, S.L. Goodrick, Review of vortices in wildland fire, J. Combust. 2011 (2011) 1–14. http://dx.doi.org/10.1155/2011/984363. [4] A.R. Pirsko, L.M. Sergius, C.W. Hickerson, Causes and Behavior of a Tornadic FireWhirlwind, USDA For. Serv. Pacific Southwest For. Range Exp. Station, Berkeley, Calif, USA, 1965, pp. 1–13. [5] M.E. Alexander, D.A. Thomas, Wildland fire behavior case studies and analyses: value, approaches, and practical uses, Fire Manag. Today 63 (3) (2003) 4–7. [6] C.M. Countryman, Fire whirls…Why, When and Where, USFS Pacific Southwest Res. Stn., Berkeley, Calif, USA, 1971. [7] S. Soma, K. Saito, Reconstruction of fire whirls using scale models, Combust. Flame. 86 (1991) 269–284. [8] J. Forthofer, B. Butler, Large scale fire whirls: Can their formation be predicted?, in: Proceedings of the 3rd Fire Behavior Fuels Conference October 25–29, Spokane, Washington, USA, 2010, pp.1–7 [9] A. Muraszew, J.B. Fedele, W.C. Kuby, The fire whirl phenomenon, Combust. Flame.
47
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Park, MD, 2011, pp. 671 [25] M. Shinohara, S. Matsushima, Formation of fire whirls: Experimental verification that a counter-rotating vortex pair is a possible origin of fire whirls, Fire Saf. J 54 (2012) 144–153. [26] K.H. Chuah, K. Kuwana, K. Saito, Modeling a fire whirl generated over a 5 cm diameter methanol pool fire, Combust. Flame 156 (2009) 1828–1833. [27] K. Satoh, K. Yang, A horizontal fire-whirl design scenario for engineering performance-based fire-code applications, Int. J. Eng. Performance-Based Fire Codes 2 (2000) 48–57. [28] R.C. Rothermel, A mathematical model for predicting fire spread on wildland fire spread in wildland fields, USDA For. Serv. Res. Pap. INT-115, 1972, pp. 40 [29] CEN EN 14918:2009, Solid Biofuels Determination of Calorific Values. [30] A. Lopes, Misturas de Biomassa. Tese de Mestrado em Engenharia Mecânica na Especialidade de Energia e Ambiente, Universidade de Coimbra, Coimbra, 2014. [31] L. Sun, X. Zhou, S. Mahalingam, D.R. Weise, Comparison of burning characteristics of live and dead chaparral fuels, Combust. Flame. 144 (2006) 349–359. [32] D.R. Weise, T. Fletcher, S. Mahalingman, X. Zhou, S. Smith, L. Sun, Correlation of mass loss rate and flame height for live fuels, in: Proceedings of the Sixth Symp. Fire and For. Meteorol., Canada, 2005 [33] E.E. Zukoski, Properties of fire plumes, in: G. Cox (Ed.)Combustion Fundamental Fire, Academic Press, London, 1995, pp. 101–219 〈http://authors.library.caltech. edu/20911/〉. [34] J.L. Dupuy, J. Maréchal, D. Morvan, Fires from a cylindrical forest fuel burner: combustion dynamics and flame properties, Combust. Flame. 135 (2003) 65–76. [35] K. Zhou, N. Liu, J.S. Lozano, Y. Shan, B. Yao, K. Satoh, Effect of flow circulation on combustion dynamics of fire whirls, Proc. Combust. Inst. 34 (2013) 2617–2624. [36] J. Copelin, Image Captured of a Fire Whirl in Missouri, Photograph, viewed 18 November 2015, 2014, 〈http://www.news.com.au/world/firenado-thecoolestscariest-thing-according-to-janae-copelin-in-missouri/story-fndir2ev1226909647318〉 [37] Fire Whirl, Photograph, viewed 18 November 2015, n.d. 〈https://upload. wikimedia.org/wikipedia/commons/9/9f/Fire-whirl.jpg〉 [38] Fire Tornado, Fire Whirl Captured on Video, Photos, Photograph, viewed 18 November 2015, n.d. 〈http://geol105naturalhazards.voices.wooster.edu/files/ 2013/03/url-1.jpeg〉
34 (1979) 29–45. [10] F. Battaglia, R.G. Rehm, H.R. Baum, The fluid mechanics of fire whirls: An inviscid model, Phys. Fluids. 12 (2000) 2859. [11] K.H. Chuah, G. Kushida, The prediction of flame heights and flame shapes of small fire whirls, Proc. Combus. Inst. 31 (2007) 2599–2606. [12] F. Battaglia, K. McGrattan, R. Rehm, H. Baum, Simulating fire whirls, Combust. Theory Model. 4 (2000) 123–138. [13] R. Meroney, Fire whirls, fire tornados and firestorms: physical and numerical modeling, in: Proceedings of the PHYSMOD2003 Int. Work. Phys. Model. Flow Dispers. Phenom., Prato, Italy, 2003 [14] H.W. Emmons, S.J. Ying, The fire whirl, in: Proceedings of the Symposium Combustion 11, 1967, pp. 475–488 [15] R. Martin, D. Pendleton, W. Burgess, Effect of fire whirlwind formation on solid fuel burning rates, Fire Technol. 12 (1976) 33–40. [16] J. Lei, N. Liu, L. Zhang, H. Chen, L. Shu, P. Chen, et al., Experimental research on combustion dynamics of medium-scale fire whirl, Proc. Combust. Inst. 33 (2011) 2407–2415. [17] J. Lei, N. Liu, L. Zhang, Z. Deng, N.K. Akafuah, T. Li, et al., Burning rates of liquid fuels in fire whirls, Combust. Flame. 159 (2012) 2014–2114. [18] J. Lei, N. Liu, J.S. Lozano, L. Zhang, Z. Deng, K. Satoh, Experimental research on flame revolution and precession of fire whirls, Proc. Combust. Inst. 34 (2013) 2607–2615. [19] J. Lei, N. Liu, K. Satoh, Buoyant pool fires under imposed circulations before the formation of fire whirls, Proc. Combust. Inst. 35 (2015) 2503–2510. [20] K. Kuwana, K. Sekimoto, K. Saito, F.A. Williams, Scaling Fire Whirls, Fire Safety J 43 (2008) 252–257. [21] K. Kuwana, S. Morishita, R. Dobashi, K.H. Chuah, K. Saito, The burning rate's effect on the flame length of weak fire whirls, Proc. Combust. Inst. 33 (2011) 2425–2432. [22] K. Kuwana, K. Sekimoto, T. Minami, T. Tashiro, K. Saito, Scale-model experiments of moving fire whirl over a line fire, Proc. Combust. Inst. 34 (2) (2013) 2625–2631. [23] K.H. Chuah, K. Kuwana, K. Saito, F.A. Williams, Inclined fire whirls, Proc. Combust. Inst. 33 (2011) 2417–2424. [24] Y. Hayashi, K. Kuwana, R. Dobashi, Influence of vortex structure on fire whirl behavior, in: M. Spearpoint (Ed.), Fire Safety Science-Proceedings of the tenth International Symposium, International Association for Fire Safety Science, College
48
Contents lists available at ScienceDirect
Fire Safety Journal journal homepage: www.elsevier.com/locate/firesaf
Fire whirls in forest fires: An experimental analysis ⁎
Cláudia Pinto , Domingos Viegas, Miguel Almeida, Jorge Raposo
crossmark
Association for the Development of Industrial Aerodynamics, University of Coimbra, Rua Pedro Hispano 12, 3030-289 Coimbra, Portugal
A R T I C L E I N F O
A BS T RAC T
Keywords: Fire whirl Fire tornado Forest fires Wildfire Extreme fire behaviour Fire safety Forest fuels
This work presents a study on the formation of fire whirls with vertical axis on wildfires at laboratory scale. A particularity of the study is the use of typical forest fuels instead of fossil fuels as seen in some of previous studies on this topic. The forest fuels tested in the experiments were dead needles of Pinus pinaster, straw of Avena sativa, dead leaves of Eucalyptus globulus and a mix of shrubs mainly composed by heather (Erica australis) and gorse (Pterospartum tridentatum). The experimental results of the tests with and without forced flow inside a fire whirl generator were compared with tests in similar conditions out of the generator. It was possible to evaluate the effects of fuel bed size, bulk density and external vorticity on several parameters like flame height and diameter, mass decay and heat release rate. The results show that forced flow increases dramatically the burning rate and reduces the time needed to achieve a high rate of energy release. Comparison with results of other sources show that the flames that are generated in the present fire whirl generator are in a transition from fire whirl to pool fire regime and that it is possible to scale up some flow and thermal properties of field scale fire whirls and to derive predictive models on the basis of laboratory scale experiments.
1. Introduction
understand and predict this phenomenon [4]. For example, Alexander and Thomas [5] studied the conditions leading to the formation of a fire whirl that occurred during a fire near Santa Barbara, California, in March 1964. These authors stated that unstable atmosphere, high temperature at the ground level, a very low relative humidity, moderate winds and a large fuel load were defined as potential conditions for this phenomenon occurrence. In a general view, one of the main factors that contributes for the formation of fire whirls is the buoyancy force. Other studies with similar statements have been reported by several authors (e.g., [6–8]). The thermodynamic and kinematic characterization of fire behaviour regarding the formation of fire whirls has been the subject of several studies on analytical [9–11], numerical [12,13] and laboratory experiments [14–25] over the past few decades. The earliest reported experimental studies were conducted by Emmons and Ying [14] in 1967 using a rotating screen to generate vorticity on a liquid (acetone) pool fire with only 10 cm diameter. The fire whirl created consisted of a rotating cylinder fuel rich inside and lean outside flame and the authors concluded that the mass loss rate increased with the increase of ambient air circulation. Other experimental studies have been performed, most of them using liquid and gas fuels. One of the few reported experimental studies of fire whirls formation using solid fuels was carried out by Martin [15] who employed cross-piled wood sticks with different dimensions and moisture content values to measure the burning rates in a cylindrical device in which the tangential inlet of air
Forest fires are a natural disaster with great social, economic and environmental impact, making the study and understanding of fire behaviour essential for their successful management. In some circumstances forest fires can cause great devastation and endanger citizens and those who are in charge of suppressing them. It is recognized that situations in which fires spread under extreme weather and topography conditions are more likely to cause wide destruction and produce accidents [1,2]. Fire whirls are a phenomenon in which the fire under certain conditions acquires a vertical vorticity and forms a swirl or a flame column with vertical orientation that is also designated as a fire tornado [3]. They are characterized by a very high rate of energy release, unexpected formation and erratic movement, making them one of the most extraordinary and dangerous phenomena associated with fire behaviour [3]. In their erratic movement fire whirls can transport fire to areas far from the place where they were formed and they can also contribute to spread the fire by spotting due to embers that are lifted in their core to great heights and transported far from the fire perimeter by wind. Several studies have been conducted in order to understand better the conditions required for fire whirls to form and their role on the fire whirl properties. Reports describing the role of topography, fuels and weather on the formation of fire whirls, are important to better ⁎
Corresponding author. E-mail address: [email protected] (C. Pinto).
http://dx.doi.org/10.1016/j.firesaf.2016.11.004 Received 9 August 2016; Received in revised form 9 November 2016; Accepted 23 November 2016 0379-7112/ © 2016 Elsevier Ltd. All rights reserved.
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
density (kg m−3)
Nomenclature
ρ
d h m t v F H L M V ṁ Q̇
Subscripts
diameter (m) height of container (m) mass loss (kg) time (s) velocity (m s−1) fan frequency (Hz) low calorific power (kJ kg−1) length of the flame (m) dry mass (kg) volume (m3) mass loss rate (g s−1) heat release rate (kW)
o b c D f fu m M i in out s
Greek symbols β
initial bulk container chaparral dead fuel flame fuel medium maximum inlet inside outside solid particles
packing ratio (-)
lateral wind velocity and developed and validated a scaling law that predicts the critical wind velocity at which the most intense fire whirl was generated. It was observed too that in actual mass fires the occurrence of fire whirls is time-dependent, in contrast to wellcontrolled small-scale laboratory experiments, but this finding was not explored. In [22] it was verified that an appropriate scale model can successfully reconstruct the occurrence of fire whirls moving along a line fire and that a critical velocity was found to be proportional to the vertical buoyant velocity, which depends on the burning rate and length scale of the burning area. One point that was not discussed in that paper was the possibility of correlating the ratio between the critical velocity and the vertical buoyant force to the existence of a critical Froude number below which no fire whirls can be observed. It was stated in this study, but not confirmed, that a curved-shape geometry is necessary to generate moving type fire whirls. An interesting study was conducted by Chuah et al. [23] on the formation of inclined fire whirls, i.e., fire whirls for which buoyancy cannot be the source of support. Fire whirls with this configuration are observed frequently in forest fires. In this study an adjustable support for producing inclined fire whirls in the laboratory was placed on the base of the fire whirl generator. It was shown that the flame length was independent of the inclination angle at a given burning rate. The upper section of the apparatus was modified to observe the influence of flame height for a certain burning rate and it was concluded that for the same burning rate the flame height can be changed. So, if the structures of fire whirls are sufficiently strong, they will be dominated by the rotation plus entrainment, rather than by buoyancy. So, the several mechanisms that can explain the formation of fire whirls have been a focus point to understand and predict the occurrence of this phenomenon, as shown by the studies of [20–23] mentioned above or, for example, the study of [25] in which it is shown experimentally that the counter-rotating vortex pair is a possible origin of fire whirls that occur downwind of a fire area and are shed downwind. Although important research has been done to better understand the mechanisms and behaviour of fire whirls generation through theoretical or/and experimental studies, it is difficult to get results that are in accordance from different studies with similar analysis conditions (theoretical vs experimental for example), as shown by Hayashi et al. [24] that try to validate, reproducing in laboratory fire whirls at small scale with methanol, the relation between the flame height and the vortex structure that was studied by Chuah et al. [26]. In this work the Burke-Schumann-Burgers (BSB) model for small-scale fire whirls based on the assumption that vortex structures of fire whirls are able to be approximated by the Burgers vortex was developed. The results indicate that the vortex around the fire whirl cannot be
at the bottom could be controlled. They found that burning rates in cribs with fire whirls were 1.4–4.2 times greater than in cribs burning without whirl formation, although they reported that fuel particle density, bulk density, surface area to volume ratio, packing ratio and moisture content influenced the results, they did not study other parameters, such as flame height, that characterize and that are necessary to better understand this phenomenon. A remark that is made on the study of fire whirls by various authors is the lack of data in this area through experimental studies at laboratory scale. In recent years great importance was given to develop experimental equipment capable of further study on this topic. As an example we mention the extensive experimental work performed recently by Lei et al. [16–19]. The experimental apparatus used in their studies was a square enclosure with 2 m×2 m×15 m made of tempered glass and open on the top. The vertical channel had 20 cm wide vertical gaps at the corners to induce tangential air entrainment by the flame produced by liquid fuel. Different sizes of fuel pans (diameters 10–55 cm) were used and the liquid fuel was n-heptane (97%). In [16] the dynamic differences between fire whirls and pool fires were shown, and correlations of the burning rate, flame height and temperature of fire whirls were established. It was concluded that the pool diameter has a great influence on the burning rates of fire whirls, as in general pool fires, and that the transition from laminar to turbulent burning occurs when the pool diameter increases. The study of quasi-steady burning rate of laminar and turbulent fire whirls was done in [17] establishing correlations based on the boundary layer theory, film theory and Chilton-Colburn analogy. Due to the flame instability observed in the previous studies, Lei et al. [18] explored the characteristics of the flame of fire whirls distinguishing the flame revolution and the flame precession. Experimental observations have shown that in the formation of a fire whirl, the entire fire is tilted and revolves around its axis of symmetry with increasing angular velocity. Other example is the work done by Kuwana et al. [20–22] which gave a great contribution to the understanding of this phenomenon by reproducing experimentally at laboratory scale, fire whirls that occurred in real fires to study the mechanisms of formation of these fire whirls in different configurations. Examples are the fire whirl that occurred in Hifukushoato [20] in an open space and the case of fire whirls generated by the interaction between a line fire and the background wind as occurred in Brazil [22]. In [20] the Hifukushoato fire whirls were reconstructed at the 1/1000th scale using a large-scale wind tunnel and where lateral wind velocity was varied from 0.5 to 2 m/s to study its effect on fire-whirl formation. Different types of fire whirls formation in open areas where there was no combustible material were observed. The authors studied the influence of the 38
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
(Fig. 1a) the base of the FWG had a height of 1.3 m while in configuration C2 (Fig. 1b) the base height was raised to 1.8 m and a set of four axial fans was installed to create a forced tangential flow at the base of the combustion chamber. The FWG was installed in the large hall of the Forest Fire Research Laboratory that has a height of 11 m and an area of 72×16 m2. The vertical distance from the FWG exit to the ceiling was of 4 m and possibly a blocking effect exists in the vertical flow in the channel.
2.2. Fuels and containers In the tests reported in this paper four different fuels were used: (1) dead needles of Pinus pinaster (one test), (2) straw of Avena sativa (two tests), (3) dead leaves of Eucalyptus globulus (one test) and (4) a mix of shrubs mainly composed by heather (Erica australis) and gorse (Pterospartum tridentatum) (sixteen tests). These fuels were chosen as they are quite common in forested areas in Central Portugal and in several Mediterranean climate regions. They are also relatively easy to acquire in order to assure repeatability of the tests. The fuel was placed in a container at the base of the FWG in each test. The containers were made of a metallic grid welded to form a cylinder open on the top. Five different containers designated as A, B, C, D and E whose shapes are shown in Fig. 2 were used. The values of the diameter dc and of height hc of each container are given in Table 1. In each test the pre-defined amount of fuel was placed inside the container to form a solid porous fuel with a bulk density ρb given by:
Fig. 1. A schematic illustration of the different configurations used on present study: a) configuration C1; b) configuration C2.
approximated by the Burgers vortex, probably because of the complexity of the flow associated with density and temperature changes near the flame. It is found that sometimes theoretical models cannot be applied correctly to the experimental data showing the importance of having laboratory experiments to explain and validate fire whirl behaviour. It is therefore important to reproduce these phenomena at a scale closer to the real scale. Using image analysis, it is important to determine parameters from real fire whirls such as shape and size of their base and flame height to develop models to predict fire whirl properties at various scales. The present work aims contributing to a better understanding on the formation of fire whirls and characterization of this phenomenon through the operation of a novel equipment the Fire Whirl Generator (FWG), and the development of a test methodology to analyse several fire whirl parameters, including the rate of fuel mass burning, the energy release rate, the flame height, the flame diameter and the influence of fuel bed dimension, bulk density and forced air inlet velocities at the base of the FWG. A particularity of this study is the use of typical forest fuels instead of fossil fuels as was done in some of the previous studies on this topic. It is important to analyse and to understand the differences, advantages and limitations of using solid fuels like those that are found in forest fires in relation to the use of gaseous or liquid fuels. We also intend to characterize differences associated to the fuel bed properties namely its dimension and porosity and of the boundary conditions namely the confinement of the flame with or without an external tangential flow. Comparison with results from other laboratory scale experiments and full size fire whirls and pool fires showed that it is possible to derive predictive models for this type of fire whirls in a wide range of scales.
Mfu
ρb =
Vfu
(1)
In this equation Mfu is the dry mass of the fuel that was estimated by the previous measurement of the moisture content of the fuel (FMC). The value of FMC was measured for one 1.5 g sample of fuel using a moisture analyser A & D MX-50 (0.01 g) immediately before the experiment. The value of Mfu was changed in some experiments but in all other tests reported here its value was kept equal to 4.0 kg. The volume of the fuel Vfu is given by the following equation:
Vfu =
π . dc 2. hfu 4
(2)
In Eq. (2), the height hfu of the fuel inside the container did not always coincide with the height of the container. For example, when a load of 4.0 kg of shrub was placed in container B the value of hfu was around 0.55 m. Typical values of bulk density for tests with shrub fuel with a load of 4.0 kg for each container are given in Table 1. We consider the packing ratio β defined as usual [28] by:
2. Experimental methodology
ρb
β=
2.1. Fire whirl generator
ρs
(3)
In this work we considered that the volumetric mass density ρs of the solid material had an average value of 700 kg/m3. The bulk density can be defined also as the ratio between the volume Vs occupied by the solid material and the total volume Vfu occupied by the fuel:
Following the work of Lei et al. [16–19] and Satoh et al. [27] we designed a Fire Whirl Generator (FWG) apparatus consisting of a vertical channel with a quadrangular section of 1×1 m2 with a height of 6 m with two sides made of tempered glass and the other two made of steel sheet. The base of the FWG has a section of 2×2 m2 and 1.8 m height tapering to the 1×1 m2 section of the main channel section that is open at the top. Each corner of the channel has a vertical opening 10 cm wide to induce tangential air entrainment. Inside the base of FWG there is a plate of 1×1 m2 covered by ceramic tiles to support the fuel container. The FWG is a new equipment that in its development was subject to several modifications that involved some changes of its configuration during the research program. In this paper we refer to two configurations that are shown schematically in Fig. 1. In configuration C1
β=
Vs Vfu
(4)
The value of Vs is equal to:
Vs =
39
Mf ρs
(5)
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 2. Containers in which the fuel was placed.
2.3. Forced flow system
Table 1 Designation and dimensions of the containers used. Container
hc (m)
dc (m)
ρb (kg m−3)
A B C D E
0.760 0.760 1.160 0.760 0.305
0.350 0.500 0.500 0.800 1.120
54.70 26.81 17.56 14.47 13.31
As was described above a set of four axial fans Rosenberg DR 560-4 of 0.81 kW was attached to the base of the FWG to induce a tangential flow to create a forced vortex inside the combustion chamber. Each fan had a duct that made the transition from its circular section (∅ 560 mm) to a rectangular section (200×300 mm2). The fans were controlled by a frequency driver Omron Sysdrive 3G3HV Inverter 3G3HV-A4037-CE, in a range between 0 Hz and 50 Hz. The average flow velocity vi at the inlet of the combustion chamber was measured with a hot wire anemometer Multifunction KIMO AMI 300. A linear relationship between vi (m/s) and the electrical frequency F (Hz) of the fans given by Eq. (6) was found (R2=0.997).
vi=0. 0995F −0. 240
(6)
In the present experiments the following frequency values were used (Hz) 0, 5, 10, 20, 30, 40 and 50, corresponding to inlet velocities of (m/s) 0, 0.31, 0.78, 1.73, 2.74, 3.86 and 4.83, respectively. 2.4. Sensors and data acquisition Fig. 3. A schematic illustration of the system used to measure the evolution of the fuel mass during the combustion process.
2.4.1. Mass loss and heat release rates The mass of the fuel in the container was measured continuously during each test using the system shown in Fig. 3. It consists of a platform on which the container is placed with a lever to allow the
Table 2 Parameters of the experiments reported in this work. Test
FMC (%)
Fuel
Cont.
In/Out test
vi (m s−1)
ṁM (g s−1)
̇ (kW) QM
LM (m)
dfm (m)
GV15 GV16 GV19 GV20 GV23 GV25 GV26 GV27 GV29 GV31 GV32 GVV6 GVV15 GVV17 GVV20 GVV23 GVV25 GVV30 GVV36 GVV37 GVV38 GVV50 GVV52 GVV54
18.2 18.2 15.1 18.5 18.0 16.0 16.0 16.0 16.0 14.4 14.4 11.4 13.6 13.6 11.3 11.3 12.5 10.9 8.4 10.2 10.2 10.5 10.5 10.5
Shrub Shrub P. needles Straw Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Shrub Eucalyptus Shrub Shrub Shrub Shrub Shrub
B A B B D D C C C E E B B B B B B B B A D B A D
In In In In In Out In Out Out In Out In In In In In In In In In In Out Out Out
– – – – – – – – – – – 0.78 2.74 3.86 1.73 4.83 – 0.31 – 2.74 2.74 – – –
32.13 22.47 28.59 29.37 47.97 67.87 50.13 38.37 39.92 85.60 79.80 43.96 67.39 58.41 46.88 63.48 50.06 41.67 19.56 46.32 78.31 – 15.68 76.97
540.3 356.4 386.0 472.6 811.6 1180.0 871.8 667.3 694.2 1521.0 1419.0 916.6 1400.9 1214.1 977.8 1323.9 1042.2 869.5 489.6 869.0 1635.9 – 350.4 1720.0
4.90 4.13 4.53 2.60 6.13 4.32 5.00 4.65 4.31 4.58 4.13 5.26 5.56 5.46 5.08 5.08 4.62 4.92 3.52 3.10 3.51 3.84 – –
0.45 0.38 0.45 0.46 0.59 0.64 0.55 0.53 0.61 0.66 0.69 0.48 0.38 0.39 0.44 0.46 0.68 0.59 0.52 0.41 0.36 0.33 0.26 0.41
40
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 4. Typical images taken by the photographic camera for five different phases of GV27 test made outside of FWG.
container C in three different conditions. In Fig. 4 five photos taken during test GV27 in which the container was placed out of the FWG are shown. The time elapsed from ignition is shown in each photo. In Fig. 5 photos of test GV 26 made with the same basket and fuel conditions but inside the FWG in its configuration C1 (Fig. 1a) are shown. In test GVV30 that is shown in Fig. 6 forced ventilation vi=0.31 m s−1 was used. The results of these tests are detailed below.
balance to be far from the heat released during the combustion process. A digital balance A & D HW-100KGL with a resolution of 10 g with a frequency of 1 Hz, by RSKey v.1.40 software, connected to a laptop was used for this purpose. This system was duplicated in order to perform tests with similar fuel containers in a quiescent ambient out of the FWG to check the effect of the boundary conditions. 2.4.2. Flame geometry analysis In each test at least two cameras were used to record images: a video camera Sony DCR-SR87 and a digital camera Canon EOS 550D. Both cameras were located at fixed places close to each other to record images in order to measure flame diameter and flame height. The flame height L and flame diameter df of the fire whirl were obtained through image processing of frames selected with a time interval of five seconds using Photoshop CC software. The flame height L from the base of the container to its visual tip was measured manually in each image using a reference length as a vertical scale. The diameter df of the fire whirl was measured in each time step with the same method at three selected heights and an average value was taken to characterize it. The highest values of the flame height LM in the series of photos for each test and the average value of the fire whirl diameter dfm were two of the parameters analysed for each test as it is summarised in Table 2. In this Table each test has a designation that is composed by a set of letters (GV or GVV) related to the test configuration and a number corresponding to chronological order. The designation GV corresponds to tests made with configuration C1 (Fig. 1a) and also to tests with the fuel outside the FWG and GVV corresponds to tests made inside the FWG with configuration C2 (Fig. 1b).
2.5.2. Mass loss analysis The mass of the fuel in the container was measured during each test with a frequency of 1 Hz and the values were corrected to a dry basis adjustment. In order to smooth the data average values at five seconds intervals were calculated sequentially along the test. The mass loss rate, ṁ, is defined here as:
ṁ =
dm dt
(7)
Its value was determined over the duration of the test at five-second intervals. In the evaluation of the maximum value ṁM of the mass loss rate it was found that in some tests this value was well defined while in others there was not a single maximum value but a series of similar values. In the latter cases the maximum rate of mass loss ṁM was taken as the average of the five higher values of ṁ in the test. The values of heat release rate Q̇ (kilowatt) were calculated using the following equation:
Q̇ = ṁ × Hfu
(8)
In this equation the mass loss rate is expressed in kg/s and Hfu is the low calorific power (kJ/kg) of the fuel used. The value of Hfu was computed using CEN EN 14918:2009 [29] and data from Lopes [30]. In Table 3 the calculated values of Hfu for each fuel are given. A typical value of 20 MJ/kg was considered in most calculations. When using Eq. (8) we are assuming that the combustion efficiency was 100% for each test, although this is not valid as some particles are released without burning completely. As the same value of Hfu was considered in each test the mass and energy release rates are proportional. Therefore, we will refer to either of them to characterize the intensity of the fire whirl. For each test the ̇ of the heat release rate was evaluated in the same maximum value QM form as was done for ṁM. The mass loss and mass loss rate values for a typical test are shown
2.5. Experimental procedure and data analysis 2.5.1. Test procedure In each test the predefined amount of fuel was collected taking into account its moisture content and packed in the container. The container was placed on the platform inside the FWG, the ignition was initiated along the perimeter at the basis of the container using a gas burner and the fans were turned on and adjusted to achieve the desired flow velocity. When the ignition was carried out, the video and photo cameras were started and the measurements initiated. Typical images of three tests GV27, GV26 and GVV30 are shown respectively in Figs. 4, 5 and 6. These tests were performed using
Fig. 5. Typical photo images for five different phases of GV26 test.
41
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 6. Typical photo images for five different phases of GVV30 test.
around the central axis due to the interaction of the ascendant flow and the entrained air from the four openings of the FWG, which indicates the formation of a fire whirl.
Table 3 Characteristic parameters of different fuels. Test
Fuel
Moisture content (%)
Test duration (s)
Flame duration (s)
Hfu (MJ/kg)
GV15 GV19 GV20 GVV36
Shrubs P. needles Straw Eucalyptus
18.2 15.1 18.5 8.4
230 300 260 230
210 300 70 230
16.9 13.5 11.7 22.5
2.5.3. Flame height and flame diameter The flame height or length L and flame diameter df of fire whirls are important parameters when studying fire whirls. They have a very unique visual aspect in comparison to ordinary flames, characterized by a relatively large height and small flame diameter. These features are relatively easy to obtain in the field during experimental and real fires and are very often registered in still and video pictures. It is therefore of great interest to be able to establish relationships between fire whirl height and diameter with other combustion process parameters that support the fire whirl like for example the fuel mass consumption rate or the equivalent heat release rate. A very common relationship that is used in the literature (e.g., [31,34]) is the following:
L = a × Q̇
b
(9)
In this equation L is the flame height and a and b are two model parameters. In the present study, according to Eq. (8) Q̇ is considered to be proportional to ṁ , therefore we will use as well the parameter ṁ to analyse the relationship between the flame height and the mass loss rate. In each experiment the flame diameter varies between a relatively narrow range as it is shown in Fig. 8 for test GV16. For this reason, we will characterize the flame diameter by its average value dm. Given the transient nature of the present experiments in the results and discussion section we shall analyse the time evolution of some parameters when they are correlated with each other in the course of time. But for other parameters like the mass loss, heat release rate or flame height, average or extreme values will be used to characterize the role of some factors and to produce models.
Fig. 7. Instantaneous results for mass loss and mass loss rate for GV31 test.
3. Results 3.1. Role of fuels Fig. 8. Evolution of the flame diameter for GV16 as a function of time and its average value dm.
In order to determine which fuel would be more appropriate to perform the test program preliminary experiments were carried out using various types of surface fuels namely shrubs, dead pine needles, eucalyptus leaves and straw, that are commonly found in forested areas. The tests were made using container B and assuring the same conditions of ambient temperature and preparation methodology, and their parameters are summarised in Table 3. In these tests the fuel load was 4 kg for all fuels. In Fig. 9a and b the corresponding values of flame length and mass loss rate are plotted. In Fig. 9a it is possible to observe that the flames of the tests with straw are very short and last for less than one minute. The other three fuels have similar values of flame length and flame duration but shrubs have higher values of maximum mass loss rate. For this reason and due to the fact that shrubs are possibly more representative of the type of fuels in which fire whirls are more likely to occur we decided to select shrubs as the basic fuel for our test program.
in Fig. 7 in which the initial mass loading was mo=4.0 kg. The instantaneous values of mass at one second intervals are shown as well as the smoothed curve with 5 s average values. The mass loss curve corresponds to average values at 5 s intervals. As reported by Sun et al. [31] the combustion process has three phases. In the initial phase that lasts for 25 s in this test, there was a slight loss in mass corresponding to the loss of moisture and start of combustion; in the second phase (lasting from 25 to 75 s there is a marked decrease of mass corresponding to flaming combustion of the fuel; the third phase corresponds to combustion of thicker particles of the fuel partially in non-flaming combustion. During the second phase in which the maximum value of mass loss rate is observed the flame initially looks like a general fire burning without vorticity but after some time the flame starts to rotate 42
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
a5
b
40
Shrubs Dead pine needles 4
Shrubs
35
Dead pine needles
Straw
30
Eucalyptus leaves 3
Straw Eucalipto
L (m)
ṁ (g.s-1)
25
2
20
15 10
1
5 0 0
50
100
150
200
250
300
0 0
t (s)
50
100
150
200
250
300
t (s)
Fig. 9. Results for different types of fuels: a) flame height of fire whirl; b) mass loss rates.
be limited by the oxygen supply. It is interesting to notice that Rothermel [28] indicates that the potential reaction velocity of excelsior combustion decreases for packing ratios greater than approximately 0.015. The same trend is observed in our work as maximum mass loss rate decreases for packing ratios greater than 0.015. In order to separate these two effects tests with constant values of β in different containers should be performed to better assess the effect of the size of the fuel bed. The average values of flame diameter for tests made inside and outside of FWG are shown in Fig. 12 as a function of container diameter. The results follow the trends shown in Eqs. (10) and (11) with a correlation coefficient equal to 0.97 and 0.89, respectively for tests inside and outside the FWG:
Table 4 Description of the fuel inside the containers. Test
Container
Packing ratio β
GV16 GV15 GV23 GV31
A B D E
0.078 0.038 0.021 0.019
3.2. Role of containers The parameters of tests GV16, GV15, GV23 and GV31 using the different types of containers are given in Table 4. The corresponding results are shown in Figs. 10 and 11. As can be seen in Figs. 10a and 11a, larger diameter containers, dc, correspond to higher values of ṁ and lower values of burning duration. These results are in agreement with other author's studies [15,16]. In Fig. 10b the values are in accordance with Fig. 10a, except for container E (dc=1.12 m), which corresponds to a lower value of flame height in relation to containers B (dc=0.50 m) and D (dc=0.80 m). This lower value could be related to the small height of the container and to its large diameter in relation to the dimensions of table base (1×1 m2). We recognize that in these tests two different factors are affecting our results. One is the container diameter and the other is the packing ration β, that is a measure of the fraction of the fuel array volume occupied by the fuel, and expresses the relationship between the fuel particles and air existing between them. In Table 4 the fuel packing ratio values for each container are given. The maximum mass loss rate values ṁM for these tests are shown in Fig. 11a and b as a function respectively of the container diameter and the packing ratio. It can be seen that ṁM increases with dc and decreases with β. This results can be related to the oxygen supply, which means that for higher values of β and for smaller dc, the ṁM can
d f =0. 42.dc 0.52
(10)
d f =0. 41.dc 0.39
(11)
3.3. Role of forced flow A set of tests (GVV25, GVV30, GVV6, GVV20, GVV15, GVV17 and GVV23) were performed inside the FWG with configuration C1 and container B to analyse the influence of forced flow. In these tests the value zero for the inlet velocity of the forced flow means that the fans were turned off, i.e., the fire whirl was formed due to the natural flow. As can be seen in Fig. 13a and b the maximum value of mass loss rate ṁM increases generally with the forced flow velocity, at the same time the average flame height increases a little (+11%) and the average flame diameter decreases significantly (−33%). The consequence of this is that the surface of the external cylinder enveloping the flame decreases. The apparent contradiction between these two observations,
Fig. 10. Instantaneous results for different diameter containers: a) mass loss rate variation; b) flame height variation.
43
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 11. Results for maximum mass loss rate as a function of: a) diameters of containers; b) fuelbed compactness.
Fig. 12. Flame diameter as a function of container diameter for tests inside and outside the FWG.
Fig. 14. The evolution of mass loss rate and flame height in the course of time for GV16 test.
could be explained by the fact that the flame surface is wrinkled because of the rotational flow, increasing in the same manner the flame surface. By retroaction the heat transfer by radiation (proportional to the flame surface) between the flame and the solid fuel increases contributing to an acceleration of the degradation of the solid fuel. Two exceptions are observed for vi=0 m s−1 and for vin=2.74 m s−1. There is a decrease of the mass loss rate when vi varies between 0 and 0.31 m s−1; the value of ṁ for vi=2.74 m s−1 is actually an absolute maximum and it decreases for vi > 2.74 m s−1. The same effect is shown in Fig. 13b in which we can see that the flame length and diameter modify their trend of variation for vi > 2.74 m s−1. This effect must be studied more carefully as it may be an indication of the existence of either a change of regime in the fire whirls development or the presence of a stagnant flow near the top of the FWG due to the presence of the ceiling constraining the development of the fire whirl.
3.4. Flame height prediction In Fig. 14 the evolution of flame height and mass loss rate (that is proportional to Q̇ ) as a function of time for test GV16 is shown. As can be seen in this figure there are two different phases in the process: (1) the growth phase in which the value of ṁ and of L increase practically at the same time; (2) the decay phase in which ṁ decreases but L decreases less rapidly. From this observation we can conclude that there is not a bi-univocal relationship between L and ṁ and two separate laws can be established for phases 1 and 2, respectively. This is illustrated in Fig. 15a and b in which the data points corresponding to phases 1 and 2 of this test are shown. The corresponding values of ai and bi are: a1=1.09 and b1=0.34; a2=1.61 and b2=0.21. Similar results were found in the other tests,
)
Fig. 13. Results for different air forced flow: a) maximum values of mass loss rate; b) maximum values of height flame and diameter flame of fire whirl.
44
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 15. A schematic illustration of the different phases that correlate the flame height with mass loss rate related to GV16: a) the growth phase (phase 1); b) the decay phase (phase 2).
distinguishing phases 1 and 2. The corresponding result for test GV16 is shown in Fig. 16. The power law fitting yielded the following results: ao=1.29 and bo=0.29. Unless stated otherwise in the following discussion the results are presented in the form of ao and bo. In Table 5 the values of ao, a1, a2 and of bo, b1, b2 for the relevant tests of the present study are shown. 3.5. Role of boundary conditions Flame height and heat release rate are two variables that are strongly dependent on each other, Zukoski [33], cited by Sun et al.[31]. Sun et al.[31] correlated the flame height with heat release rate for dead chaparral fuels using the following equation: 2/5 LD=0. 20QḊ
Fig. 16. A schematic illustration of the relation between flame height and mass loss rate for GV16 test.
They compared it with correlations from earlier studies for ovendried fuels and concluded that the two-fifths power fits well for dead or dry fuels. Dupuy [34] studied combustion dynamics and flames properties through experimental measurements to test models of a forest fire to analyse the variations of the maximum flame height with maximum heat release rate for forest fuels (Pinus pinaster needles and excelsior), these values are shown in Fig. 17a. In the present study we performed tests with the fuel outside the FWG and others inside the FWG in configuration C1 without forced flow. In the second case we had fire whirls due to the boundary conditions while in the first case a fire whirl was not formed. The results of the flame height of these tests are shown in Fig. 17a and b, respectively, and the following correlations were found for each set of tests:
Table 5 Coefficients of Eq. (9) relating Flame Height with mass loss rate. Test
a0
b0
R02
a1
b1
R12
a2
b2
R22
GV15 GV16 GV19 GV20 GV23 GV25 GV26 GV27 GV29 GV31 GV32 GVV6 GVV15 GVV17 GVV20 GVV23 GVV25 GVV30 GVV36
1.73 1.30 1.53 1.24 1.26 0.67 1.09 0.72 0.65 0.21 0.18 1.66 1.15 1.15 1.81 1.92 1.06 0.58 2.10
0.21 0.29 0.28 0.03 0.31 0.44 0.33 0.48 0.47 0.66 0.72 0.21 0.35 0.37 0.20 0.17 0.36 0.58 0.12
0.61 0.72 0.72 0.003 0.68 0.90 0.68 0.86 0.88 0.82 0.85 0.42 0.53 0.86 0.23 0.40 0.83 0.81 0.45
0.70 1.09 0.19 1.26 1.11 1.43 0.60 0.58 0.69 0.03 1.04 1.82 1.84 1.50 2.64 1.64 1.26 0.40 0.67
0.39 0.34 0.81 0.25 0.35 0.23 0.48 0.52 0.44 1.17 0.28 0.23 0.26 0.31 0.15 0.30 0.33 0.66 0.44
0.73 0.48 0.25 0.90 0.39 0.70 0.90 0.98 0.96 0.91 0.68 0.47 0.52 0.92 0.23 0.75 0.87 0.86 0.76
1.95 1.61 1.46 0.08 1.31 0.68 1.17 1.09 0.61 0.27 0.22 1.55 0.63 0.84 0.96 2.03 0.97 2.73 1.37
0.21 0.21 0.31 0.86 0.27 0.46 0.33 0.34 0.49 0.59 0.63 0.20 0.52 0.46 0.38 0.12 0.38 0.12 0.31
0.60 0.49 0.77 0.90 0.88 0.91 0.95 0.93 0.76 0.89 0.96 0.78 0.91 0.96 0.76 0.34 0.84 0.54 0.70
(12)
0.46 Lout =0. 18Q̇out
(13)
̇ 0.38 Lin=0. 35Qin
(14)
Eq. (13) corresponds to the flame height outside the FWG as a function of the heat release rate while Eq. (14) corresponds to the tests inside the FWG in which fire whirls were formed. In Fig. 17a it is shown that our correlation fits well with the correlation of Dupuy [34] and that both are greater than the results obtained by Sun [31]. This difference may be due to the combustion regime; both ours and Dupuy experiments are not steady combustion regimes while the study of Sun [31] correspond to steady state. As can be seen in this figure our experiments cover a much wider range of heat release rates in comparison to the previous two studies. In Fig. 17b it can be seen that the results for fire whirls yield higher values of L for the same value of Q̇ in relation to test performed out of the FWG. The relative confinement of the FWG can contribute to facilitate the development of the vertical flow and also the vertical openings induce tangential flow that enhances the fire whirl effect putting in evidence the role of boundary conditions in the flame development.
with b1 close to 0.4 and b2close to 0.3. Weise et al. [32] made tests with different shrub fuels and found that in the general case the growth of flame height had a time lag in relation to the ṁ growth. This time lag decreased with the dryness of the fuel; in the case of dead fuels the time lag was practically equal to zero as we found in the present experiments. When we observe an image of a fire whirl it is not easy to assess if it is in phase 1 or 2, therefore for practical purposes we will develop an overall relationship between L and ṁ for the entire test without 45
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Fig. 17. Flame height versus heat release rate: a) for tests made outside the FWG b) for tests made inside the FWG without ventilation (configuration C1).
3.6. Comparison with full scale fire whirls
Table 6 Parameters of the tests shown in Fig. 18a and b.
Using the present research, a relationship between the maximum heat release rate and the diameter of the fuel container is proposed in the form of Eq. (15). A complementary relationship between the maximum values of the flame height and the heat release rate is proposed in Eq. (16) that is similar to Eqs. (12)–(14).
̇ = a3dcb3 QM
(15)
b
̇4 LM = a4QM
(16)
Our results together with those found in the literature are shown in Fig. 18a and b for these two sets of parameters. The numbering used in graphs 18a and 18b is described in Table 3. Results from tests GV27, GV29, GV25 e GV32 performed inside the FWG are fitted by curve 1 in both figures; results from tests GV27, GV29, GV25 e GV32 performed outside the FWG correspond to curve 2 and results from tests GVV38, GVV15 e GVV37 with a forced flow of vi=2.74 m s−1 correspond to curve 3. Results from fire whirls with liquid fuel (heptane) obtained by other authors are shown in this figure and fitted by curves 4 and 5; results from pool fires also with heptane are fitted by curves 6, 7 and 8 in those figures. As can be seen in these figures the reported tests cover a wide range of scales of fuel size, flame height and heat release rate. Our results are in the middle of this range and seem to be in a transition between small laboratory scale tests and large scale field tests. ̇ and L As can be seen in these figures the results for both QM corresponding to pool fires are usually higher than those for fire whirls. The results of our tests outside the FWG are similar to pool fires while those inside the FWG are in a transition from fire whirls to pool fires. The tests with forced flow are more similar to fire whirls showing the
Ref.
Description
a3
b3
R32
a4
b4
R42
[1] [2] [3] [4] [5] [6] [7] [8]
Our study (inside FWG) Our study (outside FWG) Our study (forced flow) Lei et al. [16] Zhou et al. [35] Klassen and Gore (1994)a Zhou et al. [35]b Kosekia
1213 1339 2047 3307 3224 2295 1211 2401
1.19 0.95 0.74 2.25 2.20 2.41 2.52 2.13
0.98 0.97 0.87 0.99 0.99 0.99 0.99 0.99
2.82 8.61 0.17 0.75 1.49 0.22 0.21 –
0.08 −0.10 0.44 0.28 0.18 0.35 0.36 –
0.10 0.60 0.22 0.98 0.97 0.99 0.97 –
a b
Values from Table 2 of Lei et al.[16]. Values from Table 2 of Zhou et al.[35].
Fig. 19. Formation of a fire whirl during field experiments at Gestosa.
8
3 7
1 5
2
4
6
Fig. 18. Results of different studies: a) maximum heat release for different diameters of containers b) maximum heat flame versus maximum heat release. The sources of data that are indicated by a number are given in Table 6.
46
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
result indicates that it is possible to determine predictive models to estimate the properties of fire whirls on the basis of laboratory scale tests. The values of coefficients a5 and b5 for the series of data points that are shown in Fig. 20 are given in Table 7. We see also that the flames that are formed in our FWG are in the transition from a fire whirl to a free flame regime. In order to overcome this difficulty, we intend to increase the forced flow velocity or to create an opening on the ceiling of the Laboratory to eliminate the stagnation effect that the presence of the ceiling imposes on the flow. 4. Conclusions A laboratory scale Fire Whirl Generator using forest type fuels was developed and the proposed experimental methodology to analyse the properties of fire whirls was described. It is possible to study systematically the role of various factors like the fuel type, the size and porosity of the fuel bed, the presence of enclosing walls and the effect of forced flow on various parameters that can be used to characterize fire whirls like for example the mass loss or heat release rates, the flame length and diameter. The fuel used in these experiments was dried shrub and the maximum heat release power was of the order of 1 MW which is higher than the reported values for similar laboratory tests. Given the transient nature of the experiments relationships between correlated parameters or with average or extreme values were established. Comparison with similar experiments showed that the properties of the flames generated with the present configuration of the FWG are similar to fire whirls produced in other laboratory studies and with full scale fire whirls. It was found nevertheless that further changes must be explored in further work to possibly minimise the effect of the flow stagnation that is produced by the presence of the ceiling. The effect of fuel bulk density must be studied more systematically as well. Further work will be performed to explore the potential of the FWG extending the range of parameters tested so far using other sensors to analyse the flow and temperature field inside the flames which will provide us data to better analyse and understand the physical mechanisms associated to the formation and to the complex dynamics of fire whirls.
Fig. 20. Results of different studies of maximum height flame versus diameter of containers or the diameter of the base of the fire whirls. The sources of data that are indicated by a number are given in Table 7. Table 7 Parameters of the tests shown in Fig. 20. Ref.
Description
a5
b5
R52
[1] [2] [3] [4] [5] [6] [7] [8]
Our study (inside FWG) Our study (outside FWG) Our study (forced flow) Lei et al. [16] Zhou et al. [35] Klassen and Gore (1994)a Zhou et al. [35] [Table 2]b Real fire whirls
5.11 4.19 4.86 7.48 7.48 3.48 2.72 1.09
0.10 −0.10 0.33 0.64 0.44 0.86 0.90 1.06
0.20 0.60 0.02 0.96 0.97 0.99 0.96 0.98
a b
Values from Table 2 of Lei et al.[16]. Values from Table 2 of Zhou et al.[35].
importance of the boundary conditions namely the external vorticity and the dimension of the fuel bed. These results show that in our experiments there is a transition from fire whirl to pool fire regime. The values of coefficients a3, b3, a4 and b4 of Eqs. (15) and (16) are given in Table 6. The above results allow us to propose a relationship between the maximum flame height and the size of the fuel bed given by:
LM =
a5dcb5
Acknowledgements
(17)
The collaboration given to the experimental program that supports this paper by several collaborators – namely, by Mr. N. Luís, Mr. R. Oliveira, Mr. A. Cardoso, Mr. J. Lourenço, Mr. D. Lopes and Miss Ana Rosário – is gratefully acknowledged. The first author is thankful to IEFP and to ADAI for the professional internship. The authors are grateful for the referees' helpful comments.
It is easy to see that:
a5 = a4 × a3b4
(18)
b5 = b3 × b4
(19)
Photos or video images of fire whirls can be used to obtain data on the two parameters that are used in Eq. (17). Provided that there is some scale factor to assess vertical and horizontal dimensions and assuming that the flaming area at the base of the fire whirl – that is usually a circle – can be assimilated to the fuel bed or fuel container in the laboratory experiments we can check the validity of Eq. (17) to determine the properties of real scale fire whirls based on laboratory scale experiments. We analysed the images of nine full scale fire whirls (e.g., [36–38]) for which both the flame height and the diameter of the base of the fire whirl could be measured. As an example, Fig. 19 shows one of the images analysed from a test performed by our team in Gestosa (Portugal). The dimensions of the base and the height of the fire whirl are obtained through a relationship with the known dimensions of the fire truck. These data points are shown in Fig. 20 together with the results from laboratory experiments. The data numbering used in graph 20 is described in Table 7. It is interesting to notice that in spite of the fact that the dimensions of full size fire whirls extend the range of scales in at least one order of magnitude they follow the same trend of the laboratory experiments but with lower values of L. This
References [1] D. Viegas, Extreme fire behaviour, in: A.C. Cruz, R.E. Correa (Eds.), For Management Technology Practice Impact, Nova Science Publishers Inc., New York, 2012, pp. 1–56. [2] D.X. Viegas, A.R. Figueiredo, M.A. Almeida, V. Reva, L.M. Ribeiro, M.T. Viegas, et al., Relatório do Incêndio Florestal de Tavira/São Brás de Alportel, 2012 [3] J.M. Forthofer, S.L. Goodrick, Review of vortices in wildland fire, J. Combust. 2011 (2011) 1–14. http://dx.doi.org/10.1155/2011/984363. [4] A.R. Pirsko, L.M. Sergius, C.W. Hickerson, Causes and Behavior of a Tornadic FireWhirlwind, USDA For. Serv. Pacific Southwest For. Range Exp. Station, Berkeley, Calif, USA, 1965, pp. 1–13. [5] M.E. Alexander, D.A. Thomas, Wildland fire behavior case studies and analyses: value, approaches, and practical uses, Fire Manag. Today 63 (3) (2003) 4–7. [6] C.M. Countryman, Fire whirls…Why, When and Where, USFS Pacific Southwest Res. Stn., Berkeley, Calif, USA, 1971. [7] S. Soma, K. Saito, Reconstruction of fire whirls using scale models, Combust. Flame. 86 (1991) 269–284. [8] J. Forthofer, B. Butler, Large scale fire whirls: Can their formation be predicted?, in: Proceedings of the 3rd Fire Behavior Fuels Conference October 25–29, Spokane, Washington, USA, 2010, pp.1–7 [9] A. Muraszew, J.B. Fedele, W.C. Kuby, The fire whirl phenomenon, Combust. Flame.
47
Fire Safety Journal 87 (2017) 37–48
C. Pinto et al.
Park, MD, 2011, pp. 671 [25] M. Shinohara, S. Matsushima, Formation of fire whirls: Experimental verification that a counter-rotating vortex pair is a possible origin of fire whirls, Fire Saf. J 54 (2012) 144–153. [26] K.H. Chuah, K. Kuwana, K. Saito, Modeling a fire whirl generated over a 5 cm diameter methanol pool fire, Combust. Flame 156 (2009) 1828–1833. [27] K. Satoh, K. Yang, A horizontal fire-whirl design scenario for engineering performance-based fire-code applications, Int. J. Eng. Performance-Based Fire Codes 2 (2000) 48–57. [28] R.C. Rothermel, A mathematical model for predicting fire spread on wildland fire spread in wildland fields, USDA For. Serv. Res. Pap. INT-115, 1972, pp. 40 [29] CEN EN 14918:2009, Solid Biofuels Determination of Calorific Values. [30] A. Lopes, Misturas de Biomassa. Tese de Mestrado em Engenharia Mecânica na Especialidade de Energia e Ambiente, Universidade de Coimbra, Coimbra, 2014. [31] L. Sun, X. Zhou, S. Mahalingam, D.R. Weise, Comparison of burning characteristics of live and dead chaparral fuels, Combust. Flame. 144 (2006) 349–359. [32] D.R. Weise, T. Fletcher, S. Mahalingman, X. Zhou, S. Smith, L. Sun, Correlation of mass loss rate and flame height for live fuels, in: Proceedings of the Sixth Symp. Fire and For. Meteorol., Canada, 2005 [33] E.E. Zukoski, Properties of fire plumes, in: G. Cox (Ed.)Combustion Fundamental Fire, Academic Press, London, 1995, pp. 101–219 〈http://authors.library.caltech. edu/20911/〉. [34] J.L. Dupuy, J. Maréchal, D. Morvan, Fires from a cylindrical forest fuel burner: combustion dynamics and flame properties, Combust. Flame. 135 (2003) 65–76. [35] K. Zhou, N. Liu, J.S. Lozano, Y. Shan, B. Yao, K. Satoh, Effect of flow circulation on combustion dynamics of fire whirls, Proc. Combust. Inst. 34 (2013) 2617–2624. [36] J. Copelin, Image Captured of a Fire Whirl in Missouri, Photograph, viewed 18 November 2015, 2014, 〈http://www.news.com.au/world/firenado-thecoolestscariest-thing-according-to-janae-copelin-in-missouri/story-fndir2ev1226909647318〉 [37] Fire Whirl, Photograph, viewed 18 November 2015, n.d. 〈https://upload. wikimedia.org/wikipedia/commons/9/9f/Fire-whirl.jpg〉 [38] Fire Tornado, Fire Whirl Captured on Video, Photos, Photograph, viewed 18 November 2015, n.d. 〈http://geol105naturalhazards.voices.wooster.edu/files/ 2013/03/url-1.jpeg〉
34 (1979) 29–45. [10] F. Battaglia, R.G. Rehm, H.R. Baum, The fluid mechanics of fire whirls: An inviscid model, Phys. Fluids. 12 (2000) 2859. [11] K.H. Chuah, G. Kushida, The prediction of flame heights and flame shapes of small fire whirls, Proc. Combus. Inst. 31 (2007) 2599–2606. [12] F. Battaglia, K. McGrattan, R. Rehm, H. Baum, Simulating fire whirls, Combust. Theory Model. 4 (2000) 123–138. [13] R. Meroney, Fire whirls, fire tornados and firestorms: physical and numerical modeling, in: Proceedings of the PHYSMOD2003 Int. Work. Phys. Model. Flow Dispers. Phenom., Prato, Italy, 2003 [14] H.W. Emmons, S.J. Ying, The fire whirl, in: Proceedings of the Symposium Combustion 11, 1967, pp. 475–488 [15] R. Martin, D. Pendleton, W. Burgess, Effect of fire whirlwind formation on solid fuel burning rates, Fire Technol. 12 (1976) 33–40. [16] J. Lei, N. Liu, L. Zhang, H. Chen, L. Shu, P. Chen, et al., Experimental research on combustion dynamics of medium-scale fire whirl, Proc. Combust. Inst. 33 (2011) 2407–2415. [17] J. Lei, N. Liu, L. Zhang, Z. Deng, N.K. Akafuah, T. Li, et al., Burning rates of liquid fuels in fire whirls, Combust. Flame. 159 (2012) 2014–2114. [18] J. Lei, N. Liu, J.S. Lozano, L. Zhang, Z. Deng, K. Satoh, Experimental research on flame revolution and precession of fire whirls, Proc. Combust. Inst. 34 (2013) 2607–2615. [19] J. Lei, N. Liu, K. Satoh, Buoyant pool fires under imposed circulations before the formation of fire whirls, Proc. Combust. Inst. 35 (2015) 2503–2510. [20] K. Kuwana, K. Sekimoto, K. Saito, F.A. Williams, Scaling Fire Whirls, Fire Safety J 43 (2008) 252–257. [21] K. Kuwana, S. Morishita, R. Dobashi, K.H. Chuah, K. Saito, The burning rate's effect on the flame length of weak fire whirls, Proc. Combust. Inst. 33 (2011) 2425–2432. [22] K. Kuwana, K. Sekimoto, T. Minami, T. Tashiro, K. Saito, Scale-model experiments of moving fire whirl over a line fire, Proc. Combust. Inst. 34 (2) (2013) 2625–2631. [23] K.H. Chuah, K. Kuwana, K. Saito, F.A. Williams, Inclined fire whirls, Proc. Combust. Inst. 33 (2011) 2417–2424. [24] Y. Hayashi, K. Kuwana, R. Dobashi, Influence of vortex structure on fire whirl behavior, in: M. Spearpoint (Ed.), Fire Safety Science-Proceedings of the tenth International Symposium, International Association for Fire Safety Science, College
48