Model Scale Fire Tests on a Vehicle Deck on Board a Ship

Full text

(1)Ida Larsson Haukur Ingason Magnus Arvidson. Model Scale Fire Tests on a Vehicle Deck on Board a Ship. SP Swedish National Testing and Research Institute SP Fire Technology SP REPORT 2002:05.

(2) Ida Larsson Haukur Ingason Magnus Arvidson. Model Scale Fire Tests on a Vehicle Deck on Board a Ship. SP Swedish National Testing and Research Institute SP Fire Technology SP REPORT 2002:05.

(3) 2. Abstract This report describes model scale tests conducted to investigate the fire development on a Ro-Ro ferry vehicle deck. The objective was to investigate the conditions that can arise during a fire on a vehicle deck and to suggest suitable tactical measures when fighting the fire manually. The tests mainly focused on the effect of natural and mechanical ventilation as well as the effects of sprinklers on the fire development. The model used was constructed in scale 1:8 of a typical Ro-Ro vehicle deck. Its size, 11,4 m x 2.8 m and 0,6 m (length x width x height) correspond to an actual vehicle deck of about 90 m x 22 m x 5 m. The model incorporated stairwells, ventilation shafts, large door openings, scuppers (drainage for water), a fan to provide about ten air changes per hour and a sprinkler system. The fires source consisted of a wood crib, equivalent when scaled up, to a burning truck with a fire output of about 70 MW.. Key words: Vehicle deck, ships, model scale fire tests Sökord: Fartyg, brand, modellskala, försök. SP Sveriges Provnings- och Forskningsinstitut SP Rapport 2002:05 ISBN 91-7848-893-1 ISSN 0284-5172 Borås 2002. SP Swedish National Testing and Research Institute SP Report 2002:05 Postal address: Box 857, SE-501 15 BORÅS, Sweden Telephone: +46 33 16 50 00 Telefax: +46 33 13 55 02 E-mail: info@sp.se Internet: www.sp.se.

(4) 3. Table of contents Abstract. 2. Table of contents. 3. Preface. 4. Sammanfattning. 5. 1 1.1 1.2 1.2.1 1.2.2 1.2.3 1.2.4 1.2.5. Introduction Background Current IMO requirements Structure Fire detection systems Fire extinguishing systems The ventilation system Precaution against ignition of flammable vapours. 7 7 8 8 8 9 9 10. 2 2.1 2.2 2.2.1 2.2.2 2.2.3 2.2.4 2.3 2.4 2.5. Description of the tests The test hall The scale model The water spray system The scuppers The fan The ventilation shaft and the staircase The fire source Instrumentation and documentation The test procedure. 11 11 11 12 13 13 13 13 14 14. 3. Analysis of the test results. 16. 4 4.1 4.2. Theoretical calculations Comparison with the experiments Large scale vehicle decks. 21 21 24. 5. Answers to the questions raised. 28. 6. Conclusions. 30. 7. References. 32. Appendix A – A detailed analysis of the test results A.1 Effect of ventilation on the heat release rate A.2 Smoke layer height A.3 Gas concentrations A.4 Gas temperatures A.5 Radiation. 33 33 36 37 38 39. Appendix B – Theoretical aspects. 41. Appendix C – Scaling laws. 48.

(5) 4. Preface SP Fire Technology has carried out a series of model-scale tests of fires on a Ro-Ro vehicle deck. In order to preserve a high quality of the work a consultative advisory group was established. The group consisted of the following people: Lars Adrian, Räddningstjänsten Göteborg Carl Christensson, Svenska Skum AB Sören Fogelström, Sveriges Fartygsbefälsförening Sten Gattberg, Stiftelsen Sveriges Sjömanshus Pelle Hybring, Räddningstjänsten Göteborg Hans Holmqvist, Sjöbefälsförbundet Krister Ingvarson, Sjöfartsverket Roger Karlsson, MARINVEST Jonas Lyborg, Sveriges Ångfartygs Assurans Förening Staffan Ålund, Sjöfartens Brandskyddskommitté We gratefully acknowledge the input from the embers of the advisory group. The authors would also like to acknowledge the free-lance journalist Claes Hindenfeldt for his contribution to the summary in Swedish. The project was financed by the Swedish Agency for Innovation Systems (VINNOVA)..

(6) 5. Sammanfattning Fartygsbränder på bildäck är ovanliga, trots att riskerna är stora. På bildäck står långtradare med varierande laster som ensamma eller tillsammans kan ge upphov till mycket intensiva bränder. Där finns också personbilar med eller utan husvagn som ofta har en gasoltub. En brinnande trailer med matolja bredvid en buss eller en trailer är ingen önskedröm ut bekämpningssynpunkt, särskilt som trailer och bussar avger mycket giftiga gaser när de brinner. En brand på bildäck på ett fartyg kan därför få katastrofala följder och kan leda till totalhaveri. Ett rökfyllt däck med dödligt giftiga gaser och stigande temperatur är vad som kan möta besättning och räddningspersonal om olyckan är framme. Ett bildäck är ofta mellan 100 och 200 meter långt och cirka 30 meter brett med öppning i för och akter och kan därför snabbt rökfyllas, uppskattningsvis inom 10 minuter. Syftet med projektet var att undersöka konsekvenserna av att det brinner och vad besättningen kan förvänta sig. Hur syretillförsel och ventilation påverkar brandutvecklingen samt om, och i så fall hur, den ska angripas. Några av frågeställningarna som studerades var: •. Inverkan av den mekaniska ventilationen på brandförloppet.. •. Inverkan på brandförloppet av mindre öppningar såsom dörrar till trapphus, ventilationsschakt och scuppers.. •. Inverkan på brandförloppet av större öppningar såsom port i akterskeppet.. •. Rumsvolymens inverkan på brandens storlek.. •. Genomsnittlig gastemperaturer i hela volymen och direkt ovanför branden.. •. Inverkan av sprinkler.. För att kunna genomföra nödvändiga tester byggdes en modell av ett bildäck i skala 1:8. Modellen hade måttena 11,4 m x 2,8 m x 0,6 m (längd x bredd x höjd) vilket i verkligheten skulle motsvara ca 90 m x 22 m x 5 m. Många färjor idag har betydligt längre bildäck, ibland uppåt 200 m, men på grund av platsutrymme i SPs brandhall kunde modellen inte byggas större. Modellen utrustades också med trapphus, ventilationsschakt, scuppers (dränering för vatten), en fläkt som gav en luftomsättning på 10 luftomsättningar per timme, ett sprinklersystem och en rad fönster för observationer under försöken. Brandkällan i dessa försök utgjordes av en träribbstapel som i skalmodell motsvarar en brinnande lastbil med en brandeffekt på ca 70 MW. Vid några försök undersöktes även brandspridningen mellan fordon och i dessa försök placerades flera träribbstaplar bredvid varandra. Totalt genomfördes 18 försök där olika parametrar varierades. Nedan ges en sammanställning av de slutsatser som vi har dragit ifrån projektet: •. Försöken visar att ventilation och syretillförsel är helt avgörande för brandförloppet på bildäck på ett fartygs. Trots att bränder på slutna bildäck kan bli ventilationskontrollerade så blir de initialt mycket stora och kan utgöra en fara för besättning och passagerare..

(7) 6. •. Försöken visar också att risken för en stor brand och ett totalhaveri är större under lastning och lossning med fartygets portar öppna än under gång.. •. Lågorna spreds mycket snabbt längs taket och att hela bildäcket rökfylldes inom två och en halv minut (sju minuter i full skala) och de giftiga gaserna skapar ett livshotande läge. I det verkliga fallet så kommer flammorna att spridas i utrymmet mellan långtradarnas kapell/tak och taket, ungefär som en toppskogsbrand.. •. Försöken visade att trapphus och ventilationsschakt är för små för att tillföra tillräckligt med syre för att försörja en stor brand. Vid ett försök fick man en pulserande brand. När branden ökade minskade syremängden och branden tenderade att självslockna. När syremängden sedan ökade tog sig branden igen med hög intensitet.. •. Sprinklersystemen var till stor nytta och reducerade släckningstiderna avsevärt. I försöken var emellertid inte vattentryck och vattendroppstorlek skalenliga så släckningsförmågan är troligtvis betydligt sämre i verkligheten. Den kyleffekt ett riktigt sprinklersystemet skulle kunna ge bedöms dock vara värdefull.. •. De enda gångerna det blev en fullt utvecklad brand var när en kortsida (motsvarande en port i akterskeppet) var öppen och medgav god syretillförsel. Riskerna för en stor brand är därför betydligt större vid lastning och lossning än under gång.. •. Resultaten från försöken har använts för att genomföra beräkningar av brandförlopp och brandgastemperaturer för ett verkligt bildäck med måtten 180 m x 30 m x 5 m (höjd). Beräkningarna visar att en ökad volym avsevärt ökar den maximala brandeffekten (når upp till nära 80 MW) och att den genomsnittliga brandgastemperaturen blir hög, 250 – 300 °C. Direkt ovanför branden kommer temperaturen och värmestrålningen att vara mycket hög.. •. Praktiska råd är beroende av om fartyget är till sjöss, eller ligger vid kaj med portarna öppna. Till sjöss gäller att snabbt aktivera sprinklersystemen, som därmed kyler ner brandgaserna och sänker temperaturen. Därefter kan man bedöma möjligheterna att göra en manuell insats. Det är viktigt att inte ventilera och beträda däcket innan brandgaserna är nedkylda. Risken är annars att branden, som sannolikt avtagit på grund av syrebrist och sprinkler, startar igen. Konsekvensen blir snabbare förlopp och högre intensitet på grund av den värme som lagrats i omgivande konstruktioner.. •. En brand vid lastning och lossning är betydligt mera kritisk, eftersom bildäcket behöver utrymmas och branden inte kommer att bli ventilationskontrollerad i första taget. Om inte besättningen eller räddningstjänsten kan kontrollera branden kan hettan från branden medföra brandspridning till andra delar av fartyget.. •. Det är värt att betona att ovanstående slutsatser är baserade på att bildäckets integritet bibehålls. Det kan finnas fall, till exempel om en gasolflaska exploderar, där det kan uppstå så pass stora hål att branden blir mycket intensiv..

(8) 7. 1. Introduction. Although fires on vehicle decks onboard ships are rare, the consequences can be disastrous. Trucks or trailers are parked closely together, carrying a wide range of loads, which, single or together, can burn intensively. There are probably also private cars, with or without caravans, which often carry an LPG bottle. A burning tanker carrying for example edible oil, parked adjacent to a bus or trailer is something of a nightmare from the point of view of fighting the fire, particularly as trailers and buses can generate highly toxic gases when burning. A fire on a Ro-Ro car deck (vehicle deck) can therefore create catastrophic consequences, and may even result in loss of the ship. Once a fire has started, crew and rescue personnel attempting to fight it can be faced with zero visibility from smoke and toxic gases and rising temperatures. A typical vehicle deck, often between 100 - 200 m long and circa 30 m wide and with openings at bow and stern, can quickly be filled with smoke.. 1.1. Background. Today’s knowledge about heat release rates from fires in single vehicles is fairly good. Tests have shown that a standard car burn with a heat release rate of up to approximately 5 MW, a small truck up to 15 MW and a large heavy goods vehicles over 45 MW [1]. However, such measurements have only been conducted during normal ventilation conditions. A heavy goods truck on fire inside a tunnel with a high degree of ventilation (air speeds typically of 5 - 6 m/s) could develop heat release rates up to 120 MW [2]. In opposite we know very little how fire spread between vehicles and which size fire that could occur when vehicles are parked close together. On the vehicle deck on board a large Ro-Ro ferry there could be a few dozens of heavy goods trucks and busses and over fifty cars. The fire load is therefore very high but the access of oxygen can be limited. During loading and unloading, ports are open, which could affect the fire scenario. A possible fire scenario immensely depends on the ventilation of the deck. If the vehicle deck is closed when the fire starts, the oxygen will successively be consumed. If no fresh air is supplied, the fire decreases considerably and the fire will probably self-extinguish. If the mechanical ventilation runs at the start of the fire or if a door stands open to a staircase the result could be different. The heat from the fire rises upwards in the staircases and if there are openings towards the surroundings on upper deck, a very dangerous situation could occur. Natural ventilation then supports the fire with oxygen. It is therefore very important that we gain knowledge of how to handle a fire on a vehicle deck at different ventilation conditions. Today we don’t know how large the fires can become on vehicle decks. We know that if the fire is not controlled, there is a great risk for spread between the vehicles. How fast the fire will spread is not very well known. The objective of this research project is to simulate possible fire scenarios that could occur on vehicle deck and examine what parameters can affect the fire development. By understanding possible fire scenarios, important information can be obtained to form a suitably tactic for fire fighting on vehicle deck. Some of the questions that were asked before this research project were: • •. Is it preferable to seal the vehicle deck at a fire and let the water spray system cool the fire or should manual extinguishments be undertaken? What are the consequences if a door is opened to the deck during a fire?.

(9) 8. • • •. Should the mechanical ventilation be switched off and in that case, when? How high can the fire gas temperatures reach? How is the fire affected if the deck is divided into sections?. These questions cannot be fully answered by the experimental test series carried out here. The results can, however, provide qualitative knowledge of where the problems may be. Based on this information, preliminary judgments regarding tactics can be made.. 1.2. Current IMO requirements. The basics for the fire protection of Ro-Ro vehicle decks of passenger ships is given in Chapter II/2, Part B of the SOLAS [3] convention. The spaces are denoted special category spaces. For High Speed Crafts (HSC) a special code of safety has been developed [4].. 1.2.1. Structure. The fire integrity of bulkheads and decks for special category spaces is covered under Regulation 37. For bulkheads, the division is required to be of steel and capable of being closed and reasonably tight. For the decks, the division is dependent on the category of the adjacent space, for example shall divisions between Ro-Ro cargo spaces and control spaces be A-60, service spaces A-30 and machinery spaces of category A, A-60. High Speed Crafts (HSC) differs from conventional vessels at the construction is mainly made from lightweight materials, such as aluminium alloys.. 1.2.2. Fire detection systems. To detect fires, a fixed fire protection system, complying with Regulation 13, is required in the space. Both heat, smoke and detectors combining these features are allowed, however, smoke detectors are not required. Flame detectors are only permitted in addition to heat and smoke detectors. The detection system shall be divided into sections. Unless it is accepted by the administration, the sections are not allowed to cover more than one side of the ship. The activation of a detector or manually operated call point shall initiate a visual and audible signal at a control panel located on the navigation bridge or in the main fire control station. In addition, the section in which a detector or manually operated call point operated shall be denoted on an indicating unit. If signals have not received attention within two minutes, an audible alarm shall automatically sounded through the crew accommodation and service spaces, control stations and machinery spaces. It is worthwhile noticing that a fire detection system is not allowed to be used for any other purpose than the closing of fire doors and similar functions. The requirements for HSC regarding fire detection are with a few exceptions identical..

(10) 9. 1.2.3. Fire extinguishing systems. In the Annex of Resolution A.123(V), Recommendation on fixed fire extinguishing systems for special category spaces [5] from 1967, details are given for an acceptable installation for the protection of vehicle decks. The SOLAS and HSC Code references to Resolution A.123(V) have been identified and are as follows, se table 1. Table 1. SOLAS and HSC Code references to Resolution A.123(V). REG. II-2/37.1.3 Reg. II-2/53.2.2 Reg. II-2/54.2.1.3 Reg. II-2/54.2.9 MSC 36(63) 7.8.2. SPECIAL CATEGORY SPACES ON PASSENGER SHIPS Ro-ro cargo spaces on cargo ships Underdeck cargo spaces on cargo ships carrying dangerous goods Ro-ro cargo spaces on cargo ships carrying dangerous goods Special category spaces on high speed craft. The fire extinguishing system described shall provide an average water application rate of at least 3,5 mm/min for spaces with a deck height not exceeding 2,5 m and at least 5 mm/min for deck heights in excess of that. It is also required that the system covers the full width of the vehicle deck, but it may be divided into sections of at least 20 m in length. An exception where the requirement of the full coverage may be accordingly reduced, is where the deck is subdivided with longitudinal “A” Class divisions, forming boundaries of staircases, etc. The water pump(s) are required to be capable of a simultaneous supply of all the nozzles on the vehicle deck or at least to sections. Distribution valves shall be located outside the vehicle deck and the water pump(s) are required to be possible to operate by remote control, from the space where the distribution valves are located. To avoid that a free water surface is built up, necessary provisions shall be taken with regard to drainage and pump arrangement of the deck. In addition to fixed fire extinguishing systems, one portable fire extinguisher shall be located at each access way of the space. It is also required that at least three water fog applicators and one portable foam applicator is available on the ship. For fixed fire extinguishing systems on HSC the same requirement apply, the requirement for portable fire extinguishing equipment is through slightly different.. 1.2.4. The ventilation system. Vehicle decks shall be provided with a separate ventilation system, sufficient to provide at least 10 air changes per hour (based on the gross volume). The ventilation fans shall normally be run continuously. However, if this is impractical, daily, after which the space shall be proven gas free. In case of fire, it is required that it should be possible to rapidly be able to shut down and effectively closure the ventilation system. On HSC the ventilation system shall provide for at least 10 air changes per hour while navigating and 20 air changes per hour at the quayside during loading and unloading..

(11) 10. 1.2.5. Precaution against ignition of flammable vapours. To prevent ignition of flammable vapours, Regulation 37 also contains requirements on the electrical equipment and wiring shall be of a type suitable for use in explosive petrol and air mixtures. This is also required in exhaust ventilation ducts. Other equipment, which may constitute a source of ignition, is not permitted. The same principles are used on HSC..

(12) 11. 2. Description of the tests. 2.1. The test hall. The test was conducted at SPs fire test laboratory located in Borås, Sweden. The test hall, where the tests took place, measures 13,2 m by 8,8 m with a height of approximately 8,5 m. The facility is equipped with a ventilation system with air inlet at floor level and air exhaust in the ceiling of the building. During the tests, the ventilation system was run at a speed which provided an air exchange of approximately 40 000 m3 per hour.. 2.2. The scale model. To be able to test different fire scenarios on a vehicle deck, a model in scale 1:8 was built. The model measured 11,425 m by 2,786 m by 0,625 m (length, width and height), which are equivalent to approximately 91 m by 22 m by 5 m in large scale. The length was somewhat short as compared to a real vehicle deck that can measure up to 200 m but the lack of space in the fire test hall limited the possibilities to build larger. This should however, not have any significant affects on the results of the fire tests.. Figure 1. A schematic picture of the exterior of the scale model.. Figure 2. The location of the fire source and the instrumentation. TC indicates the location of thermocouples whereas O2, CO2 and CO indicates the location of the gas sample probe. All instruments were located at the centreline of the model..

(13) 12. The walls, the ceiling and the floor of the scale model consisted of 10 mm or 12 mm nominally thick Promatect H boards (calcium silica boards). Promatect H is a fire resistant material that also withstands water relatively well, something that was necessary when using a water spray system. The thermal data of the board is as follows; thermal conductivity: 0,19 W/m K, density: 870 kg/m3 and heat capacity: 1130 J/kg K. The main reason for using Promatect H is its strength during fire testing. It does not fulfil the thermal inertia requirements of scaling (Promatect H -> steel). By using two-zone computer model we were able to convert the model scale test results to large-scale situation since the computer model can calculate the temperature rise in the surrounding structure. The Promatect H boards had a size of 1200 mm by 3000 mm and the floor and ceiling were therefore built in sections of these dimensions. The scale model was also equipped with a staircase, a ventilation shaft, scuppers (openings for drainage of the deck), and a fan that provided an air change of 10 times the volume per hour of the deck, a water spray system and windows for insight during the tests.. 2.2.1. The water spray system. Under the isolated ceiling, a carbon steel pipe-work was arranged. The pipe-work consisted of two 25 mm main lines with sixteen 17 mm branch lines on each main line. The two main lines were then connected outside the model and were fed from a pressure tank.. Figure 3. The layout of the sprinkler system.. The water spray system had a nozzle-to-nozzle spacing of 300 mm by 300 m (equivalent to 2,4 m by 2,4 m in large scale) and the nozzles were located in pendent position at the ceiling over the area of the wood crib. A total of 64 Lechler 212.245.11.CC nozzles were installed. The vertical distance from the ceiling to the tips of the nozzles was approximately 40 mm (equivalent to 320 mm in large scale). The vertical distance from the tips of the nozzles to top of the wood crib was approximately 80 mm, which is equivalent to 640 mm in large scale..

(14) 13. The pressure tank had a pressure of approximately 2,16 bar, which provided an average total water flow of 10,2 L/min (). This water flow is equivalent to a discharge density of 5 L/(min, m2) in large scale.. 2.2.2. The scuppers. On every vehicle deck there are a number of drainage openings, “scuppers” for water drainage. In practice, the capacity of the scuppers is designed for the total water flow rate of the water spray system and the outlets are equipped with check valves so that seawater cannot flow in to the deck. On the contrary, these check valves do not prevent fresh air from entering the deck, which potentially could influence the behaviour of a fire. In the scale model, instead of drilling a large number of holes along the deck, the total area of the holes were added up together and only two larger holes were made. The two holes had an area of 0,0083 m2, which is equivalent to 30 scuppers with a diameter of 150 mm in large scale.. 2.2.3. The fan. In the scale model a fan was installed. The volume flow rate of the fan was adjusted to provide an air exchange rate of 10 times the volume per hour, which is the minimum requirement according to the SOLAS convention.. 2.2.4. The ventilation shaft and the staircase. One ventilation shaft was built. The shaft had a height of 2,8 meters which is equivalent to 22,4 m in large scale. This height corresponds to a ventilation shaft that reaches from the vehicle deck and ends three decks above. The ventilation shaft had a diameter of 160 mm. This measure was chosen so that the speed of the smoke through the shaft wouldn’t be unrealistically fast. The staircase was built of Promatect H boards and went from vehicle deck to the deck above. The vertical distance from the top of the lower door opening to the bottom of the upper door opening was approximately 0,475 m.. 2.3. The fire source. The fire source in the tests consisted of a wood crib. The crib was approximately 1,15 m long, 0,35 m wide and 0,36 m high and in free burning tests the measured heat release rate (HRR) was approximately 400 kW. In large scale this HRR is equivalent to about 70 MW. The wood crib was located between section 4 and 5 (from the right hand side in the drawings, see Figure 1) and in order to protect the ceiling above the crib was insulated with 50 mm of mineral wool. To ignite the wood crib, two trays (16 cm by 16 cm) with a total area of 500 cm2 were filled with 250 ml of heptane. The trays were abutted together, centrally under the wood crib on a Promatect H board. The distance from the wood crib to the trays was approximately 5 cm. The stand penetrated through the floor of the model where it was placed on a load cell. With the load cell the reduction in weight caused by the fire could be monitored during the tests..

(15) 14. Figure 4. The fire source consisted of wood cribs.. 2.4. Instrumentation and documentation. During the tests, gas temperatures were measured at the ceiling, directly above the wood crib. In addition, gas temperatures were measured inside the staircase and at different distances from the fire source, see Figure 2. All ceiling gas temperatures were measured using wire thermocouples. In addition, a Plate Thermometer was located at the ceiling, directly above the wood crib. The ceiling gas thermocouples were located 150 mm below the ceiling, at the centreline of the long sides of the model, in the West – East direction, see Figure 2. In addition, the total water flow rate, the water pressure and the weight loss of the wood crib was measured. All thermocouples were of type K (chromel-alumel) and made from 0,5 mm wire welded together. Additionally, the ceiling surface temperature was measured using a Plate Thermometer. The Plate Thermometer was developed at SP and consists of a 100 mm by 100 mm by 0,7 mm thick plate, insulated at the backside. A thermocouple is welded to the middle of the plate. The Plate Thermometer primarily responds to radiation, and to a lower degree, to convection. The thermocouples, the pressure transducers and the water flow meter were connected to a Solotron Orion logger. The data was recorded at a rate of about one scan per second. The test was also documented using two video cameras, one camera located at the short side window, and the second camera located at the middle of the long side, directed towards the wood crib.. 2.5. The test procedure. The test procedure was as follows. The heptane (250 ml) was poured into the trays, the data acquisition system was started and the heptane was lit after two minutes. In the tests where the water spray system was used, the fire was allowed to develop in the wood crib and after 03:00 (min:sec), measured from the ignition, the valve to the system was opened allowing water to the nozzles..

(16) 15. The water was left on to the end of the test. In all scenarios where the water spray system was used it managed to extinguish the fire. A total of eighteen tests were performed in the model. The different variables that were used in the test were: • • • • • • • •. Fan (on/off) The sprinkler system (on/off) The door to the staircase (opened/closed) Ventilation shaft (opened/closed) Gate (opened/closed) Two windows with the dimensions 250 x 500 mm on each short side of the model (opened/closed) Opening with the dimension 500 x 50 mm, located next to the staircase (opened/closed) Scuppers (opened/closed). Table 1 The test protocol listing all the parameters varied within the test series. Test no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 164 175 186. 1. Fan (On/Off) Off On Off Off On Off On Off Off Off Off Off On Off Off Off Off Off. Sprinkler (On/Off) Off Off Off Off Off Off Off Off Off On3 On On On On On On Off Off. Staircase (Op/Cl) Closed Closed Open Closed Closed Closed Closed Open1 Closed Closed Closed Closed Closed Closed Closed Closed Closed Closed. Ventilation shaft (Op/Cl) Closed Closed Closed Open Open Closed Closed Open Closed Open Closed Closed Closed Open Closed Closed Closed Closed. Gate (Op/Cl) Closed Closed Closed Closed Closed Closed Closed Closed Open Closed Open Closed Closed Closed Closed Open Closed Closed. Window (Op/Cl) Closed Closed Closed Closed Closed Open Closed Open2 Closed Closed Closed Open Closed Closed Open Closed Closed Closed. Opening (Op/Cl) Closed Open Closed Closed Closed Closed Closed Closed Closed Closed Closed Closed Closed Closed Closed Closed Closed Closed. Scuppers (Op/Cl) Closed Closed Open Open Open Closed Open Open Open Open Open Closed Closed Open Closed Open Closed Closed. The door to the staircase is opened when the fire have spread properly, at 02:05 (min:sec) after ignition In this test, only one window is opened, the window located on the long side of the scale model, near the fire source. This occurred at 06:17 (min:sec) after ignition. 3 The sprinkler system is turned on when TC 5 reaches 800ºC. Otherwise the sprinkler system is turned on 3 minutes after ignition. 4 Three wood cribs are located adjacent to each other to simulate fire spread between vehicles. 5 The vehicle deck is divided into sections. Three wood cribs are located adjacent to each other within the 5 m long section. 6 A free burning test, but with reflected radiation. At the ceiling, all modules are removed except for the two located just above the wood crib. The gables on the model are also removed. 2.

(17) 16. 3. Analysis of the test results. In the majority of the eighteen model tests, the fire either became ventilation controlled or was self-extinguished due to the low oxygen concentration within the hot smoke layer. This occurred after the hot smoke layer had descended down to floor level. The oxygen concentration measured at the upper part of the hot gas layer at that time was as low as 13 – 15 %. At the lower part we estimated the oxygen concentration to be in the range of 15 – 17 %. The maximum carbon dioxide (CO2) measured in the upper layer was about 6 – 7 % and the carbon monoxide (CO) concentration was about 0,4 %. In Figure 5 we can observe the heat release rate development for different nonsprinklered tests. The test number refers to the tests summarized in Table 1. In Figure 6 we can see how the oxygen concentration varies as a function of time for tests with natural ventilation (openings of different sizes).. Heat Release Rate (kW). 500. test 03 test 04 test 05 test 06 test 07 test 08 test 09 test 17 test 19. 400. 300. 200. 100. 0 0. 5. 10. 15. 20. 25. time (min). Figure 5. The measured heat release rate as a function of time from ignition. Test 19 is the free burning test under a calorimeter and test 9 is the test where one portal (short end wall) was open.. It took about two minutes for the smoke to fill the entire volume and the fire growth rate started to decelerate less than a minute afterwards (approximately when the oxygen concentration was 15 %). The self-extinguishment phase started with a sudden reduction in the gas temperature above the fire source. This occurred after about four to five minutes from ignition and about two minutes later we were able to verify that the fire was extinguished by examining the mass loss rate curves. Therfore we can say that the self-extinguishment occurred after about six to seven minutes from ignition. In large scale, for a vehicle deck of the same size, this would correspond to 17 to 20 minutes from ignition. This presumes that the fire growth rate is comparable in both scales. The maximum average heat release rate obtained when the ventilation rate was restricted was 181 kW, which corresponds to 33 MW in large scale..

(18) 17. 22. test 01 test 03 test 04 test 06 test 08 test 09. Fire extinguished. Oxygen concentration (%). 20. 18. 16. 14. 12. 10 0. 5. 10. 15. 20. 25. Time (min). Figure 6. The oxygen concentration measured 0.475 m above floor level for naturally ventilated fires except for test 01 which had no openings.. In test 17 we wanted to investigate the influence of compartmentisation of the vehicle deck. Test 17 was performed in a 5 m long section of the scale model (total volume of 8,7 m3). The compartment created was less than half the size of the normal test rig (19,85 m3). The maximum heat release rate that was obtained was 126 kW (corresponds to 23 MW). As expected the volume of the enclosure appears to be a parameter that can influence the maximum heat release rate. This is an important observation as vehicle decks usually are quite large, and in most cases larger than tested here. Thus, in cases when we have normal ventilation, we can expect fires on vehicle decks that are a great deal larger than the ones obtained in the test series carried out here. Flames spread very rapidly along the ceiling. The maximum radial length of the flames along the ceiling was about 1.4 m (11 m in large scale) from the centre of the initial fire source. In reality, the flames would fill the shallow space between the ceiling above and the top of the trucks, and would spread between vehicles in the same way as a forest crown fire. The maximum gas temperature above the fire was in the range of 800 to 1000 ºC, whereas the maximum gas temperature in the hot smoke layer was about 170 ºC. According to the scaling laws given in Appendix C, we should expect the same gas temperatures in both scales. This is only possible if the thermal properties of the surrounding walls were correctly scaled but that was not the case here. Therefore, we should expect slightly higher gas temperatures in large scale than 170 oC. In the case of well-ventilated fires (one portal open) the fire became fully developed (not ventilation controlled) and we were forced to extinguish the fire after about 13 minutes from ignition. This situation can arise during loading and unloading of the vehicle deck. The temperature measurements above the fire source are shown in Figure 7. The gas temperature in the hot gas layer was measured both 3 m and 5.4 m from the fire source. In Figure 8 the measured gas temperature 3 m from the fire source and 0.475 m.

(19) 18. from the floor level (0.15 m under the ceiling) is presented. The gas temperature falls very rapidly only a few meters from the initial fire source. This indicates that if we obtain a ventilation-controlled fire the gas temperature will not be very high further away from the initial fire. We can, however, expect that the local steel temperature at the ceiling may become high (assuming no-insulation in the ceiling) and therefore creating a risk for further fire spread. This conclusion is, however, highly dependent on the fire duration.. test 01 test 03 test 04 test 06 test 08 test 09. 1200. o. Ceiling temperature ( C). 1000. 800. 600. 400. 200 fire extinguished 0 0. 5. 10. 15. 20. 25. time (min). Figure 7. The measured temperature directly above the fire source for naturally ventilated fire tests.. 400 350. Fire extinguished. test 01 test 03 test 04 test 06 test 08 test 09. o. Upper Layer Temp ( C). 300 250 200 window open test 08. 150 100 50 0 0. 5. 10. 15. 20. 25. time min. Figure 8. The gas temperature in the hot gas layer measured 3 m from the fire source (TC04) and 0.15 m from the ceiling..

(20) 19. If the fire become ventilation controlled as in the present case there will probably be no problems with the steel temperatures. If we have larger vehicle decks the fire duration may become somewhat longer resulting in higher steel temperatures. As long as we have a limited degree of ventilation and the sprinkler system are able reduce the ceiling temperature we should not consider this as a major problem. If the ventilation becomes stronger than in the present study and if the sprinkler system fails to fully control the fire, we may expect great problems. During one of the tests with mechanical ventilation, test 02, the fire seemed to pulsate and oscillate around a particular equilibrium. As the intensity of the fire increased, the amount of available oxygen decreased, and the intensity of the fire declined. At this lower intensity, it consumed less oxygen, so that after a while the fire increased again. The flow rate exchange during this test was ten air changes per hour. In the other tests with mechanical ventilation no oscillation phenomena were observed. What is specific with is that the door opening at the stairwell, which measured 5 cm wide and 50 cm high, was open. This situation may create problems for the on board staff as the situation become highly dependent on the ventilation rate. In general we can say that ventilationcontrolled fires may create large problems for the on board rescue team as the fire will easily accelerate again if something will fail or if somebody makes wrong decision about the ventilation. In this situation it would be preferable to start the water sprinklers in order to reduce the temperature and reduce the risk for re-ignition. Numerous tests were carried out using a simplified sprinkler system. The sprinkler system was usually started three minutes after ignition. This would correspond to 8.5 minutes in large scale. The sprinkler system that was installed was found to be very effective. This can be observed in Figure 9 and Figure 10. The sprinkler system substantially reduced the extinguishing time, relative to that of the ventilation-controlled fires, or the one which self-extinguished. However, in the open portal case (test 16), the sprinklers controlled the fire, but did not extinguish it entirely as can be discerned in Figure 10. In that particular test we used three wood cribs. It must be remembered, though, that it was possible only to scale down the water flow rate in the sprinkler system. Due to practical reasons we were not able to scale down the pressure, or the size of the water droplets, which is compulsory in order to obtain correct results. The consequence of this was that the performance of the scale model sprinkler system was better than the performance of a real sprinkler system would have been. The results, however, clearly show that a sprinkler system can cool down the temperatures effectively and thus facilitate for the on-board rescue team fighting the fire. Even smaller amount of water than would have done a very good job. The combination of a ventilation-controlled fire and a water sprinkler system appears to be a good solution for the tested situations..

(21) 20. 1000 Test 11 Test 12 Test 13 Test 14 Test 15 Test 16. o. Gas temperature ( C). 800. 600. 400. 200. 0 0. 5. 10. 15. Time (min). Figure 9. The gas temperature measured above the fires source for the sprinklered tests.. Test 11 Test 12 Test 13 Test 14 Test 15 Test 16. 150. o. Upper Layer Temp ( C). 200. 100. 50. 0 0. 5. 10. 15. Time (min). Figure 10. The gas temperature measured 3 m from the fire for the sprinklered tests..

(22) 21. 4. Theoretical calculations. The variation in the volume of the vehicle decks and the ventilation conditions for different types of ferries is large. Consequently, the experimental data obtained from the test series carried out here have only limited use as it corresponds to only one type of vehicle deck. Of practical reasons we had to make the model smaller than is customary. It is therefore of interest to convert the results from the model scale tests carried out here to the situation on board large scale ferries. There are different methods available. The simplest one is to use the scaling laws presented in Appendix C. We have used these scaling laws earlier in the text in order to give the reader an idea of what to expect concerning the heat release rates, gas concentrations, extinction times and gas temperatures at corresponding vehicle decks i.e. a vehicle decks that is approximately 90 m long, 20 m wide and 5 m high. The test results can therefore only be valid for vehicle decks comparable in size. Another way is to use simple mathematical models such as the ones presented in Appendix B. These can give rough estimates of the situation for different vehicle decks of different volumes and with different ventilation conditions. The third method and the one we will discuss further in this chapter are to use a computer model. There are different types of models available. We can have complex and accurate models such as CFD models [6] or we can have less accurate models such as zone models. For the purpose of this study we will use a simple zone model code called BRANZFIRE [7]. The basic principal of the model is that it calculate the temperature, gas concentrations and smoke layer height as a function of time for a given heat release rate, ventilation and enclosure geometry. It can also calculate the surface temperatures and the thermal radiation and pressure within the enclosure.. 4.1. Comparison with the experiments. Before we use the model to predict the conditions in other type of vessels we want to investigate the accuracy of the computer model. Therefore, we have compared the results from simulations with BRANZFIRE with two tests, test 04 and test 06, respectively. They were chosen as they represent two different conditions, a self-extinguished fire and a ventilation-controlled fire, respectively. The actual measured heat release rates were used in the simulations. In test 04, the ventilation shaft and the scuppers at floor level were open. The shaft was simulated by assuming a horizontal opening at the ceiling. After about 6 minutes from ignition in test 04 the fire self-extinguished. In test 06, two doors, 0.25 m wide and 0.5 m high, were opened at both end walls of the model. In test 06 the fire became ventilation-controlled. A comparison of the calculated and measured values for tests 04 and 06 are presented in Figure 11. The agreement appears to be reasonably good except that there is a slightly time shifting in the results, especially at initial stage of the fire. The calculated values tend to yield somewhat higher initial and maximum temperatures. In general we can conclude that the difference is acceptable for the purpose of this study..

(23) 22. Calculated test 04 Measured test 04 200. Calcuatated test 06. o. Upper Layer gastemperature ( C). Measured test 06. 150. 100. 50. 0 0. 5. 10. 15. 20. 25. Time (min). Figure 11. Comparison between calculated and measured gas temperature in the upper hot smoke layer. The measured values were taken the thermocouple 4, 3 m from the fire and 0.475 m above floor level.. The height of the smoke layer was also determined in the experiments. This height was converted to a non-dimensional height, h/H, by dividing the actual height, h, by the total ceiling height, H=0.625 m. The experimental results are shown in Figure 21 in Appendix A. In Figure 12 we can see a comparison between the calculated non-dimensional smoke layer height and the measured one for tests 04 and 06. The agreement is very good except at the floor level. In the experimental part the non-dimensional smoke layer height was estimated from temperature measurements at h/H=0.76, 0.48 and 0.24, assuming a smoke layer where the excess temperature rise was equal to or greater than 5 ºC. The time it took for the smoke layer to reach the floor level was estimated by assuming a linear relationship of the smoke descend between h/H=0.48 and h/H=0.24, see Appendix A for more detailed information. Apparently this procedure tends to yield overestimation in the time it took for the smoke to descend down to the floor level. One of the most important parameters in this study is the depletion of the oxygen within the volume of the vehicle deck. Thus, it is of great interest to compare the calculated and measured oxygen concentrations in the upper layer. In Figure 13 we can see that the agreement is excellent the first four minutes of the tests. Apparently, in test 04 there occurs a phenomenon in the experiments, which the theoretical model cannot consider. Probably, this has to do with the leakage of air into the scale model after the gas concentrations in the vicinity of the fire has reached levels that are close to the flammability limits..

(24) 23. Nondimensional smoke layer height, h/H. 1 Calculated test 04 0.8. Measured test 04 Calculated test 06 Measured test 06. 0.6. 0.4. 0.2. 0 0. 0.5. 1. 1.5. 2. 2.5. 3. time (min). Figure 12. Comparison between the calculated, non-dimensional smoke height and measured one for test 04 and test 06.. Calculated test 04. Oxygen concentration in the upper layer (%). 22. Measured test 04 Calculated test 06. 20. Measured test 06 18. 16. 14. 12. 10 0. 5. 10. 15. 20. 25. Time (min). Figure 13. Comparison between calculated and measured oxygen concentration in the upper hot smoke layer. The measured values were taken 5.2 m from the fire and 0.475 m from above level.. In general we can conclude that the agreement between the calculated and the measured values is satisfactory for the purpose of this study. Consequently, we can make calculations for fires on a large scale vehicle deck with good confidence in the results..

(25) 24. 4.2. Large scale vehicle decks. Now we want to compare the vehicle deck we have tested (in large scale) to other vehicle decks, especially those that are larger than the one we have tested. First we make a calculation using the scaled heat release rate from test 06, where we had a ventilationcontrolled fire. The corresponding large scale vehicle deck is 91.2 m long, 22.3 m wide and 5 m high. In order to convert the measured model scale heat release rate to large scale heat release rate we simply multiply the heat release rate in kW with the factor 85/2 and the time scale with 81/2. The scaling laws used can be obtained from Table C1 in Appendix C. In Figure 14 we can observe the scaled heat release rate and the calculated upper layer gas temperature for test 06. Interesting observation is that the upper layer gas temperature is about twice the temperature measured in the model scale test. According to the scaling laws the temperatures should be about the same in both scales assuming that we can scale the thermal response of the wall material. We did not succeed in doing that (see Appendix C) and that is reflected in our results. In the large scale simulation we assumed 50 mm of insulation of the ceiling and 5 mm thick steel sheet in the walls and at the floor. Since we know from comparison with experiments that the program does simulate quite accurately this situation, we can conclude that the temperature at the vehicle deck will be about twice the temperature obtained in the model scale tests.. 500. 40. 400. 30. 300. 20. 200. 10. 100. 0. o. 50. Gas temperature in upper layer ( C). Heat release rate (MW). In the simulated large scale vehicle deck we will obtain the maximum heat release rate, which is about 33 MW, which will reduce down to 10 – 20 MW during a period of 60 minutes. The average gas temperature within the hot gas layer is in the range of 250 ºC to 350 ºC and the thermal radiation flux down to the floor is about 4.4 kW/m2.. 0 0. 10. 20. 30. 40. 50. 60. 70. 80. time (min). Figure 14. Calculated heat release rate (left abscissa) and upper layer gas temperature (right abscissa) for the scaled test 06..

(26) 25. 25. 5. 20. 4. 15. 3. 10. 2. 5. 1. 0. Height of the hot smoke layer (m). Oxygen concentration (%). The oxygen concentration and the smoke layer heights are plotted in Figure 15 for the scaled test 06. As expected the oxygen concentration and the smoke layer height corresponds very well to the model scale results. This was expected since we know that the scaling law results for gas concentrations and smoke dynamics should correspond quite accurately. Thus, we have ensured ourselves that the results, even scaled, are reasonable and thus we can predict other situations for other vehicle decks using the program BRANZFIRE. This is an important conclusion, as we know it is difficult to perform large scale tests on vehicle decks. It is very expensive and it may be many practical problems involved.. 0 0. 10. 20. 30. 40. 50. 60. 70. 80. time (min). Figure 15. The calculated oxygen concentration and the height of the smoke layer for scaled test 06.. In the following we will make a calculation of the fire development in a much larger vehicle deck than the one scaled from the model tests. The volume of this large vehicle deck is 180 m long, 30 m wide and 5 m high. The volume is more than twice the one used earlier (91.2 m x 22.3 m x 5 m). We assume the same openings as in test 06, i.e. two doors that measures 2 m wide and 4 m high, respectively. Thus, the ventilation conditions are expected to be about the same but the volume is increased by factor of 2.6. This means that the maximum heat release rate will be higher before the fire starts to decelerate. In most of the tests the fire developed up to about 181 kW (average value) and shortly afterwards the fire started to decelerate until it either became ventilation controlled or self-extinguished. This process appeared to start when the average gas concentration in the hot smoke layer was about 15 %. Therefore we believe that in the case of ventilation controlled fires or when self-extinction occurs we can roughly estimate the maximum heat release rates with the aid of this information. This assumption requires that the hot smoke layer has filled up the volume entirely when the maximum heat release rate occurs. The advantage of using this information is that we can estimate the maximum heat release rate for different vehicle decks. A simple theoretical method to do this is outlined in Appendix B. The fire growth until the fire starts to decelerate can be described according to the following equation:.

(27) 26. Q(t ) = 0.23t 2. (1). where the constant was found by scaling up the measured heat release rate from test 06. This initial fire growth rate can be regarded as general for all the tests performed. This fire growth rate corresponds to an Ultra Fast fire according to the NFPA definition [8]. By assuming that the fire growth rate will continue to follow this growth rate until the average oxygen concentration has reach to a level of 15 % we can determine the time this will occur by numerically solve equation (B15). For the case presented here it will be t=490 sec which corresponds to a heat release rate of 55.3 MW. For the smaller volume the maximum heat release rate was 33 MW, an increase by 67 %. When the fire reach the maximum heat release rate it starts to decay and finally, for the situation simulated here, it will be ventilation controlled due to the large door openings. If we use equation (B10) we find that the heat release rate when the fire become ventilation controlled is about 20 MW. If we use the fire growth rate in equation (1) and a maximum heat release rate of 55.3 MW as input in BRANZFIRE and we find out that the oxygen concentration when the maximum heat release rate according to our simple calculation is slightly higher than 15 %. Thus we made some corrections in the input to BRANZFIRE and find out that the more correct time is 587 sec. This means that the maximum heat release rate is 79,3 MW i.e. 140 % higher than for the smaller volume. Assuming that the assumption about 15% oxygen levels is correct we should expect this to be reasonably truthful results. The difference in volume size of the 91 m (91.2 m x 22.3 m x 5m) long vehicle deck and the 180 m long (180 m x 30 m x 5 m) vehicle deck is 167 %, which is quite close to the difference in maximum heat release rate, 140 %.. 100. Heat Release Rate (MW). 91 m bildäck 180 m bildäck. 80. 60. 40. 20. 0 0. 5. 10. 15. 20. 25. 30. time (min). Figure 16. The heat release rates used in the simulation for the different vehicle decks.. The heat release rate curves used in the simulations of the two vehicle decks are presented in Figure 16. In Figure 17 we can see the smoke layer height for two different vehicle decks, 91 m (10 167 m3) and 180 m (27 000 m3). The time to reach to 0.5 m from floor differs by a factor of two. This may play an important role for the personnel on the vehicle deck..

(28) 27. 5 91 m 180 m. Smoke layer height (m). 4. 3. 2. 1. 0 0. 5. 10. 15. 20. 25. 30. 35. Time (min). Figure 17. The calculated smoke layer height for a 5 m high vehicle deck with two different lengths, 91 m and 180 m, respectively.. 91 m. 300. 180 m. o. Gas temperature ( C). 250. 200. 150. 100. 50. 0 0. 5. 10. 15. 20. 25. 30. Time (min). Figure 18. The calculated gas temperatures in the upper smoke layer for the 91 m and 180 m long vehicle decks.. The gas temperatures in the upper layer for the two different vehicle decks are shown in Figure 18. The maximum temperatures are similar in both cases but there is a time lag, which is related to the volume of the vehicle. The reason for the similar maximum temperatures is due to the increase in maximum heat release rates. In general we observe that the volume of the vehicle deck is an important parameter for the fire development and the consequences of the fire..

(29) 28. 5. Answers to the questions raised. The primary purpose of this project was to provide answers to the following questions: 1. Is it preferable to seal the vehicle deck at a fire and let the water spray system cool the fire or should manual extinguishments be undertaken? 2. What are the consequences if a door is opened to the deck during a fire? 3. Should the mechanical ventilation be switched off and in that case, when? 4. How high can the fire gas temperatures reach? 5. How is the fire affected if the deck is divided into sections? Based on the experimental and theoretical work carried out in this study we are able to present the following answers to the questions raised above: 1. It is preferable to seal the vehicle deck and let the water sprinkler system cool the fire gases. The tests show that the sprinkler systems were of great help, and significantly reduced the time taken to extinguish the fire. However, it must be remembered that the pressure and droplet sizes in the trials were not to scale, so the extinguishing performance in reality would probably be poorer. Nevertheless, it is felt that the cooling effect of a real sprinkler system would be considerable. From a tactical point of view it is important that the sprinkler system is activated as soon as possible. 2. The tests show that it requires large openings in order to influence the fire development. The opening of one or two doors should not jeopardize the control of the fire. Actually, one of the most important findings in this study is that in relation to the volume of the model, the openings in the form of stairwells, ventilation shafts and scuppers were far too small to supply sufficient oxygen in order to maintain a fully developed fire at the vehicle deck. When relatively larger openings than those mentioned above were provided, the fire became ventilation controlled i.e. the fire size was dictated by the air flow through the openings. In only one case we obtained nonrestricted fire development i.e. when one of the short ends of the model were fully open. 3. The mechanical ventilation should be switched off. The experiments do not show any advantages of using the mechanical ventilation. During one of the tests with mechanical ventilation, the fire seemed to pulsate and oscillate around a particular equilibrium. This situation may become hazardous if the ventilation is increased further. As the smoke layer will descend relatively fast down to floor level there are no obvious benefits of using the mechanical ventilation. It will only increase the risk for things to go wrong. 4. The temperatures within the hot smoke layer further away from the initial fire source will be in the range of 150 – 300 oC prior to the activation of the sprinkler system. According to the tests the temperatures at the ceiling directly above the initial fire source can reach temperatures, which are in the range of 800 – 1100 oC. The highest temperatures obtained will be highly dependent on the type of vehicle burning at the.

(30) 29. early stage of the fire. 5. There are positive feedbacks of compartmentisation. The maximum heat release rate will be lower and the fire will self-extinguish faster. The drawback is that the hot smoke will descend faster down to the floor level than in the other case. This might influence the escape situation..

(31) 30. 6. Conclusions. In total, 18 tests were carried out, varying different parameters. Additionally, theoretical work using computer models was done. The following is a summary of the conclusions from the project: •. The trials showed that the degree of ventilation and the supply of oxygen are decisive in determining the growth and progress of a fire on a vehicle deck. Nevertheless, despite the fact that the ventilation can control fires on a closed vehicle deck, fires can be very large (30 – 80 MW) and present a real danger to crew and passengers. The volume of the vehicle deck appears to be an important factor for the fire development and its consequence.. •. The trials also showed that the risk for a serious fire and even a total loss of the ship is greater during loading and unloading, when the loading doors are open, than during the voyage.. •. Flames spread very rapidly across the ceiling, and the entire vehicle deck was filled with smoke within two and a half minutes (7 minutes in large scale), with toxic gases creating a life-threatening situation. In reality, the flames would fill the shallow space between the deck above and the top of the trucks, and fire would spread between vehicles in the same manner as a forest crown fire.. •. The trials showed that stairways and ventilation shafts are too small to provide sufficient oxygen for a major fire. One of the trials created a pulsing fire. As the fire grew, the quantity of oxygen was reduced and the fire tended to die down. When the amount of oxygen available increased, the fire flared up again with high intensity.. •. The sprinkler systems significantly reduced the time taken to extinguish the fire. However, it must be remembered that the pressure and droplet sizes in the trials were not to scale, so the extinguishing performance in reality would probably be poorer. Nevertheless, it is felt that the cooling effect of a real sprinkler system would be considerable.. •. The only time a major fire took hold was when one short side of the scale model, representing a stern door, was open, permitting a good supply of oxygen. The risks for a major fire are therefore probably greater when the ship is in harbour, with the doors open, than when at sea with the doors shut.. •. Practical advice varies, depending on whether the ship is at sea, or is at the quayside with the loading doors open. The best response at sea is to quickly evacuate the vehicle deck (i.e. crew, as passengers normally are not allowed on the deck during the voyage), to enclose the fire and activate the sprinklers which, at that stage, would not affect smoke development or the layer of air, but which would cool the fire gases and thus lower the temperature.. •. It is important not to ventilate the deck or to attempt to enter it before the fire gases have been cooled. If not, it is likely that the fire, which would probably have died down due to lack of oxygen and the effect of the sprinklers, would start again. This time, there would be more rapid growth and higher intensity due to the heat stored in.

(32) 31. the construction of the ship from the first fire. •. If a fire occurs at the quayside, the vehicle deck must be evacuated and the sprinkler system started immediately in order to reduce the spread of the fire. A fire can grow to full size in only 1 - 10 minutes, depending on the ignition source and on the vehicles involved.. •. With the ship in harbour, the fire and rescue services can bring their full resources to bear to assist the ship's crew. However, there is greater risk of a major fire and loss of the vessel due to the ventilation situation. Fire and heat can spread throughout the vessel, as the temperature in the hull can become very high..

(33) 32. 7. References. 1. Ingason, H., An Overview of Vehicle Fires in Tunnels, Fourth International Conference Madrid, Spain, 2-6 April, 2001, pp. 425 - 434.. 2. SP Report 1994:54, Proceedings of the First International Conference on Tunnel Fires, Borås, 1994.. 3. SOLAS, The International Convention for Safety of Life at Sea, International Maritime Organisation, London, England, Third edition 2001. 4. HSC CODE, International Code of Safety for High-Speed Craft, adopted on 20 May 1994. 5. Resolution A.123(V), Recommendation on fixed fire extinguishing systems for special category spaces, International Maritime Organization, London, England, October 25, 1967. 6. Bilger, R. W., Computational Field Models in Fire Research and Engineering, Fire Safety Science – Proceedings of the fourth international symposium, pp. 95-110, 1994.. 7. Wade, C.A., BRANZFIRE Technical Reference Guide, Study Report No. 92 (2000), ISSN:0113-367.. 8. NFPA 72E, Standard on Automatic Fire Detectors, National Fire Protection Association, Quincy, 1984.

(34) 33. Appendix A – A detailed analysis of the test results A.1. Effect of ventilation on the heat release rate. Fire in an enclosure such as the one presented in this study occurs in three idealized modes [12]. At the initial stage of the fire, i.e. in the first mode, the flame is surrounded by un-vitiated air (no smoke layer). Fresh air is entrained into the fire plume and as the fire develops in the enclosure, a stratified layer of vitiated gas may be formed at the ceiling and its thickness will grow in time if this space is not well ventilated. In the second mode the flame intrudes into the vitiated gas layer and part of the flame will entrain pure air and part vitiated gas. The third mode is reached when the vitiated ceiling layer descends below the base of the fire and the entire flame is immersed in a vitiated atmosphere. The condition described above is usually defined as the pre-flashover stage [9, 10]. There are three final conditions that can be established in the pre-flashover situation: • • •. The fuel controlled fire The ventilation controlled fire The self-extinguished fire. A pre-flashover situation means that the fire will not develop to such extend that there will occur a flashover within the room. Flashover is characterized by a sudden and dramatic involvement of most of the exposed fuel surfaces and rapid increase in the growth rate of the fire. This means that when flashover occurs, the entire room is engulfed in flames and everything, which can burn, will be involved in the fire. This situation is called the post-flashover situation [9,10]. In most of the non-sprinklered experiments carried out in this project we can observe all these three phases but only during the pre-flashover period. The fire did not reach a flashover. In Figure 5 in the main text of this report we see a plot of the measured heat release rate for test performed under different ventilation conditions. Tests 01 and 02 are excluded here, as the heat release rate measurements were found to be unreliable due to technical error. In Figure 5 we can observe that for tests 03, 04, 05, 07 and 17 the fire was selfextinguished whereas for tests 06 and 08 the heat release rate was reduced to a level, which is about half the maximum value obtained. The maximum heat release rate obtained for tests 03 – 08 was in the range of 162-198 kW with an average of 181 kW. In large scale this would correspond to a fire of about 33 MW. In test 06 we had two door openings, one on each short side, which measured 0.25 m wide and 0.5 m high. The smoke layer descended down to the floor level but in the door opening we observed that about half the upper part was filled with exhausted smoke whereas fresh air was flowing in the lower part. In test 08 we used all the openings from test 03 and 04, i.e. the stairwell, the shaft and the scuppers and at time 06:17 min:sec we opened a door (0.25 x 0.5 m) located on the long side of the model, near the fire source. The fire was nearly self-extinguished when we opened the door and the fire started to grow again. The change in the heat release rate can been seen in Figure 5 (test 08). In test 02, with a mechanical ventilation of 10 air changes per hour and a door opening at the stairwell box that measured 0.05 m wide and 0.5 m high, the fire seemed to pulsate and.

(35) 34. oscillate around a particular equilibrium. As the intensity of the fire increased the amount of oxygen available decreased, and the intensity of the fire declined. At this lower intensity, it consumed less oxygen, so that after a while the fire increased again. It is evident that in these tests that large vertical door openings on the sides of the vehicle deck can create ventilation-controlled fires without transition to a post-flashover fire. The critical size of the door openings (self-extinguished contra ventilation controlled fire) has not been clearly established by these tests. In one case (test 06) we had two door openings far away from the smoke filled vehicle deck (one door opening at each short side wall) whereas in the other case (test 08) we had single door opening close to the fire (at the long wall at approximately 90 degree angle to the fire source). In that test the ventilation shaft was open as well. The average heat release rate during the time period of 10 – 23 minutes of test 06 was 83 kW (15 MW) and for test 08 it was 66 kW (12 MW). In tests 09 and 19 we obtained a maximum value of 400 kW (73 MW), which is roughly two times the maximum value obtained when the fire became ventilation controlled or self-extinguished. Test 19 is a free burning test under a large hood calorimeter and test 09 is a test with one portal open (short side wall; 2.786 m x 0.625 m). This corresponds to the case when loading or unloading the vehicle deck. It is apparent that the fire in test 09 would have continued to grow (fuel controlled fire) if there was more fire load on the vehicle deck. The fire would have become ventilation controlled, most probably as a post-flashover fire, but that would have required 3-4 additional wood cribs. Theoretical heat release rate in the post-flashover situation with this opening would be 1162 kW (210 MW) (see Appendix B). In tests 01, 03, 04, 05, 07 and 17 we obtained self-extinguished fires. In the following we will discuss the reason for this. The hot smoke layer contains a mixture of air (O2, N2 etc), gaseous combustion products (CO2, CO etc) and smoke particles. The chemical combustion process in the flame will be influenced as the oxygen concentration in the smoke layer decreases. The reduction of the oxygen concentration in the hot smoke layer (vitiated air) is due to depletion of oxygen in the combustion process. Depending on the degree of vitiation the heat release rate start to decrease and finally it will self-extinguish if the amount of new fresh air towards the base of the fire source is to low. In Figure 19 we have plotted the ratio of heat release rate to the maximum heat release rate as a function of the oxygen concentration in the upper layer measured at h/H=0.76 (0.475 m from floor). In order to generalize we use normalized height, h/H, where h is defined as the height up to the smoke layer and H is the total height (0.625 m). The fire development is apparently not influenced until the oxygen concentration is down to a level of 14 - 17 %. The heat release rate is then decreased and at about 13 – 15 % it will self-extinguish. This usually occurred when the oxygen concentration in the upper smoke layer was as low as 13 – 15 %. The reported flammability limits for many fuels in literature [11, 12, 13] is about 13 – 15 %. Thus, the fire will self-extinguish when the flammability limits are obtained in the vitiated air surrounding the base of the fire. At this stage the fire behave very differently; lifted blue flames travelling around in the vicinity of the fire source..

(36) 35. 1. test 03 test 04 test 05 test 07. 0.8. Q/Q. max. 0.6. 0.4. 0.2. 0 10. 12. 14. 16. 18. 20. 22. Oxygen concentration (%). Figure 19. The heat release rate ratio Q/Qmax as a function of the oxygen concentration measured at h/H=0.76 for those test where self-extinction were obtained.. The oxygen concentration at h/H=0.76 is not the same as the one at h/H=0.48 or h/H=0.24. By theoretical means we found that the difference is not much higher than one or two percent, see Figure 20. This means that at the h/H=0.24, i.e. at the same height as the bottom of the wood crib we estimate the oxygen concentrations to be in the order of 15 to 17 % when the fire self-extinguish. This is slightly higher than the reported flammability limits found in the literature.. 1. test 03 test 04 test 05 test 07. 0.8. h/H. 0.6. 0.4. 0.2. 0 10. 12. 14. 16. 18. 20. 22. oxygen concentration (%). Figure 20. The oxygen concentration as a function of the non-dimensional height, h/H. The values at h/H=0.48 and 0.24 are calculated from the temperature measurement at corresponding height..

(37) 36. A.2. Smoke layer height. In the tests performed here we were not able to register accurately the position of the smoke layer at every time. Thus we have used the thermocouple readings to determine the position of the layer and then checked the video in order to see if the results were reasonable. The smoke layer position was determined by select the time when the temperature increase was 5 ºC from ambient. This appeared to give reasonable results compared to the video estimates. In order to determine the time when the stratified hot smoke layer reach to the floor level we calculated the time by assuming the same descend speed h/H = 0.48 to h/H= 0.24 as from h/H=0.24 to h/H=0. This results probably in a longer time than in reality but it is sufficiently accurate for the purpose of this project. In Figure 21 we can observe two distinct groups of curves. The faster curve group is dominated by tests were the fan was used and the other is dominated by natural ventilation or no ventilation at all. The most probable explanation to why test 08 (no fans) is in the faster curve group is that the fire growth rate was faster compared to test 04, which is identical to test 08 until the time 2:05 min:sec when the stairwell was opened.. 1. test 01 test 02 test 03 test 04 test 05 test 06 test 07 test 08. 0.8. h/H. 0.6. 0.4. 0.2. 0 0. 0.5. 1. 1.5. 2. 2.5. 3. time (min). Figure 21. The non-dimensional smoke layer height as a function of time from ignition. The time to reach floor level is based on calculated time.. The time it takes for the smoke layer to descend down to the floor level for the tests in the 19.85 m3 test rig is about 1:50 min:sec for the faster group (fan) and about 2:10 min:sec for the slower group (no fan). The smoke layer descended to floor at 1:30 min:sec in the 8.5 m3 test room (test 17). If we use the scaling laws given in Appendix C we can show that the corresponding times in large scale would be 5:11 min:sec for the faster group (fan) and 6:08 min:sec for the slower group (no fan). This would mean that the time before the vehicle deck is smoke filled for a ferry with corresponding ceiling height and volume would be in the order of 5 to 6 minutes. The time to reach the critical smoke layer height, which is about h/H=0.4 (h=2 m in large scale), would be about three and half minute to four and a half minute. This means that the time to react before the vehicle deck is filled with smoke is not very long. Here we assume that the fire growth.

(38) 37. rate in large scale is similar to the one obtained with the wood cribs. This does not necessarily be true for a real vehicle fire. This would affect the results presented above. Faster fire growth rates would result in faster smoke layer descends and vice versa.. A.3. Gas concentrations. The vitiated gas layer is a mixture of air and combustion products. Depending on the fire growth rate, the ventilation conditions and the volume of the room we will obtain different gas concentrations in the vitiated gas layer. In the test series carried out we measured the oxygen concentration (O2), the carbon dioxide (CO2) and carbon monoxide (CO) at h/H=0.76 and 5.4 m from the fire source. In Figure 6 in the main text of this report we can see how the oxygen concentration varies depending on the test conditions. In test 01 no openings were present whereas in test 03, 04, 05, 07 and 09 different openings were mounted. Apparently, the lowest oxygen concentrations tend to be in the natural ventilation cases. In test 01 and test 04 we measure very similar oxygen concentrations whereas in test 03, which have poorer ventilation than in test 04, we measure the lowest oxygen values in the entire test series without sprinkler. This was not expected as one would rather expect test 01 and test 03 to have similar values. This shows that the results are not always consistent with what to expect. It is however apparent that the tests with a fan tend to yield higher oxygen concentrations compared to the naturally ventilated cases.. 22. Oxygen concentration (%). 20. 18. 16. 14. test 01 test 02 test 05. 12. test 07 10 0. 5. 10. 15. 20. 25. Time (min). Figure 22. The oxygen concentration measured at h/H=0.76 for ventilated fires with a fan. For comparison we show test 01 which had no openings.. At the same time as oxygen is depleted in the combustion process CO2 and CO is produced. The concentrations of CO2 and CO are of interest due to toxicity limits during the escape period from the vehicle deck. In Figure 23 we can se how the CO2 and CO concentrations varies with time for test 04. The maximum CO2 measured was 6 % and the maximum CO was about 0.4 %. An interesting observation is that the quotient CO/CO2 appears to be fairly constant during the entire test and the fire does not show any.

No results found