Multiple pool fires: Occurrence, simulation, modeling and management

Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries journal homepage: www.elsevier.com/locate/jlp

Multiple pool fires: Occurrence, simulation, modeling and management S. Vasanth, S.M. Tauseef, Tasneem Abbasi*, S.A. Abbasi Center for Pollution Control and Environmental Engineering, Pondicherry University, Chinnakalapet, Puducherry 605 014, India

a r t i c l e i n f o

a b s t r a c t

Article history: Received 15 May 2013 Received in revised form 28 November 2013 Accepted 26 January 2014

When two or more pool fires burn in such close proximity of one another that they can influence each other, they are termed ‘multiple pool fires’ (MPF). The characteristics and the structure of MPFs are significantly different from that of stand-alone pool fires. Even though MPFs have known to occur fairly often in chemical process industries, much lesser work has been done towards simulation, modeling and control of MPFs as compared to stand-alone pool fires. This paper is perhaps the first-ever attempt at surveying the MPF state-of-the-art. It recounts MPF accidents and catalogs the controlled experiments that have been done to understand the mechanism and impact of MPFs. Attempts to model MPFs have been assessed and possible ways to manage MPFs have been touched upon. Ó 2014 Elsevier Ltd. All rights reserved.

Keywords: Pool fires Multiple pool fires Accident simulation Burning rate Flame characteristics

1. Introduction Pool fires are the most frequent of process industry accidents (CCPS, 2005; Lees, 2005). Pool fires are either the initiators, or the consequences, often both, of most process industry accidents involving fires or explosions (Abbasi & Abbasi, 2007a,b,c; Abbasi, Pasman, & Abbasi, 2010; Cozzani, Tugnoli, & Salzano, 2009; Crowl & Louvar, 2001; Lees, 2005). Even though pool literally means a small body of still liquid, or a collection of stakes/monies/professionals, the term pool fire has come to represent the pool of a fuel that has caught fire (Audouin, Kolb, Torero, & Most, 1995; Cetegen & Ahmed, 1993; Liu, Liao, Li, Qin, & Lu, 2003). According to Lees (2005) a pool fire occurs when a flammable liquid spills onto the ground and is ignited. A fire in a liquid storage tank is also a form of pool fire, as is a trench fire. Numerous other definitions have been given (Cowley & Johnson, 1992; Hamins, Kashiwagi, & Buch, 1996; Mudan & Croce, 1995; Nolan, 2010; Steinhaus, Welch, Carvel, & Torero, 2007; TNO, 1997) which emphasize one or the other of the characteristics common to all pool fires while stressing the main feature: ignition of pooled fuel. The term basically represents pool of liquid fuel catching fire, but it is also used to describe burning of solid fuels, for example poly methyl methacrylate (PMMA or Plexiglass) and polyethylene (Audouin et al., 1995; Cetegen & Ahmed, 1993; Liu et al., 2003), and

* Corresponding author: Concurrently Visiting Associate Professor, Department of Fire Protection Engineering, Polytechnic Institute, Worcester, MA 01609, USA. Tel.: þ91 9751468491. E-mail address: [email protected] (T. Abbasi). http://dx.doi.org/10.1016/j.jlp.2014.01.005 0950-4230/Ó 2014 Elsevier Ltd. All rights reserved.

forest fires (Hamins et al., 1996, pp. 15e41). Based on the medium on which the pool is formed, presence or absence of confinement, and the type of location, pool fires have been classified as in Fig. 1. A review of the state-of-the-art of process industry accidents (Abdolhamidzadeh, Abbasi, Rashtchian, & Abbasi, 2011; Amendola, Contini, & Nichele, 1988; CCPS, 2005; HSE, 2012; Khan & Abbasi, 1999, 1998; Koivisto & Nielsen, 1994; Lees, 2005; Tauseef, Abbasi, & Abbasi, 2011; Tauseef, Rashtchian, & Abbasi, 2011) reveals that very elaborate and extensive studies have been done on standalone pool fires (SPF) but much less attention has been paid to MPFs. Blinov and Khudiakov (1961, p. 208) studied the flammability and ignition of liquids from burning of gasoline-like liquids in pans ranging in size from a fraction of a cm up to nearly 30 m in diameter to explain a number of phenomena observed in the ignition and burning of mixtures of liquids. They also investigated the problems connected with shapes and dimensions of the flames, pulsation, temperature, radiation and various combustion regimes. Their study reveals that thermal radiation heat transfer dominates real fires, not the pan conduction effects nor the convective heat transfer that characterizes smaller scale laboratory fires. For smallscale pool fires (diameter less than 1 m), the burning rate was found to be directly proportional to the diameter of the pool and the type of fuel feeding the SPF. The study also reveals that in SPF the burning rates per unit surface area tend to be relatively constant for pan diameters greater than about 1 m, independent of whether the flammable liquid is gasoline, kerosene, or diesel. In contrast to the range and the depth of information available on SPFs, most studies on MPFs have been limited to pools of sizes

104

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

Pool fires

Pool fires on land

Confined pool fire

Slot fire

Storage tank fire

Pool fires on water

unconfined pool fire

Dike fire

Trench fire

Fig. 1. Classification of pool fires.

ranging from just a few mm to a few cm. Very recently a report has appeared on MPFs involving pools of 1.5 m diameter (Schälike, Mishra, Wehrstedt, & Schönbucher, 2013) but this report does not alter the fact that studies on the MPFs arising in pools of diameters large enough to be relevant in the management of real-life tanks are yet to be conducted. For MPFs of size ranges studied so far, the burning rates were found to vary with diameters of the pools, and the types of fuels. Highly radiating pool fires such as the ones fueled by gasoline were found to interact more strongly than the ones fueled by low radiation fuels such as diesel (Vincent & Gollahalli, 1995). Many of the past accident reports state that the interaction of flames from MPFs result in more intense radiation, and higher flame, than it would have resulted if the flames were not interacting. But in the absence of any experimental data on MPFs with pool diameters larger than 1.5 m, it is impossible to say if the burning rates and the flame interaction is a function of pool diameter and fuel type. One of the first reports on multiple pool fires is due to Broido and Mccmasters (1960) who generated 70 fires with 2.1 m high 6.09 m  4.5 m assorted scrap lumber piles spaced 3.6 m apart along three concentric rectangles. The resulting 15.2 m high flames were seen to merge at least part of the time. The authors felt that this happened because conditions were near to the critical for merging; after about 10 min when the quantity of fuel had been depleted, the merging flames separated into a group of individually burning fires. Putnam and Speich (1963, pp. 867e877) studied different arrangements of fuel piles to generate MPFs and found that when the spacing factor given by S/(Q2o/g)1/5 was about two, the individual diffusion flames began to interact. Waterman, Labes, Salzberg, Tamney, and Vodvarka (1964) burned 0.91 m2 wood cribs arranged in square arrays to find that the onset of coalescence corresponded with a dramatic peak in the rate of burning of the cribs. In an accidental fire in two stacks of timber each 45.7 m  12.1 m, separated by a 6.09 m gangway, Baldwin, Thomas, and Wraight (1964), found that flames merged for part of the time. All through the 1960s, the behavior of fire merging received much attention. Most of the researchers tried to develop empirical models for describing fire merging behavior, by adding the fire spacing and the number of fire points into general single pool fire models. Among these attempts was the model proposed by Thomas, Simms, and Wraight (1964) for describing the steady spread of fire in cribs of wood in still air. The authors found it to be in reasonable agreement with the results that of laboratory experiments. The same

authors (Thomas, Baldwin, & Heselden, 1965, pp. 983e996) later performed experiments with two pools employing, separately, timber and town gas and obtained a dimensionless equation relating the merged flame height to fire spacing. Other noteworthy studies included the ones by Baldwin (1968), Countryman (1969), Evans and Tracey (1966) and Thomas, Baldwin, Theobald, and Britain (1968). Arguably the earliest experiments on MPFs generated by pools of liquid fuels were performed by Huffman, Welker, and Sliepcevich (1969) who reported that interaction of number of fires burning in close proximity has substantial effect on the burning rate of the fuel, the size of the flame, and the rate of heat transfer from the flame to the surroundings. They observed that individual pool fires start to burn more intensely with higher flames as the distance between them is decreased. Several authors have subsequently reported different forms of interactive effects that distinguish liquid pool MPFs in contrast to stand-alone pool fires (Chigier & Apak, 1975; Delichatsios, 2007; Fukuda, Kudo, & Ito, 2005; Kamikawa et al., 2005; Steward & Tennankore, 1981; Sugawa & Takahashi, 1993; Weng, Kamikawa, Fukuda, Hasemi, & Kagiya, 2004). This paper traces the history of major accidents involving MPFs, and catalogs the few controlled experiments that have been carried out to understand the manner in which MPFs evolve. The attempts to model MPFs, assess their impacts, and find ways to control them, are also reviewed. 2. Case histories of MPFs Given the general paucity of scientific literature on MPFs, it may appear that accidents involving MPFs are few and far between. But an analysis of the past accidents reveals that such is not the case. Between 1950 and present hundreds of major MPF accidents have been reported worldwide (Hailwood, Gawlowski, Schalau, & Schönbucher, 2009; Tauseef, Abbasi, et al., 2011). A few instances which illustrate the variety of situations under which MPF have occurred; the size and intensity of the MPFs, and the damage they can cause are presented below. These and some other representative MPF events that have occurred over the last 37 years are cataloged in Table 1. 2.1. Beek, The Netherlands On November 7, 1975, there was escape of vapor and formation of a vapor cloud, during the start up of a Naphtha Cracker unit in a

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

105

Table 1 List of major MPF accidents that have occurred since 1975. Year

Location

Material/substance

Quantity Dead/injured Reported maximum Duration (kL) flame height (m) (days)

1975 Beek, Netherlands

Naphtha, propylene

1977 Umm said, Qatar

Liquid propane, butane

1977 Illinois

Diesel

1977 Texas 1981 Shuaiba refinery tank farm, Kuwait 1983 Milford Haven, UK. 1983 Newark, New jersey, US 1983 Newark, New Jersey 1985 Naples, Italy

Distillate, ethanol Petrochemical grade naphtha Crude oil Gasoline Gasoline Gasoline, diesel, and fuel oil

Not reported 37,521 19,873 Not reported 21,718 Not reported 47,000 6677 206 27,000

1986 Thessaloniki, Greece

Crude oil, fuel oil and gasoline Domestic fuel and diesel fuel Petrol and premium grade gasoline Fuel additives Jet fuel Lead-free premium gasoline, Furnace fuel oil Gasoline Heavy fuel oil Gasoline Crude Sulfate Turpentine (CST) Sodium hydrosulfide 45% solution in water Briquest Antiblaze 80 Kerosene, avtur, naphtha Crude oil Gasoline Crude oil

1987 Lyon, France

1990 Denver, Colorado, US 1991 Saint Herblain, France

1991 1992 1994 1995

Saint Ouen, France Verdun, France Ueda, Nagano, Japan Powell Duffryn Company, UK

1995 Cilacap, Indonesia 1998 Ras Gharib, Egypt 1999 Sri Racha, Thailand 1999 West Feliciana, USA 1999 Korfez, Turkey

2002 Cabras Island, USA 2005 Buncefield, UK 2008 Koln, Germany 2008 Oil refinery in big spring, Texas 2009 Pertamina national oil company, Jakarta, Indonesia 2009 IOC, Jaipur, India. 2009 Caribbean petroleum corp, San juan, Puerto Rico. 2009 Tanjung lagsat port, Malaysia 2011 Cosmo Oil Company, Japan 2011 Sharjah oil depot, UAE 2011 Shiogama and Ichihara, Japan 2012 Jefferson Davis county, Mississippi 2012 Carabobo, Venezuela 2013 Denham springs, Louisiana 2013 IOC, Gujarat, India

Naphtha fuel, crude oil, and liquefied petroleum gas Gasoline

Not reported 1900

14/107 7/13

Not reported

(Lees, 2005)

76.35

Not reported

Not reported

Not reported

14

Not reported 1/1

Not reported 50

Not reported 6

14 42

(Zalosh, 2003) (FACTSonline, 2012)

150 50 Not reported Not reported

3 1 Not reported 6

7.3 10 25 51

Not reported

Not reported Not reported

(IFW, 2012) (FACTSonline, 2012) (Zalosh, 2003) (Clark et al., 2001; FACTSonline, 2012; Hailwood et al., 2009; Maremonti et al., 1999) (FACTSonline, 2012)

Not reported 2/14

7

22.8

(Cozzani et al., 2009; Lees, 2005) (Zalosh, 2003)

0/20 1/0 1/23 5/170

100

Not reported

Economic losses Reference (million US $)

200

1

26

(BARPI, 2012)

None 1/5

Not reported 50

2 1

30.6 20

(FACTSonline, 2012) (BARPI, 2012)

0/15 None 3/1 0/11

50 Not reported Not reported 25

1200 600 60,566 7200

670 58 244.7 2385

Not reported Not reported 1 3

3.37 1.6 1.1 8

(BARPI, 2012) (BARPI, 2012) (Sozogaku, 2013) (EPA, 2012)

1287 1022 984 33,778

None

76,313 39,746 158

Not reported 8/13 Not reported

35,931

Not reported

50 20 50 Not reported 50

Not reported

3

112

4 30 1.5 25.8 Not reported Not reported 2

Not reported Gasoline, diesel, jet fuel. 82,359

Not reported 0/43

100

Ethylene Acrylonitrile Gasoline, asphalt

300 1200 11,000

None

40

0/5

50

1

Gasoline

43,000

None

100

3

200

Not reported Not reported 4.5 Not reported

(FACTSonline, 2012) (FACTSonline, 2012) (FACTSonline, 2012) (Persson & Lönnermark, 2004, p. 14) (FACTSonline, 2012)

66.25

(Persson & Lönnermark, 2004) (Hailwood et al., 2009; HSE, 2012; MIIB, 2012) (IFW, 2012)

90

(IFW, 2012)

12,500

Not reported

Gasoline, Diesel, 60,000 Kerosene Gasoline, diesel, jet fuel 19,078 or fuel oil.

11/150

30

5

0/3

100

1

Not reported

Naphtha Gasoline Gasoline Diesel Diesel Gasoline, Diesel

11,200 16,000 35,000

None

Not reported

2

Not reported

(FACTSonline, 2012)

0/6

Not reported

10

Not reported

(Daily Telegraph, 2011)

None 6/Not reported None

Not reported Not reported

Not reported Not reported 11 Not reported

(IFW, 2012) (Wikipedia, 2012)

Crude oil

81 Not reported 127

15

Not reported Not reported

(IFW, 2012)

Crude oil Crude oil

5,000 365

None 0/3

35 20

Not reported Not reported Not reported Not reported

(IFW, 2012) (One News page, 2013)

Gasoline

5,000

None

50

2

512.4

(IFW, 2012)

18.5

(RSPCB, 2010; The Hindu, 2009; Wikipedia, 2012) (CSB, 2012; HSE, 2012; IFW, 2012)

(The Hindu, 2013)

106

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

large 100,000 tons per annum ethylene plant. The vapor cloud then exploded. Besides direct damage, it started MPFs in six tanks of 1500e6000 m3 capacity within a common dike leading to flames which were as tall as 40 m. The accident killed 14 persons and injured another 107. The losses were estimated at the equivalent of US $ 61 million (FACTSonline, 2012).

fuel to the MPFs. As the fires continued, coupling gaskets in the piping deteriorated and more fuel flowed out of the storage tanks, substantially feeding the fires. Additionally, the valve controlling fuel flow in the supply line to the airport sporadically released fuel in the valve pit. Flame heights of upto 100 m were observed at the time of the accident. The damages ran into US $ 30.6 million (FACTSonline, 2012).

2.2. Shuaiba, Kuwait 2.7. Saint Herblain, France On August 20, 1981, a fire originated at a pump manifold when pumping naphtha into one of the six 25,500 kL floating roof tanks within the common dike. A strong wind and radiated heat prevented strong fire fighting efforts from succeeding. In spite of the heavy protective cover of water streams the fire spread into an adjoining row of four 11,500 kL floating roof tanks containing intermediate products and to a fixed roof 5100 kL slop tank. Eventually five of the six tanks caught fire which lasted for about 6 days and generated flames which reached 50 m into the sky. The sixth tank was empty and sustained severe damage. One person died and one was injured in the accident. The damage ran to US $ 148 million (FACTSonline, 2012). 2.3. Newark, New Jersey, US On July 1, 1983, ignition of gasoline vapors that formed due to spillage from an overfilled floating roof tank led to a VCE. It then set of four pool fires. One person was killed, 23 others injured, and caused US $ 10 million of damage to the terminal and up to US $ 25 million in legal claims for damage to rail rolling stock and adjacent properties. Although dikes contained the burning spill, two adjoining internal floating roof tanks and a smaller transmit tank ignited and eventually were destroyed along with 19,078 kL of product. 2.4. Naples, Italy On December 21, 1985, 24 pool fires started simultaneously in tanks storing gasoline, diesel, and fuel oil after a vapor cloud explosion occurred as a result of overfilling a floating roof tank over a period of 1.5 h. About 700 kL of flammables sustained the pool fires for 6 days over an impacted area of z49,000 m2. The accident killed 5 and injured 170. The damage amounted to US $ 51 million. It was reported that at the time of incident ‘extraordinary’ radiation and flame height were observed (Clark, Deaves, Lines, & Henson, 2001; Maremonti, Russo, Salzano, & Tufano, 1999). 2.5. Lyon, France On 2 June, 1987, a violent explosion resulting from the ignition of fuel additives, released in the form of spray, by a spark from a welding operation started numerous fires which eventually spread to 11 tanks. About 37,000 kL of fuel which included 1900 kL of domestic fuel and diesel, 1200 kL of petrol and premium grade gasoline, and 600 kL of fuel additives in 14 tanks were consumed in the pool fires. There were 2 fatalities and 14 injuries. The damage was worth US $ 26 million (BARPI, 2012). 2.6. Denver, Colorado On November 25, 1990, seven pool fires occurred at a tank farm containing about 61,000 kL of jet fuel. The MPFs are believed to have been started by the ignition of fuel leaking from an operating fuel pump. The electric motor for the pump served as the ignition source. A cracked supply pipe in the valve pit provided additional

On October 7, 1991, two pool fires started as a result of a VCE caused by overfilling of gasoline. The tank farm consisted of 11 storage tanks containing 1500e15,000 m3 of fuel, adding upto 80,000 m3. Two of the tanks were completely burned out. The accident damaged another 3 hydrocarbon tanks in a nearby tank farm. Flames went up to 50 m in air. Besides the loss of a life, serious burn injuries were sustained by two employees and minor injuries by three others (BARPI, 2012). 2.8. Ueda, Japan On October 9, 1994, there was leakage of oil in an inland oil terminal at a time when some repair works were being carried out. It is possible that due to intra-unit communication errors, oil transfer were commenced in a pipeline even before the repair work had been completed (Sozogaku, 2013). Static electricity or spark from a forklift ignited the leaked oil. The resulting fire soon spread to three tanks, leading to an explosion. Three workers died and another was injured. Damages ran into US $ 1.1 million (Sozogaku, 2013). 2.9. Cilacap, Indonesia On October 24, 1995, lightning struck a floating roof storage tank containing naphtha initiating a pool fire at the 48,000 kL per-day refinery. The pool fire then multiplied involving six additional storage tanks. Flame heights of upto 50 m were observed at the time of accident. The damage was worth over US $ 38 million (FACTSonline, 2012). 2.10. Buncefield, United Kingdom (UK) On December 11, 2005, the overfilling of 300 kL of unleaded gasoline over a period of 40 min at one of the tanks located in Buncefield tank storage facility led to massive spillage and formation of a vapor cloud 1e2 m deep. On coming in contact with an ignition source, the vapor cloud exploded. It caused a simultaneous initiation of 22 pool fires in various tanks storing gasoline in the facility. The MPFs lasted for about 4.5 days and impacted an area of 80,000 m2. The vegetation around the storage facility was completely consumed by fire. Flame heights of 100 m were observed at the time of accident. Besides injuring 43, the accident cost an equivalent of US $ 13 billion (Hailwood et al., 2009; HSE, 2012; MIIB 2012; Vela et al., 2009, pp. 87e97). 2.11. San juan, Puerto Rico On October 23, 2009, 17 pool fires started simultaneously at Caribbean Petroleum Corporation after a massive VCE. Another four storage tanks burned before the blaze was extinguished 72 h later. The combined effect of the MPFs sent huge flames and smoke plumes into the air. Flames reaching more than 100 m into the air could be seen from several km away (CSB, 2012; HSE, 2012; IFW, 2012).

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

2.12. Jaipur, India On October 29, 2009, a leak from a pipeline during the transfer of petrol at Indian Oil Corporation (IOC) terminal followed by a massive VCE initiated 11 pool fires which generated flames high enough to be visible from a distance of 30 km. Over time, the flames leapt higher and wider into the air and raged for about two weeks till all the fuel had been burnt off. The accident killed 12 persons and injured another 200. More than half a million people had to be evacuated from the area during the 2-week period that the fire raged. The damages ran into US $ 32 million (RSPCB, 2010; The Hindu, 2009; Wikipedia, 2012). 2.13. Sendai, Shiogama and Ichihara, Japan The massive earthquake that struck northeastern Japan on March 11, 2011, and the subsequent tsunami, triggered MPFs in several oil refineries and industrial complexes situated in that region. The tsunami upturned a 980 kL gasoline tank at JX Nippon Oil and Energy Corporation’s refinery at Sendai. The ignition of the spilled oil floating on top of inundated tsunami water helped the fire to spread to the surrounding asphalt tanks and facilities in the harbor, triggering MPFs. No fire fighting efforts could be undertaken because of the tsunami and the fire raged till March 15, 2011, when all the fuel was consumed. The fires raging in the facilities at the Port of Sendai and a petrochemical factory at Shiogama, led to a gigantic plume of smoke about 85 km long and 2 km high (Wikipedia, 2012). Simultaneously many other MPFs started in other parts of the region hit by the earthquake. For example at the Chiba branch of Cosmo Oil refinery a fire broke out and as it spread in bursts to hit one storage tank after another, several tanks suffered boiling liquid expanding vapor explosion (BLEVEs). All attempts to douse the resulting MPFs failed and they raged for 11 days until the fuel feeding them was exhausted. Besides utter devastation of the facility, six people were injured as a direct consequence of the fire (COSMO, 2013; Daily Telegraph, 2011; Wikipedia, 2012). 2.14. Sharjah, United Arab Emirates (UAE) On March 24, 2011, a VCE started a fire which lit up a 30 kL pool of diesel. The fire then moved to a neighboring oil depot which contained 37 kL of diesel, and converted it to another pool fire. The combined effect of the two pool fires led to flames which rose up to 25 m. Four lorries parked in the compound at the time of accident were burned to scrap. The damage was estimated at US $ 0.2 million (IFW, 2012). An analysis of the flame height reported at the time of accident (Table 1) shows that the height/diameter (H/D) ratio was in the range 2.5e6.5 which is 1.2e3.2 times greater than that which would have been observed if the fires were stand-alone ones (Hailwood et al., 2009). A commonly observed sequence in the MPF incidents is shown in Fig. 2. 3. Controlled experiments on MPFs Given the fact that MPFs occur quite frequently, and cause enormous damage, it is surprising that very few studies exist on the mechanism of MPF development and the factors that control it. Of special concern is the existing paucity of knowledge about such of the interactive effects of pool fires which can make MPFs more destructive than non-interacting PFs of identical numbers and sizes. The first attempt to study interaction of factors which influences wind-blown MPFs was made by Rios, Welker, and Sliepcevich

107

(1967) using 2.43 m long wood cribs of 1.27 cm2 cross section with separation distance varying from 0.60 to 1.5 m. They reported that the rate of flame propagation, depth of the burning zone and the mass burning rate increases with increasing wind velocities. In subsequent studies (Fukuda, Kamikawa, Hasemi, & Kagiya, 2004; Huang & Lee, 1967; Liu, Liu, Deng, Kohyu, & Zhu, 2007; Liu et al., 2009; Vincent & Gollahalli, 1995) it was seen that the overall flame length as well as the mass burning rate increased with decrease in the separation distance between MPFs. Closer proximity also contributed to higher flame height and heat release rate (HRR) (Delichatsios, 2007; Fukuda et al., 2004; Kamikawa et al., 2005; Liu et al., 2007; Vincent & Gollahalli, 1995; Weng et al., 2004). An overview of the work done on MPFs so far is presented in Table 2. 4. Mechanism governing MPF In a cluster of pools the rate and the extent of flame propagation from one pool to another pool, which results in MPFs may depend on a host of factors including competition for oxygen available in the interstitial space, radiation transfer from the neighboring flames, heat feedback enhancement, wind effect, pool spacing, type of the fuel, etc (Fukuda et al., 2004; Kamikawa et al., 2005; Liu et al., 2007, 2009; Sugawa & Takahashi, 1993; Vincent & Gollahalli, 1995; Weng et al., 2004). In order to model the MPFs phenomenon, assess the impacts of MPFs, and devise prevention and control strategies for MPFs, a precise understanding of these factors and the manner in which they influence MPFs is required. Based on the experiments conducted so far, the following factors have been identified as responsible for the spread of fire from one tank to the other resulting in MPFs (Fukuda et al., 2004; Kamikawa et al., 2005; Liu et al., 2007, 2009; Vincent & Gollahalli, 1995; Weng et al., 2004). 4.1. Separation distance Separation distance between the individual pools has a large influence on the interaction among them (Fukuda et al., 2004; Kamikawa et al., 2005; Liu et al., 2007, 2009; Weng et al., 2004). When the MPFs are located close enough, the restriction of air entrainment by flames results in a pressure drop in the space among the pool fires. Due to the pressure drop the flames are deflected vertically and merge together, resulting in a flame higher than that of the single pool fire (Kamikawa et al., 2005; Weng et al., 2004). An increase in the separation distance reduces the effects of neighboring pools and beyond a limit the flames of MPFs cease to interact and behave as individual pool flames. The radiative contribution of neighboring pools also decreases with an increase in separation distance, causing a fall in the burning rate of the pools. But the competition for O2 from the surroundings decreases with the increase in separation distance which causes a positive feedback to the burning rate. A favorable impact of larger amount of O2 becoming available at larger pool separations is lower soot concentration and shorter flames (Vincent & Gollahalli, 1995). When two pool fires are adjacent to each other, the fuel surface of one fire not only receives the flame heat feedback from itself, but also receives the heat feedback from the other, mainly by radiation (Liu et al., 2009, pp. 2519e2526). This heat feedback enhancement will, in response, induce additional fuel volatiles to evaporate from the liquid, and make the burning more intense (Liu et al., 2009). The accelerated burning fire, in turn, releases more heat to the surrounding flames. The opposite effect is caused by the restriction of air entrainment which occurs in closely spaced MPFs: it decreases combustion efficiency and thus reduces the heat feedback to the fuel surface, in turn decreasing the burning rate (Liu et al., 2009).

108

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

Loss of containment due to • operational error • equipment/ instrument failure • tank crack/ rupture • piping rupture/leak

Pool formation within dike

Ignition source Spark Non-explosion proof motor and tools Welding Static electricity

Liquid at or above flash point/ boiling point

Yes

• • • •

No

Vapour cloud formation

Ignition Lightening ignites the content of storage tank

Pool fire Vapour cloud explosion

Boilover Heat load

MPF

Fig. 2. Sequence of events leading to MPFs.

4.2. Wind effect Wind may facilitate the spread of fire from one pool to another and may also influence the intensity and the rate of spread of a fire. As wind blows across a fire, it pushes the flame forward and closer to the unburned fuel downwind of the fire. This increases convection and radiation, which increases the rate of heating of the unburned fuel layer and thus the rate of spread of the flame front (Whelan, 1995). The rate of spread of the flame front into the unburned region is largely independent of the speed of the prevailing wind when the unburned region is in the upwind direction, but increases with the increase of the speed of the prevailing wind if the unburned region is in the downwind direction (Huang & Lee, 1967). Rios et al. (1967) have reported that in two fires burning in close proximity, the upwind fire had a greater depth of burning zone than a single fire under wind velocities less than approximately 1 m/s because of increased preheating of the fuel. The downwind fire is largely shielded from the wind effects by the upwind fire, and its burning characteristics under calm or low wind velocity conditions are similar to a stand-alone fire. As the wind velocity increases, the depth of burning zone and flame propagation rate increases in MPFs more rapidly than in a single pool fire. This behavior is caused by differences in the air flow into and around the proximate fires as compared to the single fire. 4.3. Materials used for insulation in storage tank Fire resistant insulation reduces the risk of a fire outside a tank from igniting the flammable liquid contained in the tank. It also

reduces the temperature of the tank surface. Commonly used insulation materials include mineral wool, fiberglass, silicate and perlite. Such insulations can withstand temperatures in excess of 1000  C and act as barriers to the fires, preventing the spreading of the fires (EPA, 2012). Some of the materials have thermal conductivities four orders-of-magnitude lower than that of steel. Commercial insulations that have been certified to achieve a requisite level of fire resistance are available (Zalosh, 2003). 4.4. Characteristics and quantity of fuel present in the pools The interaction of pools depends on the fuel type. Highly radiating pool flames interact more strongly than pool flames with low radiation. The flammability and the volatility of the fuel also affect the fire spread (Vincent & Gollahalli, 1995). 4.5. Combustion products The amount of soot produced in MPFs is known to depend on the diameter of the tanks, and the amount and the type of fuel contained in each (Hailwood et al., 2009). According to Koseki and Mulholland (1991) the soot concentration decreases with the increase in separation distance between pools. When the separation distance is small, the merged flame behaves essentially like a large diameter pool flame, and the soot concentration is high. When the separation distance is increased, more soot becomes oxidized fully to CO2 because of the higher availability of oxygen. There is a critical separation distance between the pools when the soot value is maximized. Above this value of separation distance, the increase in CO2 formation rate is masked by the dilution effect of the entrained

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

109

Table 2 Experiments done on MPFs Sl no

Fuel involved

Study objectives and aspects covered

Gist of experiments

Key findings

Reference

1

n-Heptane and ditert-butyl peroxide (DTBP)

Objective: To determine the limiting distances for flame merging of MPFs Aspect covered: Flame merging

i) Two to nine fuel pans made of steel, 6 cm in diameter, were placed in test bunker of 10 m  5 m  5 m. Fire spacing was varied between 2  D/d  9. ii) Two pools made of steel, 1.5 m in diameter were tested in the field. Fire spacing was varied between 0.17  D/d  1

(Schälike et al., 2013)

2

n-Heptane

Objective: To study spatial variations of burning rates in a fire array for different values of fire spacing and array size. Aspect covered: Variation of burning rate along a fire array

Fuel pans, 5 cm in diameter and 2 cm in height, were completely filled with 98% n-heptanes and arranged in arrays of 3  3 e15  15. Fire spacing was varied from 5 to 50 cm.

3

n-heptane

Objective: To study the burning rate of square fire arrays. Aspect covered: Burning rate

Fuel pans, 5 cm in diameter and 2 cm in height, were completely filled with 98% n-heptane. They were studied in arrays of 3  3 e15  15. Fire spacing was varied from 5 to 50 cm for 3  3 e7  7 and 20e40 cm for 15  15 arrays.

4

Propane

Objective: To study the merging flame height in group fires. Aspect covered: Heat release rate (HRR)

Porous burners of diameter 15 cm arranged in arrays of 2  2e5  5 were used with propane as the fuel.

1. Flame merging was observed when the relative distance between at least two fires falls below a limiting distance. 2. Three regions for flame merging were defined and limiting distances were determined: (i) A merging region with the limiting distance Dmerg/d within which all flames merge together over the complete burning time. (ii) A transition region with the limiting distance Dtran/ d within which some flames merge together while other flames are separated. (iii) A separated region within which all flames are completely separated from each other. 3. Correlations (equations (38)e(41)) for limiting distances against the number of burning pools were also obtained. 1) A new concept termed ‘fire layer’ is introduced and is defined to characterize the complex spatial variations of burning rates for fire layers, under different conditions. Thus facilitating analysis and physical interpretation of complex spatial variation of burning rates for fire layers. 2) The nonlinear spatial distribution of average burning rates along fire array indicates that the two interaction effects e the heat feedback enhancement and air supply restriction e involve distinct spatial fluctuations in the fire array. 3) The heat feedback enhancement and air supply restriction in fire arrays are significantly affected by the two major parameters e fire spacing and fire array size. 4) The average burning rates for all fire layers vary linearly as a function of the fire area ratio. The fire area ratio denotes the ratio of the total fuel surface area to the whole fire array area. 1) Due to two competing mechanisms e heat feedback enhancement and air entrainment restriction e the average burning rate of fire array was found to be intimately connected with inter fuel pans spacing and array size in a complex way. 2) When the fires were close to each other (0.2e0.5 m), the mechanism of heat feedback enhancement was more dominant than air entrainment restriction. It induced more intense burning as compared with a single fire with the same fuel area. 3) With increasing separation distance between the fires, the heat feedback decays more rapidly compared to increase in air entrainment. Flame heights were seen to correlate with pool diameter and HRR as per expression. Lf * 2=3 D ¼ 2:7 Q

(Liu et al., 2012)

(Liu et al., 2009)

(Delichatsios, 2007)

(continued on next page)

110

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

Table 2 (continued ) Sl no

Fuel involved

Study objectives and aspects covered

Gist of experiments

5

n-heptane

Objective: To develop a burn out time data analysis method to analyze interaction effects among multiple pool fires. Aspect covered: Fire merging and fire whirls

Identical steel circular fuel pans, 6 cm in diameter and 2 cm in height were completely filled with 98% n-heptane. They were studied in arrays of 3  3e7  7. The fire spacing was varied from 20 to 50 cm.

6

n-heptane and methane

Objective: To study the characteristic of two small pool fires arranged in two different horizontal planes. Aspects covered: 1) Burning rate 2) Flame height 3) Puffing frequency.

Two brass pans of 48 mm diameter and 20 mm depth were used with heptane or methanol as fuel. The horizontal separation distance between the two pans was varied from 0 to 150 mm and vertical separation distance between the two pans was varied from 0 to 600 mm.

7

Propane,

Objective: To study the merging flames from multi-fire sources. Aspects covered: 1) Flame height 2) HRR

Square-shaped porous diffusion burners of 0.15 cm side length in arrays of 2  2e5  5 were used with propane as the fuel. The separation distance between burners was varied from 0 to 3 cm.

8

Propane

Objective: To study the flame characteristic of group fire. Aspects covered: 1) Flame height 2) HRR

Porous burners of diameter 15 cm arranged in arrays of 2  2e5  5 were used with propane as the fuel.

9

Iso-octane, jet A

Five pools made of thin sheet metal of 1.6 mm thickness and 25.4 cm diameter were used with octane and Jet A as fuels.

10

Propane, hexane and heptane

Objective: To study the effects of diameter, spacing, and shape of the pools on the characteristics of multiple liquid pool fires. Aspects covered: 1) Flame height 2) Temperature 3) Soot, oxygen and CO2 concentration 4) Pool surface regression rate 5) Burning rate peak radiation of center and outer pools Objective: To study the flame height behavior of multi-fire sources. Aspect covered: Flame height

Two rectangular fire sources, 3 and 4 circular pools of 120 mm diameter were used in parallel and symmetrical configuration respectively. A rectangular gas diffusion burner, 20 mm  400 mm or 20 mm  800 mm was used as line

Key findings where, Lf, merged flame height; D, pool diameter; Q*, dimensionless HRR. 1) It was observed that in each fire had the, burning intensity increased significantly with higher flame height, when compared with a single free-standing fire. 2) The flames from the outer fuel pans showed a strong leaning toward the center, indicating strong entrainment in-flows at the boundaries. 1) The burning rate of two pool fires arranged in a horizontal plane becomes larger than that of a single pool fire and reaches a maximum value at a separation distance of 60 mm between the two pans. 2) When two pool fires were arranged in different horizontal planes, the flame height of the upper pan was smaller than that of lower pan, contrary the puffing frequency of upper flame was higher than that of lower flame. 1) For 2  2 array when the dimensionless separation distance (S/D, where S (m) is the separation distance between burners and D (m) is the side of burner) decreased from 2 to 0 the flame height increased by 40%. 2) For 3  3 array when the dimensionless separation distance decreased from 2 to 0 the flame height increased by 60%. 3) For 4  4 array when the dimensionless separation distance decreased from 2 to 0 the flame height increased by 80%. 4) For 5  5 array when the dimensionless separation distance decreased from 2 to 0 the flame height was doubled. The influence of the separation distance on the flame merging property was smaller than that of the number of the burners and the HRR. Especially when the separation distance was under D/5, (where D is the diameter of the pool) the flame height was hardly affected by the separation distance excluding the moment when the flame began to merge. 1) Burning rate, flame height and soot concentration decreased with the increase in separation distance between the pools. 2) The interaction of pools depended on the fuel type. Highly radiating pool flames interacted more strongly than pools with low flame radiation. 3) The fuel layer thickness and the position of the pool in the cluster also affected the interaction of the pool flames. The interactive distance over which a flame could affect nearby flames was about four times its representative length. When the dimensionless separation (S/D, where S is the separation distance between burners (m) and D is the width of shorter length of rectangular burner (m) for

Reference

(Liu et al., 2007)

(Fukuda et al., 2005)

(Weng et al., 2004)

(Fukuda et al., 2004)

(Vincent & Gollahalli, 1995)

(Sugawa & Takahashi, 1993)

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

111

Table 2 (continued ) Sl no

Fuel involved

Study objectives and aspects covered

11

Heptane

Objective: To study the merging characteristics of four heptanes pool fires.

12

Natural gas

Objective: To study the interaction of diffusion flame. Aspects covered: 1) Flame height 2) Flame temperature 3) Gas concentration under the degree of swirl and burner separation

13

Wood

Objective: 1) To determine the minimum size of fire and fuel loading at which mass fire, and particularly fire storm effects, occur, so as to provide a basis for future and more sophisticated studies. 2) To explore the instrumentation problem in mass hire research, and develop instrumentation for such experimental work. 3) To acquire as much quantitative information as possible on fire systems, particularly in areas of primary interest to civil defense problems. 4) To test the validity of the descriptive model of a simple mass fire system.

Gist of experiments

Key findings

fire source with propane as fuel. Hexane was employed as fuel for circular pool fires.

rectangular fire sources or pool diameter (m) for circular pools) increased to more than 4, the flame height approached the height of an isolated fire source. It was observed that at separation distance less than 0.6 m, flame merging causes increases in burning rate and flame height.

Four tanks each having diameter of 0.8 m and height 0.4 m were used with separation varied from 0.3 to 0.6 m. Two concentric swirl burners with a 20 half angle diffuser exit with a diameter of 52.5 mm were used with natural gas as fuel. The fuel was supplied to each burner through a single hole nozzle terminating at the throat with a diameter of 31.8 mm. Six experimental or test fires were done using multiple-fuel beds to simulate urban conditions. The fuel beds were built by arranging uprooted pinyon pine and Utah juniper trees in square piles covering about 185.8 m3. Each pile of fuel was 14.23 m on a side, averaged 2.1 m high, and contained about 18,143 kg of fuel (dry weight). Fire spacing was varied from 7.62 to 35.05 m.

Aspects covered: 1) Flame spread

14

Methanol, acetone, hexane, cyclohexane and benzene

Objective: To study interaction between MPFs Aspect covered: Burning rate

15

Town gas (methane)

Objective: To find critical conditions associated with flame merging. Aspects covered: 1) Flame merging

Circular burners, 10.16 cm in diameter, were used in the 13burner pattern, whereas burners with diameters of 5.08 cm, 10.16 cm, and 15.24 cm were used in the 9burner pattern. Town gas burners each 0.30 m2, where arranged horizontally in various configurations and separation distances (S). The flame height (L) was varied by varying the gas flow for each separation until the flames were just merging, to find the critical conditions of merging.

Reference

(Koseki & Yumoto, 1989)

Due to flame interaction the overall flame height increased with a decrease in the burner separation distance. These effects were more pronounced at low swirl levels.

(Chigier & Apak, 1975)

Observations of wild land fires and prescribed burns indicate that once a fire is established in a fuel bed, the fire will burn well over a wider range of moisture content when the fuel loading is deep than when it is sparse. This probably results from better conservation of available heat in the deep fuel bed, thereby offsetting the effect of the water vapor. Observations of free-burning fires in the open leaves little doubt that surface or low level winds are a major factor in the spread of fire. Many investigations have been made into the effect of wind on fire spread, but there still remains considerable uncertainty of the exact relationships. The uncertainty is even greater for the effect of wind on the burning rate of wood fuels where fire spread is not a factor. The air flow patterns in and around the Flambeau test fires showed consistent patterns for a given fuel-bed arrangement. In the single-fuel-bed test, the fire blocked the ambient wind, with turbulence and eddies forming in the wake of the fire. Little direct inflow into the sides of the fire could be detected. The same flow pattern appeared around individual fuel beds in the multiple-fuel-bed fires, where the fuel beds were wide-spaced. The burning rates for the center and intermediate burners peaked at about the same dimensionless separation (S/ D), of about 2.5, whereas the outer burners did not reach a peak within the limits of minimum separation distance attainable with the equipment. When the fires are a long way apart the flames behave like those from individual fires. When S ¼ 0, the flames merge completely and behave like the flame of a single fire. For intermediate values of S/D, the flames lean towards each other and attain a height (L). When the flames are just merging, the increase in flame height due to the reduction of surface area of the flame and of the heat loss to the surroundings is probably quite small and to a reasonable degree of approximation the

(Countryman, 1969)

(Huffman et al., 1969)

Baldwin (1968)

(continued on next page)

112

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

Table 2 (continued ) Sl no

Fuel involved

Study objectives and aspects covered

16

Wood

Objective: To test the feasibility of designing a small-scale model that may be used in the laboratory for the study of multiple fires. Aspects covered: 1) Flame merging

17

Propane

Objective: To study the interaction of two parallel line fires. Aspect covered: Angle of inclination between two interacting flames for various heat release rates and separation distance

18

Wood cribs

Objective: To study the behavior of interacting fires under windy conditions. Aspects covered: 1) Flame propagation rate 2) Burning rate under various wind speeds

19

Timber and town gas

Objective: To study the air entrainment, heat transfer and flame merging of interacting fires. Aspects covered: 1) Flame merging 2) Burning rate

20

Wood cribs

Objective: To study the fire spread in wood cribs in still air. Aspects covered: 1) Flame spread

21

Propane

Objective: To study the effect of fuel-flow rate, spacing between diffusion flames, and/or the diameter of the diffusion flame on flame merging in MPFs. Aspects covered: 1) Flame merging

Gist of experiments

Experiments were performed for arrays of three different sizes 6  6, 10  10 and 16  16 with separation distance to diameter of the pool (S/D) ratios of ¼ or ½. In some of the experiments the square units were placed side by side with no spacing to form a large continuous fuel bed. Two identical line channel burners made of galvanized steel 1.82 m long, 17.78 cm high with a width tapering from 2.54 cm at the bottom to about 3.17 cm at the top, were used.

A wind tunnel test section approximately 7.62 m long was used with 2.43 m long wood cribs of about 1.27 cm2 cross section. The separation distance between the centers of the proximate fires was varied from 0.60 to 1.5 m. Experiments with two pools employing, separately, timber and town gas were performed. The ‘town gas’ was burned in 60  30 cm and 30  30 cm burners. The conditions of flame heights and their separation were varied to determine the conditions when the flames just merged. The larger-scale MPF employed two stacks of timber with horizontal cross section 45.70 m  12.19 m separated by a 6.09 m gangway. Wood sticks which were approximately 1.3 cm thick, spaced approximately 4 cm from each other were employed in the experiments. Different widths and heights of cribs were employed.

Two burners with multiple nozzles were used; each nozzle representing a pool fire .The first burner was 0.25 m in diameter with 7 nozzles; and the second burner was 1.21 m in diameter with 120 nozzles. Center-to-center nozzle spacing was 0.05 m and 0.10 m in the two burners, respectively. Symmetric arrays, circular arrays and line arrays were used to study the effect of such configurations on flame merging.

Key findings

Reference

flame height may be calculated from single flame data. Flame merging was observed for S/ D ¼ ¼ which is consistent with the theoretical predictions based on formula given by Baldwin (1966). No merging was observed for S/D ¼ ½.

Thomas et al. (1968)

1) The two flames leaned toward each other. It was also found that both flames were reasonably straight. 2) The heat flux going into the cooling water, computed from the mass flow rate and the temperature rise of cooling water, was always negligibly small compared with the total heat flux released from the flame, computed from the mass flow rate of fuel and its heat of combustion in air. The rate of flame propagation and the depth of the burning zone increased with increasing wind velocity.

(Huang & Lee, 1967)

It was found that L/L* w6/5 (where L* is the height of flame from an isolated fuel bed) when the flames were just merging. It was also observed that the flame from two stacks merged at a separation distance of 15.24 m. These flames were probably fully merged and therefore this result gives a lower limit for separation distance at which the two flames merged.

Thomas et al. (1965)

A theoretical model given by Thomas et al. (1964), describing the steady spread of fire in cribs of wood in still air is shown to be in reasonable agreement with the results of laboratory experiments. For a wide range of conditions the rate of fire spread R is given by Rrb ¼ 5 e9 mg cm2 sl where R is in cm/s and rb (kg/m3) is the bulk density of the fuel bed. It was found that the ratio of flame height of the array of diffusion flames to single-diffusion flame height is a function of dimensionless spacing S/ (Q2o /g)1/5, the number of diffusion flames (n), and source-shape factor. For circular arrays, if the ratio given by

Thomas et al. (1964)

ðL0 =L* Þ1 n2=5  1

¼ 2

3

6 7 function 4source shape factor; n;  S1 5 2 5 Q

o g

has a value between zero to one then source shape or number of diffusion flames were found to have no effect on

Rios et al. (1967)

Putnam & Speich (1963)

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

113

Table 2 (continued ) Sl no

Fuel involved

Study objectives and aspects covered

Gist of experiments

Key findings

Reference

flame merging When brought closer than a critical spacing factor of about two, the individual diffusion flames began to interact. At a moderate number of diffusion flames, the exact number depending on the array pattern and the dimensionless spacing, the array may be treated as a single extended fire, giving results correlating with results on stacks of wood. In the study of the linear arrays, a gradual shift of the relation of flame height to flow rate from that observed for individual sources to that predicted for a pure line source was expected. This shift did not occur. Apparently the characteristic spacing for these studies was not made small enough to obtain this expected shift. A semi-empirical correlation of the array flame height indicated that flame height reaches a finite maximum as the number of diffusion flames in a line is increased.

air. The balance between these two phenomena leads to a critical value of separation distance at which the CO2 concentration attains a peak value (Mulholland, Henzel, & Babrauskas, 1989; Vincent & Gollahalli, 1995). The combustion products of MPFs may contain more soot, carbon dioxide, carbon monoxide, nitrogen oxides, sulfur dioxide and particulates than stand-alone fires. The first reason is that multiple fires do not get adequate oxygen supply to burn the carbon produced by the pyrolysis of fuel vapor. This not only produces soot, but also lowers the overall heat released and hence the temperature, resulting in more smoke production (Hamins et al., 1996). The second cause is the lowering of the effective concentration of fuel and vapor due to the recirculation of burnt gases by the vortex that is prevalent in all such fires (Hamins et al., 1996). 4.6. Tank roof Standards have been set by agencies such as National Fire Protection Association (NFPA) and American Petroleum Institute (API), for storage tanks of designated pressure rating and specified roof design. Tanks designed to hold cryogenic liquids are dealt as a special category (Zalosh, 2003). The roof type is chosen on the basis of tank diameter and product to be stored. Cone roof tank, external/open top floating roof tank, and internal floating roof tank/covered floating roof are the ones commonly employed. Of these, cone roof storage tanks have vertical sides and are equipped with fixed cone-shaped roofs that are welded to the sides of the tank. Tanks that confirm to API standards are provided with weak seam at the joint where the roof and sides meet so that, in the event of an internal explosion, the roof blows off leaving the tank shell intact. This system allows the tank to retain its contents and any resulting fire involves the full surface of the exposed flammable liquid. In floating roof tanks, the roof of the tank rises and lowers with the stored contents thereby reducing vapor loss and minimizing fire hazard. These are commonly used in oil refineries. Fixed roof tanks are suitable for low volatility, high flash point liquids such as ethylene glycol, kerosene; Jet A fuel, and Number 6 fuel oil. These fuels generate vapor pressures which are high

enough at ambient temperatures to cause flammability limits to cross. But if there is fire exposure vaporization may be hastened to generate flammable mixtures in the tank vapor space. If there is flame entry into the tank during such situations it could result in an explosion. Hence, in fixed roof atmospheric tanks, also, the roof-toshell seams are kept week so that if they suffer an explosion in the vapor space, it would cause the roof to detach leaving the tank shell intact (Zalosh, 2003). There are also lifter-roof tanks in use which allow the roof to slide up and down such that there is a variable vapor space above the liquid surface. A liquid channel at the top of the tank wall provides a roof-to-shell seal. The vapor space is usually fuel rich except when the liquid is pumped out sufficiently rapidly to necessitate air in-breathing. Lifter-roof tanks have been used to store volatile liquids but they are a lot fewer in number compared to floating roof tanks (Zalosh, 2003). 5. Modeling of MPFs Although a number of models exist e empirical, integral, analytical, advanced analytical-computational and the so-called ‘field’ (based on computational fluid dynamics, CFD) e for modeling of stand-alone pool fires; only a few empirical and field models has been developed to model MPFs (Delichatsios, 2007; Fukuda et al., 2005; Huang & Lee, 1967; Huffman et al., 1969; Kamikawa et al., 2005; Liu et al., 2009; Rios et al., 1967; Satoh, Liu, Liu, & Yang, 2008; Satoh, Naian, Qiong, & Yang, 2007; Sugawa & Takahashi, 1993; Weng et al., 2004). An overview of these models is presented below. 5.1. Empirical models for MPFs The empirical models for MPFs developed so far have aimed to correlate the total thermal energy radiated by the MPFs with the physical characteristics of fire, aspects of overall combustion chemistry, and variation of the thermal output from different parts of MPFs. These models can be used to estimate the flame geometry (Delichatsios, 2007; Huang & Lee, 1967; Kamikawa et al., 2005; Sugawa & Takahashi, 1993; Weng et al., 2004), burning rate

114

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

(Huffman et al., 1969; Liu et al., 2009; Rios et al., 1967) and the radiation from an MPF (Delichatsios, 2007; Fukuda et al., 2005; Weng et al., 2004). Empirical models have good predictive value if the situations to which the models are applied fall within the range of validity of the models (Delichatsios, 2007; Fukuda et al., 2005). A common drawback of these models is that they are based on data obtained from very small-scale laboratory experiments. This is mainly due to the high costs as well as risk associated with conducting largerscale tests. This aspect limits the applicability of the empirical models to real-life situations. 5.1.1. Models for estimating MPF flame height 5.1.1.1. Model proposed by Putnam and Speich (1963). Putnam and Speich (1963) modified the correlation proposed by Ricou and Spalding (1961), to arrive at expression for single pool fires, given by equation (1).

L ¼ KðQo Þ2=5

(1)

where, L is the flame length of a single pool fire; K is a constant; Qo is the volume flux of injected fuel. Putnam and Speich (1963) further assumed that flames from multiple burners placed very close to each other would behave as a single pool fire and the total merged flame height (L0 ) can be given by

L0 ¼ KðnQo Þ2=5

(2)

where, n is the total number of burners. Equations (1) and (2) led to

L0 2 ¼ n5 L*

(3)

where, L0 and L* are the heights of single and multiple pool fires, respectively, fed by fuel injected at the same flow rate (Qo). Equation (3) gives the merged flame when burners are placed side by side. For the situation when the pool fires placed at a separation distance of S just start to merge, the height of the combined flame L0 , would be given by

3

2

6 S 7 source shape factor; n;  1 7 ¼ function6 5 4 * L Qo2 5 L0

(4)

g

where, the source-shape factor is some function which describes the shape of the set of small or point sources. Putnam and Speich (1963) also proposed an alternate formulation

2 3  0 * 6 L =L  1 S 7 7 ¼ function6 4source shape factor; n;  2 15 5 n2=5  1 Qo

2) The rate of mass flux of aspirated air to mass flux of injected fuel (m/ mo) at the flame tip is constant for a pool fire irrespective of whether it is isolated from, or is in close proximity to other pool fires. 3) There is no significant deviation of flame from the vertical because of the mutual aspiration of adjacent pool fires. Such a “pulling together” of the flames would decrease the actual spacing between flames. Based on these, Putnam and Speich (1963) extended the correlation proposed by Ricou and Spalding (1961) to estimate the separation distance at which flames from multiple pool fires would merge, for the cases of: (a) two pool fires in close proximity; (b) three pool fires in an equilateral triangular pattern; (c) six and seven pool fires in a hexagonal pattern; and (d) three pool fires in a line. For analysis of the experimental data (as summarized in Table 2) the symmetric flame sources were treated either as point sources or as area fire. For the symmetric flames treated as point sources, the dimensionless groups L0 /L* was plotted as a function of S/(Q2o /g)1/5 for different sets of number of flames and source shape. The experimental data as well as the predicted theoretical curves were plotted and from the plot it was found that the theoretical curves show a strong correlation between the two dimensionless groups. The experimental data do not correlate with the theoretical curves but instead show a more rapid rate of change of flame height. For the symmetric flame sources treated as an area fire, Putnam and Speich (1963) correlated their experimental data in terms of an area fire in which the fuel-flow rate is the total fuel available for combustion and the principal dimension is some dimension indicative of the total area of the fuel source. This permitted a comparison with area fire data. It was assumed that the pool fires, regardless of their array pattern, are in a square array. With spacing between diffusion flames of S, the principal dimension of the array then is equal to one side of the square given by equation (6).

 1  D ¼ S n2  1

(6)

where, n is the number of sources in the array. Equation (6) makes it possible to obtain D from S and n for configurations other than square arrays. Putnam and Speich (1963) then correlated the gaseous-fuel data generated from their experiment in a form used by Thomas (1963) in his work on solid fuel fires. The array to area conversion factors were estimated to be

L L  ¼  1 D S n2  1

(7)

and

(5)

g

This expression has the advantage of having limits of zero and one, but term (L0 /L*)  1 exaggerates small variations in the measurement of flame height. Putnam and Speich (1963) next derived the flame height for intermediate burner spacing using the following assumptions: 1) For any pool fire surrounded by other pool fires, the path of the entraining air is slightly blocked. The blockage is assumed proportional to the diameter of the surrounding pool fires. Thus, compared to a stand-alone pool fire, it takes longer for the MPF to aspirate a specific mass flow of air.

q2 =gD5 ¼





3

2

6 Qo2 =gd5o 4

n2 1 2

d5o

n  1

57 5 S 5

(8)

One of the arrays studied by Putnam and Speich (1963) was that of the line fire. The curve, passing through the data for all values of n used, was given by

2 16 4 29 

3 S Qo2 =g

7 1=5 5 

  S 1 ½1expð0:24 ðn1ÞÞ ¼ Ln 67

Equation (9) would give the maximum value of (Ln/S)

(9)

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

ðLn =SÞmax ¼

29  1=5 7 S= Qo2 =g  16

(10)

The correlation factor (fs) was seen as



fs ¼

L0

where, is the flame height of the combined flame under the condition of close proximity; L* is the flame height of single isolated flame; Ln is the height of flame array made up of (n) diffusion flames; n is the number of diffusion flames; D is the principal dimension of an area fire; Qo is the flow rate; do is the nozzle port diameter; S is the separation distance between the burners; g is the gravitational constant. 5.1.1.2. Model proposed by Thomas et al., (1965). They proposed a theoretical model for flame height of two pool fires when their flames are just merging. It was based on the assumptions that a) the radius of curvature of each flame is large compared to its length, thus, axis of flame may be taken as straight; b) the only forces acting on the flames are the buoyancy (upwards) and pressure thrust, acting normal to the axis; and c) air entrainment is unaffected by the pressure drop on one side of the flame or by the leaning of the flames. These led to the expression



S3

L ¼ 9 D DW 2

1=2 (11)

where, L is the flame height (cm); D is the dimension characteristic of burning zone e.g. tray diameter; S is the separation between fuel beds (cm); W is the length of long side of fuel bed (cm). Thomas et al. (1965) also performed experiments with two pools employing timber and ‘town gas’ separately. The ‘town gas’ was burned in 60  30 cm and 30  30 cm burners. The conditions of flame heights and their separation were varied to determine the conditions when the flames just merged. It was found that L/L* w6/5 (where L* is the height of flame from an isolated fuel bed) when the flames were just merging. The larger-scale MPF employed two stacks of timber with horizontal cross section 45.70 m  12.19 m separated by a 6.09 m gangway. It was observed that the flame from two stacks merged at a separation distance of 15.24 m. These flames were probably fully merged and therefore this result gives a lower limit for S. These experiments demonstrated that the flame height predicted by theory is realistic. 5.1.1.3. Model proposed by Sugawa and Takahashi (1993). This model is based on the data from the experiments performed for three configurations. The first configuration employed two rectangular gas diffusion burners with base area of 20 mm  400 mm and 20 mm  800 mm as fire source with propane as the fuel. The second and third configurations employed 3 and 4 circular pans, respectively, of 200 mm diameter arranged symmetrically. In the first configuration burners were set 300 mm above the floor, avoiding the floor effect on the entrainment. In the other two configurations the burner surface was set at almost floor level. In all the configurations hexane was used as the fuel. The experiment was used to study flame height and obtain a correlation between flame height, pool diameter, and separation distance between the pools. The dimensionless flame height (L/Lm) for circular pools was found to fit the equation

115

 S2 þ nD2   n S2 þ D 2

(13)

where, D is the width of the square burner or diameter of the circular burner (m); n represents number of pools; S is the separation distance (m). Based on the ‘Zukoski curve’ (Zukoski, Kubota, & Cetegen, 1981), used to establish the relationship between the dimensionless flame height and dimensionless heat release rate (HRR) on a logarithmic scale, Sugawa and Takahashi (1993) obtained the following corre*  40. lation for 1  Qm * 2=5

L ¼ 3:3 ðfs Þ2=5 Qm

Dm

(14)

* and D where, Qm m are dimensionless HRR and the diameter of burner (m) respectively. The dimensionless flame height (L/Lm) for rectangular pools in a parallel configuration was found to fit the equation

L ¼ ðf1 Þ2=3 Lm

(15)

The correlation factor (f1) for the flame height was seen as

f1 ¼

  2DW þ S2   2 DW þ S2

(16)

where, D (m) is the length of shorter side of the rectangular burner, W (m) is the length of longer side of the burner, and S (m) is the separation distance between the two burners in parallel configuration. 5.1.1.4. Model proposed by Kamikawa et al. (2005) and Weng et al. (2004). Kamikawa et al. (2005) and Weng et al. (2004) carried out a series of experiments using square-shaped porous diffusion burners of 0.15 cm side length arranged in arrays of 2  2e5  5, with propane as the fuel. The separation distance between burners was varied from 0 to 3 cm to study its effect on the flame height of MPFs. A correlation was obtained between flame height, side length of the burner, number of burners and separation distance between the burners. The dimensionless flame height (Lf/Lm) was found to be

  1B 0  pffiffiffiffipffiffiffiffi 2 N N  1 S2 þ lD2 Lf   A ¼ @  pffiffiffiffipffiffiffiffi Lm l 2 N N  1 S2 þ D2

(17)

(12)

where, Lf is the merged flame height for pool fires separated from each other (m); Lm is the merged flame height for pool fires without separation between them (m); N represents number of fire sources arranged in square array; D is the side length of the square-shaped porous diffusion burners (m); S is the separation distance between pool fires (m); B is the power constant determined from the experimental data with no separation distance between the pools. The merged flame height calculated using the above correlation was compared with the numerical results obtained using the Fire Dynamics Simulator (FDS). Close agreement was found between the two for the fire source configurations, 3  3, 4  4, and 5  5. For the 2  2 configuration the empirical model predicted slightly higher flame height for smaller separation distance between the burners than the numerical model.

where, L is the flame height of a single pool fire (m); Lm is the merged flame height (m).

5.1.1.5. Model proposed by Delichatsios (2007). Based on the experimental data reported by Fukuda et al. (2004), Delichatsios

L ¼ ðfs Þ2=5 Lm

116

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

(2007) obtained a correlation for flame height of MPFs with pool diameter and HRR. The experiment (Table 2) performed by Fukuda et al. (2004) used porous burners of diameter 15 cm arranged in arrays of 2  2e5  5 with propane as the fuel. The correlation between flame height, pool diameter and HRR was given as

Lf ¼ 2:7Q * 2=3 D

where, D is the depth of the flaming zone in wood crib measured from flame photographs.

(18)

where, Lf is the merged flame height (m); D is the pool diameter (m); Q* is the dimensionless HRR given by

Q* ¼

Q

r0 Cp T0 g1=2 D5=2

(19)

where, Q is the HRR (kW); r0 is the density of ambient air (kg/m3); Cp is the specific heat of air (kJ/kg K); T0 is the ambient temperature (K); g is the gravitational acceleration (m/s2) 5.1.2. Models for estimating MPF flame tilt 5.1.2.1. Model proposed by Huang and Lee (1967). Huang and Lee (1967) used two identical line channel burners, 1.82 m long and 17.78 cm high with a width tapering from 3.17 cm at the top to about 2.54 cm at the bottom. They studied the phenomenon of flame tilting. The angle (f) between the axis of the flame and the horizontal plane representing the flame tilt was correlated with heat fluxes per unit length of the burners of the two flames by the following expression:

f ¼ fnðM; NÞ

(20)

Of these M, N are readable from the graphs available in Huang and Lee (1967). Where,

M ¼ q=Q N ¼

 2=3 1 Q y0 g 1=3 r0 Cp T0

(21)

(22)

where, g is the gravitational acceleration (m/s2); r0 is the density of ambient air (kg/m3); Cp is the specific heat of air at constant pressure (kJ/kg K); T0 is the temperature of ambient air (K); Q is the heat flux per unit length of the burner of the flame for which f is calculated (kW/m2); q is the heat flux per unit length of the burner of the auxiliary flame (kW/m2); y0 is half of the separation distance between the flames (m). 5.1.3. Models for estimating MPF burning rate 5.1.3.1. Model proposed by Rios et al. (1967). Rios et al. (1967) used 2.43 m long wood cribs, of 1.27 cm2 cross section area, as the fuel. Unlike liquid fuels, which should be at or above their flash point to give off vapors that can form an ignitable mixture, solid fuel such as wood cribs must first pyrolize to form gases or vapors before they can burn in flame. Pool fires fueled by solid fuel generally tend to have lower burning rates compared to the ones fed by liquid fuel. The wood cribs were arranged in a wind tunnel test section which was approximately 7.62 m long. The separation distance between the centers of the wood cribs was varied from 0.60 to 1.5 m. Based on the data thus generated, correlations at steady state were obtained for flame propagation rate p, mass burning rate m, and _ with the depth of the flaming zone burning rate per unit area m, estimated from the flame photographs.

p ¼ 0:00194D0:67

(23)

m ¼ 0:015D0:72

(24)

_ ¼ 0:00522D0:29 m

(25)

5.1.3.2. Model proposed by Huffman et al. (1969). Huffman et al. (1969) performed four experiments using circular burners. For the first experiment 13 burners of 10.16 cm diameter were arranged in circular symmetry pattern. The other experiments used three sets of 9 burners with diameters 5.08 cm, 10.16 cm, and 15.24 cm arranged in circular symmetry pattern. The focus of the experiments was on the interaction between the MPFs. Two generalized correlations, based on the data from experiment using 9 burners, of burning rate data with separation distance between the burner and diameter of the burner were made. The first correlation was for the center burner alone and the second was for the average burning rate of all nine burners combined (Hc). For the center burner

m ms



DHc DHv

 2:9

rg ra

¼ f

 

 S mp rg 0:87 D ms ra

(26)

where, S is spacing between the burners (m); D is the burner diameter (m); mp is the peak burning rate per unit area of interacting fire (kg/s m2) For the average burning rate of all nine burners

m ms



DHc DHv

 2:6

rg ra

¼ f

 1:3

S mp D ms

(27)

where, m is the burning rate for unit area of interacting fire (kg/ s m2); ms is the burning rate per unit area of a single fire (kg/s m2); DHc is the heat of combustion of fuel (kJ/kg); DHv is the heat of vaporization at the boiling point plus the sensible heat for raising the fuel from ambient temperature to the boiling point (kJ/kg); rg is the density of fuel vapor at boiling point (kg/m3); ra is the density of ambient air (kg/m3). 5.1.3.3. Model proposed by Liu et al. (2009). Liu et al. (2009) used circular fuel pans, 5 cm in diameter and 2 cm in height, completely filled with 98% n-heptane, arranged in arrays of 3  3e7  7 and 15  15. The spacing between the burners was varied from 5 to 50 cm for 3  3e7  7 arrays and from 20 to 40 cm for 15  15 array. The burning rate was studied for different separation distances. For any n  n fire array with an initial constant fuel mass (m) for each fuel pan, the average burning rate was found to be

BRh PN

Nm

i¼1

BOTðiÞ

(28)

where, BOT(i) is the burn out time of the fire (s); N ¼ n2 the number of fuel pans; m is the initial constant fuel mass for each fire point (kg). The time average burning rate of the single reference fire was found to be

BRr ¼

m BOTr

(29)

Dimensionless average burning rate for the fire array was found to be

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

P BR N= N i ¼ 1 BOTðiÞ ¼ BRr 1=BOTr

BR* ¼

117

For no separation distance

(30) D ¼ ONd

For all the fire array sizes investigated, the burning rate (BR*) was correlated with rs as formulated

where, N is the number of burners with side of d. HRR for propane experiments was expressed as

3  3: BR* ¼ 6:22rs þ 1:20r 2 ¼ 0:9700

(31)

4  4: BR* ¼ 15:95rs þ 1:21r 2 ¼ 0:9817

(32)

where, q is the HRR for each burner.

5  5: BR* ¼ 14:82rs þ 1:44r 2 ¼ 0:9624

(33)

6  6: BR* ¼ 28:40rs þ 1:20r 2 ¼ 0:9800

(34)

7  7: BR* ¼ 35:62rs þ 1:12r 2 ¼ 0:9877

(35)

15  15: BR* ¼ 52:92rs þ 1:07r 2 ¼ 0:9993

(36)

5.1.4.2. Model proposed by Fukuda et al. (2005). Fukuda et al. (2005) obtained a correlation for HRR of MPFs with geometric view factor based on the data from an experiment performed using two brass pans of 48 mm diameter and 20 mm depth with heptane or methanol used as the fuel. The horizontal separation distance between the two pans was varied from 0 to 150 mm and vertical separation distance between the two pans was varied from 0 to 600 mm to study the characteristic of MPFs. Solid flame model was used to calculate the radiative heat flux (q) from one pool fire to the liquid fuel surface of the other.

where, rs denotes the dimensionless ratio of the total fuel surface area to the whole fire array area represented as

rs ¼

p=4 ½1=n þ ðn  1Þ=n D * 2

(37)

where, D* is the dimensionless fire spacing equal to D/d; d is the pool diameter (m); D is the distance between the centers of two adjacent fuel pans (m). 5.1.4. Models for estimating MPF HRR 5.1.4.1. Model proposed by Kamikawa et al. (2005) and Weng et al. (2004). Kamikawa et al. (2005) and Weng et al. (2004) obtained a correlation for HRR of MPFs with flame height and burner side length based on the experiment performed using square-shaped porous diffusion burners of 0.15 cm side length arranged in arrays of 2  2e5  5 with propane as the fuel. The separation distance between burners was varied from 0 to 3 cm. Propane and wood crib burners were operated at various HRR. Correlations obtained between HRR, flame height and burner side length were found to be

Lf =D ¼

3:11QD* 0:57

for 2  2 propane data;

(38)

Lf =D ¼ 2:85QD* 0:68 for 3  3 propane data;

(39)

Lf =D ¼ 2:98QD* 0:78 for 4  4 propane data;

(40)

Lf =D ¼ 4:07QD*1:08 for 5  5 propane data;

(41)

Lf =D ¼ 3:66QD* 0:96 for 2  2 wood crib data;

(42)

Lf =D ¼ 3:66QD* 0:82 for 3  3 wood crib data;

(43)

where, Lf is the flame height (m); D is the burner side length (m) and QD* is the dimensionless HRR, expressed as

QD* ¼

Q

r0 Cp T0 g1=2 D5=2

(44)

where, Q is the HRR (kW); r is the density of ambient air (kg/m3); Cp is the specific heat of air (kJ/kg K); T0 is the ambient temperature (K); g is the gravitational acceleration (m/s2).

Q ¼ N$q

  q ¼ 4εs Tf4  T 4

(45)

where, ε is the emissivity; s is the StephaneBoltzmann’s coefficient (J/s m2 K4); Tf is the flame temperature (K); T is the ambient temperature (K); 4 is the geometric view factor (m2), expressed as

( "sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi#  1 m m ðA  2nÞ Aðn  1Þ 1 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ pffiffiffiffiffiffiffiffi tan tan 4 ¼ pn p n A$B Bðn þ 1Þ n2  1 )  1 n1  tan1 n nþ1 (46) where, A ¼ (1 þ n)2 þ m2; B ¼ (1  n)2 þ m2; m ¼ H/R; n ¼ (L  R)/R; H is the flame height given as average single flame height (m); R is the flame radius (R ¼ D/2, D: pan diameter) (m), and; L is the distance from the center of flame to receiving element (m). 5.1.4.3. Model proposed by Delichatsios (2007). Delichatsios (2007) used the experimental data reported by Fukuda et al. (2004) and obtained a correlation for HRR with separation distance and diameter of the burner. Fukuda et al. (2004) had used porous burners of diameter 15 cm arranged in arrays of 2  2e5  5, with propane as the fuel. Correlations between HRR, separation distance and diameter of the burner were obtained as

Qg* ¼

Q* ¼

Q*



N 1=2 1 þ

 ðN1ÞS 5=2 ND

Q

r0 Cp T0 g1=2 D5=2

(47)

(48)

where, Qg* is the dimensionless HRR for the group fire; Q* is the dimensionless HRR for each individual burner; D is the diameter of the burner (m); Q is the HRR (kW); r is the density of ambient air (kg/m3); Cp is the specific heat of air (kJ/kg K); T0 is the ambient temperature (K); g is the gravitational acceleration (m/s2). 5.1.5. Models for estimating MPF Flame merging 5.1.5.1. Model proposed by Baldwin (1968). Baldwin (1968) proposed a model for the estimation of critical condition for merging of

118

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

flames in a matrix of ni  ni fires (where i ¼ 1.n), each with a square base of side D and separated from its neighbor by a channel of width S. In order to find the condition for flame merging, it was assumed that at the onset of merging of flames the deflection from the vertical would not be very large resulting in flames, from each fuel bed, with the same mean height L. It was also assumed that for each concentric sub-matrix (ni  1eni  1) of the ni  ni matrix of fires, the flow of air is the same into each of the channels separating fires on the perimeter, and that the rate of entrainment of air is unaffected by the flame leaning or by the pressure gradient across the flame. Based on these assumptions Baldwin (1968) calculated the total air entrainment into the fire; estimated the angle of tilt of the flame by balancing the pressure thrust acting on the lateral surface of the flame with the buoyancy force acting upwards and equated (S/L) with the difference of the angle of tilt for flames from two concentric ring of the matrix. This led to the critical condition for flame merging

S=D ¼ f ðL=DÞ

S=L ¼ 0:14ðW=DÞ2=3

5.1.5.2. Model proposed by Schälike et al. (2013). Schälike et al. (2013) carried laboratory scale experiments using two to nine fuel pans made of steel, 6 cm in diameter, placed in test bunker of 10 m  5 m  5 m. Fire spacing was varied between 2  D/d  9. They also carried out field experiments using two pools made of steel, 1.5 m in diameter. Fire spacing was varied between 0.17  D/ d  1 to study the flame merging. Three regions for flame merging were defined and limiting distances were determined, as detailed in Table 2. The limiting distances were correlated with the number of burning pools (N) as follows:

(49)

For two rectangular burners with the surface area W  D equation (49) may be written as

 1=3 S= W 2 D ¼ f ðL=DÞ

(50)

Dmerg;Hc ¼ 0:11N 0:68 d

(53)

Dtran;Hc ¼ 0:14N 1:03 d

(54)

Dmerge;DTBP ¼ 0:46N0:29 d

(55)

Dtran;DTBP ¼ 1:01N 0:47 d

(56)

where, d is the diameter of the pool; D is the separation distance between two pools; N is the number of burning pools; Dmerg;Hc is the limiting distance for the merging region for hydrocarbon fuel; Dtran;Hc is the limiting distance for the transition region for hydrocarbon fuel; Dmerg,DTBP is the limiting distance for the merging region for DTBP; Dmerg,DTBP is the limiting distance for the transition region for DTBP. The limiting distances for flame merging of n-heptane pool fires observed in this study are much smaller in comparison to the values of Liu et al. (2009). Schälike et al. (2013) have attributed it to the absence of an exact definition of the merging and the transition regions.

where, f is an unknown function. Baldwin (1968) also performed experiments, summarized in Table 2, and plotted the data generated by him along with the data generated earlier by others (Broido & Mccmasters, 1960; Putnam & Speich, 1963; Thomas et al., 1964; Waterman et al., 1964) to develop a graph of L/D versus S/(W2D)1/3 values. The resulting curve seemed to fit well in equation (50), leading to the regression correlation applicable for 1 < L/D < 10

 1=3 S= W 2 D ¼ 0:14ðL=DÞ0:96

(52)

(51)

5.1.6. Predictive ability of various models An assessment of the predictive ability of various models is summarized in Table 3. It reports the flame height, burning rate and HRR predicted by the empirical models described above for the

The index of L/D being z1, the onset of the merging condition can be represented by

Table 3 Comparison of the predictions for flame height, burning rate, HRR and separation distance of which the flames merge by various empirical models with corresponding experimental data reported by Koseki and Yumoto (1989) and Vincent and Gollahalli (1995) for MPF involving pool fires of diameter 25.4 cm and 80 cm respectively. Models

Diameter of the pool fires used in the experiment on which the model is based (m)

Flame height (Thomas et al., 1965)a e (Sugawa & Takahashi, 1993) 0.12 (Weng et al., 2004) 0.0015 (Delichatsios, 2007) 0.15 Burning rate (Rios et al., 1967) e (Huffman et al., 1969) 0.05, 0.10 and 0.15 (Liu et al., 2009) 0.05 HRR (Weng et al., 2004) 0.0015 (Fukuda et al., 2005) 0.048 (Delichatsios, 2007) 0.15 Separation distance of which flames merge (m) (Schälike et al., 2013) 1.5 a

MPF with 25.4 cm diameter pool fires as used in experiment by Vincent and Gollahalli (1995)

MPF with 80 cm diameter pool fires as used in experiment by Koseki and Yumoto (1989)

Model predictions

Percentage error (deviation from the data reported by Vincent & Gollahalli, 1995)

Model predictions

Percentage error (deviation from the data reported by Koseki & Yumoto, 1989)

1.58 0.859 0.823 0.921

53 20 23 14

7.2 6.1 6.4 5.8

51 22 28 16

0.0245 0.0295 0.0458

53 44 13

0.0212 0.0276 0.0431

60 49 19

174 104 118

26 24 14

e e e

e e e

e

e

0.467

22

Predictions have been made for square pools with area equal to circular pools used by Koseki and Yumoto (1989) and Vincent and Gollahalli (1995).

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

MPFs studied by Koseki and Yumoto (1989) and Vincent and Gollahalli (1995). As may be seen when the diameter of the pool fire of which characteristics are being predicted is close to the diameter of the pool fire on which the empirical model was based, the predicted are correspondingly close. 5.2. Computational fluid dynamics (CFD) models CFD models validated with adequate experimental data can be used to simulate a variety of pool fire scenarios with reasonable accuracy (Sinai & Owens, 1995). Several CFD models have been developed specifically for simulation of pool fires. The scenarios that have been modeled range from pool burning fire in the open (Chatris et al., 2001; Ferrero, Muñoz, Kozanoglu, Casal, & Arnaldos, 2006; Henriksen, Ring, Eddings, & Nathan, 2008; Muñoz, Planas, Ferrero, & Casal, 2007; Planas-Cuchi, Chatris, López, & Arnaldos, 2003; Sudheer & Prabhu, 2010; Vela et al., 2009; Woods, Fleck, & Kostiuk, 2006) to compartment fires (Hamins, Johnsson, Donnelly, & Maranghides, 2008; Jun et al., 2008; Koseki & Mulholland, 1991; Pierce & Moss, 2007; Steckler, Quintiere, & Rinkinen,, 1982; Tieszen, O’Hern, Weckman, & Schefer, 2004). Some of the most widely used CFD models for simulation of pool fires include FDS (Hostikka, McGrattan, & Hamins, 2003, pp. 383e394; McGrattan, Rehm, & Baum, 1994), J Adaptive Structured Meshes applications INfrastructure (JASMINE) (Cox & Kumar, 1987; Mawhinney & Richardson, 1997), Kameleon (Nicolette, Gritzo, Holen, & Magnussen, 1994), Smartfire (Ewer et al., 1999; Wang, Jia, Galea, Patel, & Ewer, 2001) and Simulation Of Fires In Enclosures (SOFIE) (Bressloff, Moss, & Rubini, 1996; Carlsson, 1990; Lewis, Moss, & Rubini, 1997). Some research and special application codes have also been used for simulation of pool fires (Olenick & Carpenter, 2003). All these models have shown to work well for a variety of fire scenarios and have been validated extensively using data from stand-alone pool fire experiments. But all these models are specific to stand-alone pool fires. Very few CFD-based studies have been done on MPFs. Satoh et al. (2007) and Weng et al. (2004) used CFD to simulate the merging of flames from MPFs. Weng et al. (2004) used largeeddy simulation (LES) to simulate the merged flame from MPFs, and compared the numerical results with the experimental data (Table 2). They have found that numerical results accorded well with the experimental data although the predicted flame height was slightly lower than the flame height reported in the experiment. They studied the effect of the size of the square array, interfire distance and HRR on the pool fires merging. The simulated profiles of isothermal surfaces of merging fires were quite similar to the ones experimentally determined by Satoh et al. (2007). The critical merging distance in the simulations was also close to that reported in the experiment. In another CFD study, Satoh et al. (2008) have investigated the swirling conditions of fires in square arrays in the presence of wind by varying the inter-fire distance, HRR, and mass flow rate. It was found that the fire whirl generation was highly affected by the inter-fire distance in the array, the total HRR and also the mass flow rate. Since the experimental data of Satoh et al. (2008) was limited, direct comparison between the experiments and the numerical simulation results was not made but the swirling profiles between the experiment and the numerical simulation were quite similar. Till about a decade back, the use of CFD was restrained due to the long and expensive computational times that were involved, but those constraints have been increasingly reduced over the years due to the falling costs and rising capabilities of computational equipment. It is now possible to employ CFD much more inexpensively and extensively in simulating/forecasting process

119

industry accidents. But, as of now, very little experimental data is available on MPFs to validate CFD-based models and to calibrate them. What little data that does exist has been mostly generated with very small-scale desktop experiments. This lack of adequate data has restricted the use of CFD in MPF simulation to a couple of studies mentioned above. Hence there is an urgent need to organize experiments on MPFs and obtain data pertaining to different fuels, pool areas and depths, inter-pool distances and positions, turbulence effects, etc, on sizes, shapes, intensities, radiation loads, soot formation and other aspects of MPFs. 6. Prevention/control of MPFs As of now no accident prevention/control strategies and codes of practice exist that are specific to MPFs. Given that even stand-alone fires are extremely dangerous, requiring specialized knowledge and equipment for effective control, the challenge posed by MPFs is even greater. The possibility of re-flash and re-ignition that exists with stand-alone pool fires (MIIB 2012) is even greater with MPFs. Efforts are needed to develop safeguards applicable to situations where occurrence of MPFs is possible. There is also a need to develop codes of practice for such situations. The following aspects are particularly relevant to MPFs, in addition to all other measures associated with the prevention and control of stand-alone pool fires:  Keeping adequate provision to ensure safe and quick drain-off of the fuel in case a fire starts. The history of MPFs is replete with the fires that began in one of the tanks in a farm and later spread to other tanks. Provisions to block such spread can drastically reduce the risk of MPFs.  Positioning of the storage tanks in a manner that ensures safe distance between the tanks based on the considerations of local meteorology, especially wind directions and speeds. This safeguard would also reduce the likelihood of the fire in a tank from spreading to other tanks.  Extra precautions against overfilling/spillage. Almost always pool fires occur due to accidental spillage. Spillage is also the main cause of vapor cloud explosions which either start pool fires or follow them.  The walls of catch basin should have adequate height and it should be properly maintained. The catch basin should be located at an adequate distance from the storage tanks with provisions made to eliminate the possibility of fire spreading to it from adjacent pool fires.  Pumping stations should be provided with drainage systems capable of quickly and safely draining away the flammable should a spill occur.  Lightning has been a significant ignition source for storage tank fires, as it is, indeed, for all fires. Lightening endangers storage tanks in two ways, by direct strike and secondary effects. The latter take the forms of the molecular/atom bound charges; electrostatic pulses, and earth currents (Carpenter, 1996). Secondary effects carry greater probability of starting a fire than direct strikes. Codes such as the ones contained in BS 6651: 1985 have been set to minimize lightening hazard but they have proved fallable on many an occasion (Carpenter, 1996; Chang & Lin, 2006).  Snow and ice can damage outdoor storage tanks and the piping connected to the tank. This can result in leaks and spills and give rise to fire or explosion hazard. The accumulation of snow and ice around storage tanks and piping can also reduce accessibility of trouble shooters to those units in case of emergency. Temporary or permanent covers should be used to protect damage to tanks and piping from snow and ice (NHDES, 2008).

120

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121

 Heavy rain, or events like a tsunami, which cause flooding may displace stored flammables and damage electrical systems, enhancing risk of fires (EPA, 2012). 7. Summary and conclusion When two or more pool fires burn in close enough proximity to influence one another they are termed multiple pool fires (MPFs). A survey of past accidents involving MPFs reveals that MPFs are fairly common and are known to lead to extraordinarily high flame heights and exceptionally intense thermal radiation compared to stand-alone pool fires. This fact makes MPFs far more hazardous than stand-alone pool fires, warranting efforts to study their mechanism and model them so that strategies to control them can be devised. But there is a surprising paucity of information on MPFs. The present review is perhaps the first attempt to fill this gap with a survey of the MPF state-of-the-art. It is hoped that the retrospective it provides would sensitize and catalyze further work towards the understanding and management of MPFs. Acknowledgement SMT thanks the Council of Scientific and Industrial Research (CSIR), New Delhi for Senior Research Associateship. References Abbasi, T., & Abbasi, S. A. (2007a). Accidental risk of superheated liquids and a framework for predicting the superheat limit. Journal of Loss Prevention in the Process Industries, 20, 165e181. Abbasi, T., & Abbasi, S. A. (2007b). The boiling liquid expanding vapour explosion (BLEVE): mechanism, consequence assessment, management. Journal of Hazardous Materials, 141, 489e519. Abbasi, T., & Abbasi, S. A. (2007c). Dust explosions-cases, causes, consequences, and control. Journal of Hazardous Materials, 140, 7e44. Abbasi, T., Pasman, H. J., & Abbasi, S. A. (2010). A scheme for the classification of explosions in the chemical process industry. Journal of Hazardous Materials, 174, 270e280. Abdolhamidzadeh, B., Abbasi, T., Rashtchian, D., & Abbasi, S. A. (2011). Domino effect in process-industry accidents - An inventory of past events and identification of some patterns. Journal of Loss Prevention in the Process Industries, 24, 575e593. Amendola, A., Contini, S., & Nichele, P. (1988). MARS: the major accident reporting system. In Preventing major chemical and related process accidents: Vol. 445 (pp. 1121e1141). Audouin, L., Kolb, G., Torero, J., & Most, J. (1995). Average centreline temperatures of a buoyant pool fire obtained by image processing of video recordings. Fire Safety Journal, 24, 167e187. Baldwin, R. (1966). Some tentative calculations of flame merging in mass fires. Fire Research Station. Baldwin, R. (1968). Flame merging in multiple fires. Combustion and Flame, 12, 318e324. Baldwin, R., Thomas, P., & Wraight, H. (1964). The merging of flames from separate fuel beds. Fire Research Station. Blinov, V. I., & Khudiakov, G. N. (1961). Diffusion burning of liquids. Report #T-1490 (a-c). Moscow: Academy of Sciences. Bressloff, N. W., Moss, J. B., & Rubini, P. A. (1996). CFD prediction of coupled radiation heat transfer and soot production in turbulent flames. Symposium (International) on Combustion, 26, 2379e2386. Broido, A., & Mccmasters, A. W. (1960). Effects of fires on personnel in shelters. Berkeley: California, Department of Agriculture Forest Service Pacific Southwest Forest and Range Experimental Station. Technical Paper 50. Buncefield Major Incident Investigation Board (MIIB). (2012). Buncefield Investigation [Online]. Available: http://www.buncefieldinvestigation.gov.uk/index. htm Accessed 11.01.12. Bureau for Analysis of Industrial Risks and Pollutions (BARPI). (2012). The analysis, research and information on accidents (ARIA) database [Online]. Available: http://www.aria.developpement-durable.gouv.fr/The-ARIA-Databasee5425. html Accessed 12.01.12. Carlsson, J. (1990). Fire modelling using CFD-an introduction for fire safety engineers. Journal of Heat Transfer, 112, 945e951. Carpenter, R. B. (1996). Lightning protection for flammables storagefacilities. Boulder, CO. USA: Lightning Eliminators, Consultants. CCPS. (2005). Guidelines for chemical process quantitative risk analysis (2nd ed). New York: Centre for Chemical Process Safety, AIChE. COSMO. (2013). Fire and explosions at Chiba refinery [Online]. Available: http:// www.cosmo-oil.co.jp/eng/csr/sustain/pdf/2011/sus2011e_2.pdf Accessed 28.11.13.

Cetegen, B. M., & Ahmed, T. A. (1993). Experiments on the periodic instability of buoyant plumes and pool fires. Combustion and Flame, 93, 157e184. Chang, J. I., & Lin, C.-C. (2006). A study of storage tank accidents. Journal of Loss Prevention in the Process Industries, 19, 51e59. Chatris, J. M., Quintela, J., Folch, J., Planas, E., Arnaldos, J., & Casal, J. (2001). Experimental study of burning rate in hydrocarbon pool fires. Combustion and Flame, 126, 1373e1383. Chemical Safety Board (CSB). (2012). Caribbean petroleum refining tank explosion and fire [Online]. Available: http://www.csb.gov/investigations/detail.aspx? SID¼87 [Accessed 13.01.12. Chigier, N. A., & Apak, G. (1975). Interaction of multiple turbulent diffusion flames. Combustion Science and Technology, 10, 219e231. Clark, S., Deaves, D., Lines, I., & Henson, L. (2001). Effects of secondary containment on source team modelling. HSE Contract Research Report. Countryman, C. M. (1969). Project flambeau e An investigation of mass (1964e1967) (Vol. 1). DTIC Document. Cowley, L. T., & Johnson, A. D. (1992). Oil and gas fires: Characteristics and impact. Offshore Technology Information. OTI 92 596. Cox, G., & Kumar, S. (1987). Numerical model of fire in road tunnels. Tunnels and Tunnelling International, 19, 55e57, 59. Cozzani, V., Tugnoli, A., & Salzano, E. (2009). The development of an inherent safety approach to the prevention of domino accidents. Accident Analysis and Prevention, 41, 1216e1227. Crowl, D. A., & Louvar, J. F. (2001). Chemical process safety: Fundamentals with applications (2nd ed). New Jersey: Prentice Hall. Daily Telegraph. (2011). Japan oil refinery blaze finally put out [Online]. Available: http://www.dailytelegraph.com.au/japan-oil-refinery-blaze-finally-put-out/ story-e6freuyi-1226025714794 Accessed 12.08.12. Delichatsios, M. A. (2007). A correlation for the flame height in “Group” fires. Fire Science Technology, 26, 1e8. Environmental Protection Agency (EPA) U.S.A.. (2012) [Online]. Available: http:// www.epa.gov/oem/docs/oil/newsletters Accessed 18.10.12. Evans, E., & Tracey, E. (1966). Observations of Mass Fire 460-14 at Mono Lake. US Naval Radiological Defense Laboratory. Ewer, J., Galea, E. R., Patel, M. K., Taylor, S., Knight, B., & Petridis, M. (1999). SMARTFIRE: an intelligent CFD based fire model. Journal of Fire Protection Engineering, 10, 13e27. FACTSonline. (2012). TNO industrial and external safety [Online]. Available: http://www.factsonline.nl/browse-chemical-accidents-in-database Accessed 05.11.12. Ferrero, F., Muñoz, M., Kozanoglu, B., Casal, J., & Arnaldos, J. (2006). Experimental study of thin-layer boilover in large-scale pool fires. Journal of Hazardous Materials, 137, 1293e1302. Fukuda, M., Kudo, Y., & Ito, A. (2005). Characteristics of two small pool fires arranged two different horizontal planes. In Proceedings of the 5th Asia-Pacific conference on combustion. Fukuda, Y., Kamikawa, D., Hasemi, Y., & Kagiya, K. (2004). Flame characteristics of group fires. Fire Science Technology, 23, 164e169. Hailwood, M., Gawlowski, M., Schalau, B., & Schönbucher, A. (2009). Conclusions drawn from the Buncefield and Naples incidents regarding the utilization of consequence models. Chemical Engineering & Technology, 32, 207e231. Hamins, A., Johnsson, E., Donnelly, M., & Maranghides, A. (2008). Energy balance in a large compartment fire. Fire Safety Journal, 43, 180e188. Hamins, A., Kashiwagi, T., & Buch, R. R. (1996). Characteristics of pool fire burning. ASTM Special Technical Publication, 1284. Health and Safety Executive (HSE). (2012). Control of major accident hazards (COMAH), [Online]. Available: http://www.hse.gov.uk/comah/index.htm Accessed 05.11.12. Henriksen, T. L., Ring, T. A., Eddings, E. G., & Nathan, G. J. (2008). Puffing frequency and soot extinction correlation in JP-8 and heptane pool fires. Combustion Science and Technology, 180, 699e712. Hostikka, S., McGrattan, K. B., & Hamins, A. (2003). Numerical modeling of pool fires using LES and finite volume method for radiation. Worcester, MA. Huang, W., & Lee, S. (1967). Interaction of two parallel line fires. State Univ. of New York at Stony Brook, College of Engineering. Huffman, G. K., Welker, R. J., & Sliepcevich, M. C. (1969). Interaction effects of multiple pool fires. Fire Technology, 5, 225e232. Industrial Fire World (IFW). (2012). Available: http://www.fireworld.com/ IncidentLogs.aspx Accessed 12.01.12. Jun, F., Yu, C. Y., Ran, T., Qiao, L. F., Zhang, Y. M., & Wang, J. J. (2008). The influence of low atmospheric pressure on carbon monoxide of n-heptane pool fires. Journal of Hazardous Materials, 154, 476e483. Kamikawa, D., Weng, W., Kagiya, K., Fukuda, Y., Mase, R., & Hasemi, Y. (2005). Experimental study of merged flames from multifire sources in propane and wood crib burners. Combustion and Flame, 142, 17e23. Khan, F. I., & Abbasi, S. (1998). Techniques and methodologies for risk analysis in chemical process industries. Journal of Loss Prevention in the Process Industries, 11, 261e277. Khan, F. I., & Abbasi, S. (1999). Major accidents in process industries and an analysis of causes and consequences. Journal of Loss Prevention in the Process Industries, 12, 361e378. Koivisto, R., & Nielsen, D. (1994). FIREda database on chemical warehouse fires. Journal of Loss Prevention in the Process Industries, 7, 209e215. Koseki, H., & Mulholland, G. W. (1991). The effect of diameter on the burning of crude oil pool fires. Fire Technology, 27, 54e65.

S. Vasanth et al. / Journal of Loss Prevention in the Process Industries 29 (2014) 103e121 Koseki, H., & Yumoto, T. (1989). Burning characteristics of heptane in 2.7 m square dike fires. Fire Safety Science, 2, 231e240. Lees, F. P. (2005). Lee’s loss prevention in the process industries: Hazard identification, assessment, and control (3rd ed). Oxford, UK: Elsevier. Lewis, M., Moss, J., & Rubini, P. (1997). CFD modelling of combustion and heat transfer in compartment fires. In Proc. 5th int. symp. on fire safety science (pp. 463e474). Australia: International Association for Fire Safety Science. Liu, J., Liao, G., Li, P., Qin, J., & Lu, X. (2003). Experimental study on the interaction of fine water mist with solid pool fires. Science in China Series E: Technological Sciences, 46, 218e224. Liu, N., Liu, Q., Deng, Z., Kohyu, S., & Zhu, J. (2007). Burn-out time data analysis on interaction effects among multiple fires in fire arrays. Proceedings of the Combustion Institute, 31, 2589e2597. Liu, N., Liu, Q., Lozano, J. S., Shu, L., Zhang, L., Zhu, J., et al. (2009). Global burning rate of square fire arrays: Experimental correlation and interpretation. Montreal, QC. Liu, N., Liu, Q., Lozano, J. S., Zhang, L., Deng, Z., Yao, B., et al. (2012). Multiple fire interactions: a further investigation by burning rate data of square fire arrays. Proceedings of the Combustion Institute, 34, 2555e2564. McGrattan, K., Rehm, R., & Baum, H. (1994). Fire-driven flows in enclosures. Journal of Computational Physics, 110, 285e291. Maremonti, M., Russo, G., Salzano, E., & Tufano, V. (1999). Post-accident analysis of vapor cloud explosions in fuel storage areas. Process Safety and Environmental Protection, 77, 360e365. Mawhinney, J., & Richardson, J. (1997). A review of water mist fire suppression research and development, (1996). Fire Technology, 33, 54e90. Mudan, K., & Croce, P. (1995). Fire hazard calculations for large open hydrocarbon pool fires. Quincy: National Fire Protection Association. Mulholland, G., Henzel, V., & Babrauskas, V. (1989). The effect of scale on smoke emission. In Proceedings of the second international symposium on fire safety science (pp. 347e357). New York: Hemisphere. Muñoz, M., Planas, E., Ferrero, F., & Casal, J. (2007). Predicting the emissive power of hydrocarbon pool fires. Journal of Hazardous Materials, 144, 725e729. New Hampshire Department of Environmental Services (NHDES). (2008). weatherRelated Public Service Announcement “Snow & Ice Can Damage Fuel Storage Tanks”. Available online http://des.nh.gov/media/pr/2009/091210.htm Accessed 13.8.2013. Nicolette, V., Gritzo, L., Holen, J., & Magnussen, B. (1994). Comparison of the KAMELEON fire model to large-scale open pool fire data. Albuquerque, NM (United States): Sandia National Labs. Nolan, D. P. (2010). Handbook of fire and explosion protection engineering principles: For oil, gas, chemical and related facilities (2nd ed). Norwich: William Andrew Publishing. Olenick, S. M., & Carpenter, D. J. (2003). An updated international survey of computer models for fire and smoke. Journal of Fire Protection Engineering, 13, 87e110. One News page. (2013). Oil tank fire near Denham Springs spreads to second tank, crews plan to inject foam [Online]. Available: http://www.onenewspage.com/n/ US/74vudsxhj/Oil-tank-fire-near-Denham-Springs-spreads-to.htm Accessed 12.05.13. Persson, H., & Lönnermark, A. (2004). Tank fires. SP Report. Pierce, J., & Moss, J. (2007). Smoke production, radiation heat transfer and fire growth in a liquid-fuelled compartment fire. Fire Safety Journal, 42, 310e320. Planas-Cuchi, E., Chatris, J. M., López, C., & Arnaldos, J. (2003). Determination of flame emissivity in hydrocarbon pool fires using infrared thermography. Fire Technology, 39, 261e273. Putnam, A., & Speich, C. Y. (1963). A model study of the interaction of multiple turbulent diffusion flames. Symposium (International) on Combustion. Elsevier. Rajasthan State pollution Control Board (RSPCB). (2010). Report of the committee on environmental impacts of fire in Indian Oil Corporation Deport. Ricou, F., & Spalding, D. (1961). Measurements of entrainment by axisymmetrical turbulent jets. Journal of Fluid Mechanics, 11, 21e32. Rios, J., Welker, J., & Sliepcevich, C. (1967). Interaction effects of wind-blown flames from wood crib fires. Fire Technology, 3, 129e136. Satoh, K., Liu, N., Liu, Q., & Yang, K. (2008). Numerical and experimental study of fire whirl generated in 15  15 square array fires placed in cross wind. ASME. Satoh, K., Naian, L., Qiong, L., & Yang, K. (2007). Numerical and experimental study of merging fires in square arrays. ASME. Schälike, S., Mishra, K. B., Wehrstedt, K. D., & Schönbucher, A. (2013). Limiting distances for flame merging of multiple n-heptane and di-tert-butyl peroxide pool fires. Chemical Engineering Transactions, 32, 121e126.

121

Sinai, Y., & Owens, M. (1995). Validation of CFD modelling of unconfined pool fires with cross-wind: flame geometry. Fire Safety Journal, 24, 1e34. Steckler, K. D., Quintiere, J. G., & Rinkinen, W. J. (1982). Flow induced by fire in a compartment. Symposium (International) on Combustion, 19, 913e920. Steinhaus, T., Welch, S., Carvel, R. O., & Torero, J. L. (2007). Large-scale pool fires. Journal of Thermal Science, 11, 101e118. Steward, F. R., & Tennankore, K. N. (1981). The measurement of the burning rate of an individual dowel in a uniform fuel matrix. Symposium (International) on Combustion, 18, 641e646. Sudheer, S., & Prabhu, S. V. (2010). Measurement of flame emissivity of gasoline pool fires. Nuclear Engineering and Design, 240, 3474e3480. Sugawa, O., & Takahashi, W. (1993). Flame height behavior from multi-fire sources. Fire and Materials, 17, 111e117. Sozogaku. (2013). Failure knowledge database [Online]. Available: http://www. sozogaku.com/fkd/en/cfen/CC1000062.html Accessed 18.08.13. Tauseef, S., Rashtchian, D., & Abbasi, S. (2011a). CFD-based simulation of dense gas dispersion in presence of obstacles. Journal of Loss Prevention in the Process Industries, 24, 371e376. Tauseef, S., Abbasi, T., & Abbasi, S. (2011b). Development of a new chemical processindustry accident database to assist in past accident analysis. Journal of Loss Prevention in the Process Industries, 24, 426e431. The Hindu. (2009). Big fire in IOC depot near Jaipur [Online], Jaipur. Available: http://www.thehindu.com/news/states/other-states/ioc-fire-may-abate-onthursday/article43289.ece?css¼print Accessed 05.11.12. The Hindu. (2013). Major fire at IOC’s Hazira terminal [Online]. Available: http:// www.thehindu.com/news/national/major-fire-at-iocs-hazira-terminal/ article4276397.ece Accessed 12.05.13. Thomas, P. (1963). The size of flames from natural fires. In Symposium (International) on Combustion, 1963 (pp. 844e859). Elsevier. Thomas, P., Baldwin, R., & Heselden, A. (1965). Buoyant diffusion flames: some measurements of air entrainment, heat transfer, and flame merging. Symposium (International) on Combustion. Elsevier. Thomas, P., Baldwin, R., Theobald, C., & Britain, G. (1968). Some model-scale experiments with multiple fires. Fire Research Station. Thomas, P., Simms, D., & Wraight, H. (1964). Fire spread in wooden cribs. Fire Research Station. Tieszen, S., O’Hern, T., Weckman, E., & Schefer, R. (2004). Experimental study of the effect of fuel mass flux on a 1-m-diameter methane fire and comparison with a hydrogen fire. Combustion and Flame, 139, 126e141. TNO. (1997). Methods for the calculation of physical effects due to releases of hazardous materials (liquids and gases) (3rd ed). Netherlands, CPR 14E: TNO. Vela, I., Chun, H., Mishra, K. B., Gawlowski, M., Sudhoff, P., Rudolph, M., et al. (2009). Prediction of the thermal radiation of large hydrocarbons and peroxide pool fires by CFD simulation. In Vorhersage der thermischen Strahlung großer Kohlenwasserstoff- und Peroxid-Poolfeuer mit CFD Simulation (Vol. 73). Vincent, J., & Gollahalli, S. (1995). An experimental study of the interaction of multiple liquid pool fires. Journal of Energy Research and Technology, 117. Wang, Z., Jia, F., Galea, E. R., Patel, M. K., & Ewer, J. (2001). Simulating one of the CIB W14 round robin test cases using the SMARTFIRE fire field model. Fire Safety Journal, 36, 661e677. Waterman, T., Labes, W., Salzberg, F., Tamney, J., & Vodvarka, F. (1964). Prediction of fire damage to installations and built-up areas from nuclear weapons. Final Report, Phase III, Experimental Studies, Appendices AG. IIT Research Institute Report for National Military Command System Support Center. Contract No. DCA-8. Weng, W. G., Kamikawa, D., Fukuda, Y., Hasemi, Y., & Kagiya, K. (2004). Study on flame height of merged flame from multiple fire sources. Combustion Science and Technology, 176, 2105e2123. Whelan, R. J. (1995). The ecology of fire. Cambridge University Press. Wikipedia. (2012). Jaipur (Indian Oil) fire [Online]. Available: http://en.wikipedia. org/wiki/(2009)_Jaipur_fire Accessed 13.01.12. Woods, J. A. R., Fleck, B. A., & Kostiuk, L. W. (2006). Effects of transverse air flow on burning rates of rectangular methanol pool fires. Combustion and Flame, 146, 379e390. Zukoski, E., Kubota, T., & Cetegen, B. (1981). Entrainment in fire plumes. Fire Safety Journal, 3, 107e121. Zalosh, R. G. (2003). Industrial fire protection engineering, West Sussex. Wiley.