ICHEME SYMPOSIUM SERIES NO. 141 A MODEL FOR PREDICTING THE HAZARDS FROM LARGE SCALE COMPARTMENT JET FIRES

Transcription

1 A MODEL FOR PREDICTING THE HAZARDS FROM LARGE SCALE COMPARTMENT JET FIRES M.A.Persaud and G.A. Chaberlain, Shell International Oil Products, Shell Research and Technology Centre, Thornton, P.O.Box 1, Chester, CHI 3SH. C. Cuinier, Therodynaics Departent, Institut National des Sciences Appliquees de Rouen, B.P.08 Place Eile Blonde!, Mont Saint Aignan Cedex, France. Shell Internationale Research Maalschappij B.V SYNOPSIS Prediction of the hazards associated with jet fires is an iportant part of the Safety Cases subitted for offshore platfors. The effects of confineent on the behaviour of jet fires has, however, received little attention to date. Recent experiental studies have suggested that in addition to the well-known hazards of high heat fluxes, there exist several potentially new hazards including external flaing, increased soke, and increased CO levels [1]. These experients also suggest that the confined fire approaches steady state quite quickly, in the order of 5-10 inutes in ost scenarios. This paper presents a physically based zone odel to enable the hazards to be estiated at steady state and at a scale representative of offshore odules. For engineering purposes the odel has been ipleented in a spreadsheet for which is quick to run and easy to use. Keywords: Copartent Fire, Jet Fire, Fire Hazards, Offshore Safety INTRODUCTION When a fire burns inside a copartent, the cobustion starts as though the fire were in the open. There is enough air already present to satisfy the stoichioetric requireent for coplete cobustion and the fire is fuel controlled. The hot cobustion products rise to the ceiling and spread out as a ceiling jet. There is a net outflow of aterial through the copartent openings as the gases heat up and expand. After a short tie a hot gas layer of cobustion products builds up in the upper part of the copartent which grows and descends as gases continue to flow into it. A relatively well defined interface norally fors between the upper hot layer and cool air below. When this interface descends below an opening there is a sudden outflow of soke, cobustion products or flae. If the copartent openings are sall, the fire ay not be able to entrain enough air for coplete cobustion of the fuel inside the copartent. The fire is then said to be ventilation controlled. The ajor hazards associated with copartent fires include all those norally associated with open fires. However, additional hazards exist due to the effect of confineent. For personnel these include: 163

2 The extent of external flaing Ipaired visibility along escape routes through soke obscuration Increased hazard fro carbon onoxide (CO) Explosion hazard fro unburned fuel if the fire extinguishes due to insufficient oxygen Heat loading onto vessels, pipework and structures can be greater in copartent fires due to the effects of increased soot foration during ventilation controlled conditions, and additional heat radiated fro the hot surroundings/walls. The aount of theral radiation externally radiated fro the fire is very sensitive to the aount of soot present. Generally, the greater the degree of ventilation control, the greater is the soot concentration, and the greater is the radiation hazard. However in heavily under ventilated fires the reduced cobustion intensity can result in lower plue teperatures and the soot produced can also obscure radiative heat fro the flae. The priary physical paraeters affecting copartent fire behaviour include: 1. Geoetry and size of copartent, position and size of vents, and degree of theral insulation present on the boundaries. 2. Type of fire (jet or pool) and fuel involved. 3. Release conditions of fuel - ass flow rate, orifice/nozzle diaeter or pool diaeter, flow regie, orientation of release for jet fire. 4. The ass transfer fro the fire feeding the soke layer and losses to the outside environent through vents. The aount of air present and available for cobustion is the controlling factor in deterining soot and CO production and flaability of the soke layer. 5. Radiation and convection heat fluxes to ipinged roof, walls and objects. The radiation heat flux is closely linked to soot volue fraction and flae cheistry, particularly the acetylene (C 2 H 2 ) and CO reaction kinetics in the soke layer. The convection heat flux is usually uch lower than the radiation except where a jet fire ipinges directly. The aount of heat lost through the roof and walls depends on the effectiveness of insulation. This in turn affects internal teperatures and therefore the cobustion behaviour. 6. Mitigating circustances, such as passive fire protection coatings on objects inside the copartent and/or the roof and walls. 7. Water deluge, if present. The processes discussed above are not independent, but are coupled together such that a change in any one process can have an iediate effect on all others. In a rigorous analysis of copartent fire behaviour, all pertinent physical and cheical processes need to be considered. However, physically based zone odels incorporating epirical correlations for

3 the cheistry and soot behaviour can allow good prediction of the fire hazards and their evolution with tie. Recently, experiental work has enabled a uch siplified approach to be adopted for estiating the steady state behaviour of copartent fires. The forulation presented is suitable for ipleentation in a coputer spreadsheet and gives estiates of global soke layer properties at steady state, such as depth and teperature, radiative heat and ass transfer losses, and the extent of external flaing. THE STEADY STATE MODEL The copartent geoetry is approxiated by an idealised box structure of length L, height H, and width W, and can have either one or two open vents at one end (see Figures 1 and 2). The vent areas and positions are defined by a height h v, width w v, and distance between vent base and floor level, z v. The copartent itself ay be therally insulated or totally uninsulated. The ass flow rate of fuel released into the copartent, r, is an input paraeter allowing jet fires to be considered and, if the total ass burning rate is known, pool fires also can be considered. Heat transfer between the copartent surface and outside environent is assued to be purely radiative and the effects of convection are neglected. This is a reasonable approxiation for large scale odules in which the fire velocity near to boundaries can be assued sall (neglecting direct jet ipingeent), and where wind cooling effects on die outside surfaces are ignored in a worst case analysis. A steady state heat balance is then perfored by atching die heat released during ventilation controlled cobustion with the total heat loss, Qj oss, fro the syste. The latter is represented by radiative losses fro the copartent walls and through the vent areas in addition to heat loss due to ass flow of hot cobustion products to the outside. Two cases are considered: 1. A copartent wi one vent having a neutral plane which intersects the vent plane at steady state (Figure 1). 2. A copartent with two vents having a neutral plane which lies in between the vents at steady state (Figure 2). The governing equations for die steady state ass and heat transfer processes, for each of these cases, are as follows: Mass Transfer For both cases it is assued that the gases in the upper soke layer are well stirred, which is a good approxiation for large copartents. Hot gases/soke exit the copartent above the neutral plane and cold air enters below it. The depth to which the neutral plane extends fro die ceiling or 'height' of the neutral plan is denoted by h n and is related to the distance above floor level, z n by: K = H-z n (1) 165

4 -. h l I z n i V 1 Figure 1 Single vent copartent (Case 1) Width of copartent = w Width of vent = w. z v ii * r 1 Figure 2 - Double vent copartent (Case 2) Width of copartent = w Width of lower vent = w, Width of upper vent = w 2

5 The stoichioetry or equivalence ratio, (j), is an iportant paraeter defining the overall state of ventilation and is defined by: a '/ (2) where r is the stoichioetric ass ratio for coplete cobustion and a is the ass flow rate of air supplied to the fire through the vents. For <jkl the fire is fuel controlled or over ventilated, <j>=l corresponds to stoichioetric burning, and for (j)>l the fire is ventilation controlled. Assuing the flow of gases is driven by buoyancy forces and there is no ixing of the inflowing and outflowing gases, then the pressure difference between the inside and outside of the copartent controls the flow. By integrating Bernoulli's equation over the area of vent overlapping with the soke layer above the neutral plane, the ass flow rate of gas out of the copartent is calculated by: K = jot^ftfe-p^fir^ +**-*&] (3a) for one vent (Case 1) and «0 - Q[2gp,(p fl -P,)]' /2 WJ^K, +z, 2 -z P -(z, 2 -zj*] (3b) for two vents (Case 2). The expression in Equation 3b assues that the neutral plane lies in between the upper and lower vents (Figure 2). The density of the soke layer/cobustion gases is assued to behave as an ideal gas and can be approxiated by: T P,=Pof W where the teperature of the soke layer, T s, is initially set to 1200 K in the spreadsheet and deterined ore accurately in the heat transfer calculation. Siilarly for the ass flow rate of air into the copartent: *> *%CjfoJPt-*$[rJk njp] (5a) for the single vent case and: *. = C,[2sp.(p -p,)] *»;,[(*,, -jjp -(z. -,, -Kf] (5b) for the double vent case. At steady state 0 = a +j-and Equations 1, 3, 4 and 5 can be solved to deterine the depth of the soke layer, h. Variations in T s have a sall effect on the calculated ass flows 167

6 and in practice little iteration is required to atch the ass flows into and out of the copartent while siultaneously satisfying the steady state heat transfer requireent (next section). Heat Transfer Steady state conditions require that the internal heat generated through cobustion, A re /, atches the total heat losses fro the copartent resulting in no net change in teperature. Neglecting convection, the net heat power generation within an uninsulated copartent is: Qnel ~ &EKI ~ [Qrad-twall + Qwd-tven, + Qrwl-^jlonr + Qassflow ) ~ 0 (6) and for an insulated copartent is: Q na = AE rd - (fl^, + &*-»«+ Q-f ) = ( 7 ) It is assued that the floor always acts as a heat sink. The heat release rate due to cobustion depends on whether the fire is fuel or ventilation controlled. For fuel controlled fires: AE nl = AH c r (for < > < 1) (8) and for ventilation controlled fires: A r W = ^ (for< >>l) (9) Considering each ter in brackets in Equations 6 and 7: &*«* = W*(j ) [LW + 2h n (L + fv)-w v (h v + z v -z n )] (10a) for the uninsulated single vent copartent and: GL«= w { ^ ) [LW + 2h (L + W v )-W v2 h v2 ] (10b) for the uninsulated two vent copartent. In Equations 10a and 10b it is assued that the copartent walls and ceiling are coposed of thin sheets of steel which radiate as grey bodies fro both sides, with an eissivity e w. Multiple reflections between the soke layer and walls are neglected which is a reasonable assuption given that the soke layer is, in reality, a volue eitter with negligible reflectivity. For large copartents it also reasonable to assue the soke layer radiates as a perfect black body. The heat capacity of the walls is assued to be sall copared with the theral conductivity such that there are no teperature gradients within the walls or ceiling at steady state. Background radiation contribution fro all sources at abient teperature are sall and therefore neglected in this analysis. With these assuptions, the equilibriu teperature of the uninsulated walls and ceiling is approxiated siply as T wa n=tj\2. 168

7 Neglecting view factors with the botto of the hot soke layer, the radiative heat loss directly through the vent is deterined by the spatial overlap between the soke layer and open vent area. For the single vent: a,,- = */U?*.{k, +*v -*J (l la) and for the double vent: 0^,, =e, 7:XA, (Hb) Heat radiated fro the botto of the soke layer can be iportant and if it is assued that the floor acts as a black body (e.g. coated in soot) and eits radiation at teperature Tfl oor, then for all cases: Qr^^r-^Mt-KjLW (12a) In reality the floor in an uninsulated copartent often coprises of high heat capacity aterial or steel in contact with an effective heat sink. In offshore odules the floor ay also be a steel grating and thus ainly transparent to theral radiation. Therefore it is reasonable to neglect Tjj oor fro Equation 12 in these cases, giving: Equation 12b effectively assues that the floor is at a teperature of which is a good approxiation for uninsulated copartents where back radiation fro the floor is sall. For a perfectly insulated copartent, the net radiative loss to the floor at steady state is siply: CL^Jtar-0 (12C) However, a perfectly insulated floor rarely occurs in practice and it can be assued that the floor always reains at a sufficiently low teperature such that Equation 12a or Equation 12b applies. The final ter in Equations 6 and 7 accounts for the ost significant heat loss process in practice, arising fro theflow of hot cobustion products to the outside through the vent. For the single and double vent cases the appropriate expression is: flu»»=(«/+*-m5~rj (i3) where it is assued that the fuel is released at abient teperature, has an isobaric specific heat capacity equivalent to that for air at abient teperature, and the fuel is either vapour or vapourises iediately upon entering the copartent (without a change in 169

8 enthalpy). These assuptions are reasonable given that the heat release through cobustion doinates over the heat required for liquid fuel vapourisation. The heat capacity of the cobustion products/soke layer is also assued to be equivalent to that of air. External Flae Unless the flae is directly oriented toward or is deflected through a vent, fuel controlled copartent fires burn within the confines of the copartent. However for ventilation controlled fires (<(>>1) the hot soot and partial cobustion products can burn on ixing with the air outside the vent. The vertical length of the external flae can be estiated fro the correlation of Yokoi, recoended by Thoas and Law [2]. Also, noting that external flaes easured in large scale experients by Chaberlain [1] rise at an angle of about 60 to the horizontal in still air, the total extent of external flaing siplifies to: /,, =0.01 W, f {\-y$ C P P K (14) Equations 1-14 have been coded into a spreadsheet progra allowing values for T s, T wa //, <t>, and Jgfto be deterined fro the copartent geoetry and ass flow release rate of the fuel. In practice an initial value of 7^ is estiated and entered into the spreadsheet to deterine a value of z which obeys 0 = a +j: A value for is then calculated allowing AR rel to be evaluated. A value for T s, such that Q t = 0, is then obtained fro the spreadsheet. This new value for T s is then used to optiise the values of z n and 0 (& a ). A flow chart showing the procedure used is shown in Figure 3. In practice a single iteration is required to satisfy all requireents for steady state conditions. Figure 3. Flow chart showing calculation procedure used in the coputer spreadsheet Calculate z for 0 = a + f & Calculate ass flows 1 Calculate phi and Heat release I I Evaluate 7 S such that Q e,=0 I 170

9 RESULTS The steady state odel was tested against prograes of experients carried out at two scales. One prograe was carried out by Shell Research Liited at large scale, in a steel copartent sited at the Norwegian Fire Laboratory, operated by SINTEF at Trondhei. al details can be found in reference [1]. The other prograe was recently carried out by Shell Research Liited at sall scale, in a 33 3 steel copartent sited at the Sheffield University Test Site in Buxton. Further details of these experients will be published in due course. All the large scale (Chaberlain) tests were conducted in an insulated copartent with one wall vent and used gaseous propane fuel. The sall scale (Persaud) tests were carried in an uninsulated copartent with either one or two vents in one wall, again using gaseous propane fuel. The height and orientation of the fuel release were different for the tests, but these differences are not treated in the odel. A suary of the experiental paraeters used in the odel is given in Table 1. Table 1. A suary of Test Paraeters used in the Steady State Model TEST ID CHAMBERLAIN lb 1 vent CHAMBERLAIN 27 1 vent CHAMBERLAIN 2 1 vent PERSAUD 27 1 vent PERSAUD 28 1 vent PERSAUD 17 2 vents PERSAUD 18 2 vents Copartent H(),W(), () 3.54,3.91, , ,3.91, ,2.355, ,2.355, ,2.355, ,2.355, Insulated or Uninsulated Insulated Insulated Insulated Uninsulated Uninsulated Uninsulated Uninsulated Vent /» v (), W v (), z v () 2.0, 2.5, , 2.5, , 2.5, , , , 2.355, ,2.355, , 2.355, ,2.355, f (kgs-i) Fuel Release / Nozzle Orientation sonic propane vertical subsonic propane vertical subsonic propane vertical subsonic propane vertical subsonic propane horizontal subsonic propane vertical subsonic propane vertical Model predictions based on the copartent geoetry, vent size and fuel ass flow rate are shown in Table 2 with the experiental values taken at, or near to, steady state. 171

10 Table 2. al and predicted values for ean soke layer teperature, wall teperature, roof teperature, stoichioetry and external flae lengths. Approxiate floor teperatures derived fro test data are also shown. 1 Steady state odel with floor teperature effectively set to using Equation 12b (no back radiation). ^ Steady state odel with floor teperature ade equal to experiental value at end of test using Equation 12a. 3 Steady state odel with floor teperature set to final equilibriu value assuing perfect insulation using Equation 12c. + denotes teperature still rising at terination of the test * estiated value of "steady" flae length (peak interittent values approxiately x2 greater) # values for axiu observed flae length TEST ID or Theory ( C) T wall ( C) 'roof ( Q Tfloor ( Q k lef () CHAMBERLAIN lb insulated, 1 vent CHAMBERLAIN 27 insulated, 1 vent CHAMBERLAIN 2 insulated, 1 vent PERSAUD 27 uninsulated, 1 vent PERSAUD 28 uninsulated, 1 vent PERSAUD 17 uninsulated, 2 vents PERSAUD 18 uninsulated, 2 vents Theory 2 Theory 3 Theory 2 Theory < # no data # * * 1.1 <0.5* 0 <0.5* 0 The results show that the steady state odel predictions are in good agreeent with easureent, given the liitations of the spreadsheet odel. Using Equation 12b to describe the heat transfer to the floor, then the soke teperatures and wall surface teperatures are generally in good agreeent with the observed values, although the agreeent is better for uninsulated than for insulated copartents. However, if the actual floor teperatures are used in the odel (Equation 12a) for the insulated copartents, then the agreeent is 172

11 excellent. This observation gives confidence to the overall ethodology eployed in the analysis. The zero back radiation assuption does not significantly affect the accuracy of predictions for the uninsulated copartents because the floor reains relatively cool and the T 4 eission is negligible. In practice offshore odules have inial theral insulation. Given that the therophysical properties of the floor and heat sinks ay not be easily quantified, the assuption of back radiation eission fro the floor at either zero or abient teperature reains a reasonable one for all cases, even for insulated copartents. Values for overall flae stoichioetry are also well predicted for ost tests and provides a useful basis for further calculations in ore coplex odels in which the cobustion cheistry can be treated to give CO or soot concentration. Values for external flae lengths show fairly good agreeent with experient. It should be noted that in practice the flaes were often attached to the vertical surfaces of the copartent (e.g. PERSAUD Test 27 and Test 28) resulting in longer visible flae lengths than expected, but this interittent effect is not included in the values given in Table 2. Also, in the large scale tests [1], values quoted for l e j represent the axiu observed extent of external flaing. Overall the odel predictions are rearkably good given the underlying siplicity of the theory. Future work will concentrate on developing a fully tie dependent zone odel for use in the field, incorporating correlations based on the global cobustion cheistry and soot evolution. Such a odel will be capable of predicting CO and soke source ters in addition to the hazards considered here. SUMMARY A physically based zone odel has been developed enabling the hazards of copartent fires to be estiated at steady state and at a scale representative of offshore odules. For engineering purposes the odel has been ipleented in a spreadsheet for which is quick to run, transparent and easy to use. For a given scenario the odel successfully estiates global soke layer properties at steady state, such as depth and teperature, radiative heat and ass transfer losses, and the extent of external flaing. The results copare favourably with experients to a level of precision sufficient for coparison and ranking of offshore hazard scenarios. REFERENCES 1. P. Thoas, M. Law, 1974, "The projection of flaes fro buildings on fire", Fire Prevention Science and Technology, 1Q, G.A. Chaberlain, 1994, "An experiental study of large-scale copartent fires", Trans I.Che.E, 22, Part B,

12 NOMENCLATURE Sybol Units Description Q C P Mrel A// c g H h I >ef Qnet Qrad wall Qrad floor Qrad vent Qass flow r T W 2 S * P CT Subscript a ef / n s v, vl, v2 wall diensionless Jkg-'K-' Is-' Jkg-' s' 2 kgs-i is-' Js-l is-' Js-l Js-l diensionless K diensionless diensionless kgnr 3 Js-'-2K- 4 Discharge coefficient Isobaric specific heat capacity of air Net energy release rate due to cobustion inside the copartent Heat of cobustion of fuel Acceleration due to gravity Height of copartent Height of vent or soke layer Length of copartent Length of external flae Mass flow rate Net. power generation inside copartent Radiative power transfer to wall or ceiling of copartent Radiative power transfer fro (he botto of the soke layer towards the floor Radiative power transfer through the vent Net. power transfer carried by ass flow through vent Stoichioetric air/fuel ass ratio for coplete cobustion Teperature Width of copartent or vent Distance above floor level Eissivity Stoichioetry or equivalence ratio Density Stefan-Boltzann constant (5.67 x 10" 8 ) Description Abient air External flae Fuel Neutral plane Soke/gas layer Single vent, lower vent, upper vent Wall or ceiling 174