- Email: [email protected]
Effect of ventilation procedures on the behaviour of a fire compartment scenario
Nuclear Engineering and Design 235 (2005) 2155–2169
Effect of ventilation procedures on the behaviour of a fire compartment scenario H. Pr´etrel ∗ , J.M. Such Institut de Radioprotection et de Sˆuret´e Nucl´eaire (IRSN), Service d’Etude et de Recherches Exp´erimentales sur les Accidents (SEREA), Laboratoire d’Exp´erimentation des Feux (LEF), Centre de Cadarache, 13108 C´edex Saint Paul-lez-Durance, France Received 29 July 2004; received in revised form 23 February 2005; accepted 3 March 2005
Abstract This contribution presents a study on the consequences of applying ventilation procedures during a fire scenario involving a TPH/TBP pool fire in a ventilated enclosure. This research is addressed to fire safety in the nuclear industry in which ventilated enclosures remain a configuration frequently encountered. This work presents experiments comprising a 300 kW liquid pool fire in a 400 m3 vessel connected to an industrial ventilation system featuring one inlet and one exhaust branch. The investigated ventilation procedures consist in closing the inlet branch only or closing both inlet and exhaust branches. The analysis compares fire behaviour with and without the implementation of a ventilation procedure and points out the effects of said procedures on the combustion rate, fire duration and gas temperature within the vessel. It highlights pressure variations within the vessel when both the inlet and exhaust ventilation branches are closed. Conclusions provide practical answers that would be useful when designing appropriate ventilation strategies limiting fire hazards. © 2005 Elsevier B.V. All rights reserved.
1. Introduction The scenario comprising a fire within a ventilated enclosure remains one of the key issues for fire safety assessment in the nuclear industry. Indeed, partitioning is frequently encountered irrespective of the building’s designated use, building reactor, fuel processing or reprocessing plants. Moreover, to ensure confinement, these enclosures are equipped with a ventilation ∗ Corresponding author. Tel.: +33 4 42 25 26 67; fax: +33 4 42 25 48 74. E-mail address: [email protected] (H. Pr´etrel).
0029-5493/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.nucengdes.2005.03.003
air cleaning system that allows negative pressure to be created in the compartments and thus prevent radioactive materials or harmful products from escaping out of the compartment (Jiang et al., 2002; Nuclear Air Cleaning Handbook, 2003). It is for this reason that configurations involving a fire in a mechanically ventilated enclosure are extensively studied in the framework of nuclear fire safety research (Audouin and Tourniaire, 1999; Pr´etrel et al., 2001). In such fire scenarios, the oxygen feeding into the compartment is mainly controlled by the ventilation system; consequently, one procedure that could limit the fire hazards would be to switch the ventilation
2156
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Nomenclature cv Cp C m m ˙ n P ˙ Q r t T V v˙ X Y
heat capacity at constant volume heat capacity at constant pressure coefficient mass rate of mass flow exponent pressure rate of heat release oxygen to fuel stoichiometric ratio time temperature volume rate of volumetric flow molar fraction mass fraction
Indices c ex ext f g in leak o O2 th v w
combustion exhaust extinction fuel gas inlet leakages reference oxygen theoretical ventilation losses
Greek symbols γ isentropic coefficient ρ density system to a special control mode. There are many possible approaches; these depend on the particularities of the enclosure and the expected efficiency (Klote, 1993; Vaari and Hietaniemi, 2000). A typical example is the rapid extraction of smoke (using positive pressure ventilation (PPV) concept) by mechanical fans. This approach limits the increase in gas toxicity and gas temperature (Svensson, 2002). In the case of one ventilated compartment equipped with inlet and exhaust branches, a possible ventilation procedure would be to close the valves situated either upstream or down-
Fig. 1. Schematic representation of the investigated fire scenario and the PLUTON facility.
stream of the fire compartment. In the nuclear industry, this procedure is of particular interest because it allows the compartment to be isolated and thus protects the rest of the installation against the consequences of fire, such as the plugging of filters with smoke or the release of radioactive dust. In order to determine the efficiency of such procedures and to identify the most appropriate one, the behaviour of the compartment and of the pool fire has to be investigated. To address this issue, an experimental study was undertaken within the framework of the FLIP research programme developed by IRSN in collaboration with COGEMA. Fig. 1 illustrates a fire scenario representing actual situations in reprocessing plants where some enclosures are dedicated to chemical processes using flammable liquids. The fire scenario is a pool fire located against one side of a rectangular enclosure. The ventilation network comprises an inlet branch and an exhaust branch each equipped with a closing valve. A fan located on the exhaust branch is used to create negative pressure in the compartment thus ensuring confinement. The location of the pool fire next to the wall, its shape and the ventilation characteristics, are selected in order to be representative of a typical scenario found in the nuclear industry. The research programme, performed at IRSN’s fire test laboratory at Cadarache (France), involved large-scale experiments designed to mimic the real situation as closely as
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
possible. The aims of these tests are to analyse the fire scenario under two ventilation procedures and to point out the consequences on parameters key to fire safety: the rate of combustion, the fire duration, the gas temperature and the pressure within the enclosure.
2. Description of the tests 2.1. Test set-up Tests are carried out in the IRSN PLUTON facility (cf. Fig. 1), which is a 400 m3 reinforced concrete vessel with 0.25 m thick walls. The vessel dimensions are 7.5 m in height and 54 m2 in area. The vessel is connected to an industrial ventilation network made of an inlet and an exhaust branch. The inlet branch is connected to the atmosphere. The flow inlet is located in the lower part of the enclosure (1.5 m from the ground) and is directed towards the pool. The inlet cross-section is 0.4 m2 . The decision regarding the position of the inlet, right in front of the pool, is made to ensure the symmetry of the inlet airflow entering the vessel. A butterfly type valve (from AMRI) is installed 1 m upstream from the vessel inlet. Equipped with an air control actuator, the valve allows the opening to be varied from a control room. The exhaust branch is connected in the upper part of the enclosure, and is equipped with a valve positioned flush with the vessel ceiling that acts as the valve seat. The valve operates vertically via an ac motor enabling the opening to be varied. Downstream, a dilution branch is connected to the exhaust branch thus allowing a wider range of airflow operating conditions. The exhaust branch ends with a 55 kW fan upstream where eight high efficiency particles air (HEPA) filters are installed to collect the soot particles produced by the fire. The pool consists of a rectangular container filled with 50 mm of a liquid fuel mixture named TPH/TBP, proportions 70% of hydrogenated tetra-propylene (C12 H26 ) and 30% of tri-butyl phosphate (C12 H27 O4 P) by volume. It is positioned against a wall at an equal distance from the two opposite sides as shown in Fig. 1. The fire is ignited with a propane burner directed toward the fuel surface.
2157
2.2. Instrumentation The mass fuel loss rate is determined based on pool mass loss measurements taken by the weighing system, SARTORIUS IS-300, located under the pool. The fuel mass loss rate noted mc is calculated based on the pool weight measured at each time interval (approximately 0.5 s). The fire heat release rate, HRR, is considered to be the product of the fuel mass loss rate and the total heat of combustion taken to be 36 MJ/kg for the mixture TPH/TBP. Gas temperature measurements are performed by means of thermocouples trees set in various places within the enclosure in order to characterize the fire plume, the ceiling jet and the vertical gas temperature stratification. Gas temperatures are also measured in the ventilation branches. K type thermocouples, diameter 1.5 mm, are used. The pressure within the compartment is measured with differential pressure transducers mounted flush with the wall enclosure. The airflow rates within the ventilation branches are measured with differential pressure probes (ANNUBAR type) connected to a differential pressure transducer. The air temperature is also measured at the probe location to take the density change into consideration in the airflow calculation. Oxygen concentration is measured with SERVOMEX4100 gas analysers connected to gas sampling systems. 2.3. Ventilation control sequence Two ventilation control sequences, based on actual operating procedures applied in the nuclear industry, are studied. The first sequence, referred to in this paper as inlet valve closure (IVC), involves closing the inlet valve 100 s after pool fire ignition without modifying any of the other ventilation system devices; the fan on the exhaust continues to operate. The 100 s time period corresponds to an estimate of the time that would lapse between fire ignition and fire detection (by detectors that automatically trigger inlet valve closure). The time taken for the inlet valve to close is approximately 5 s in the PLUTON facility, this is realistic as regards valves used in nuclear industry. The second sequence, referred to as inlet and exhaust valves closure (IEVC), involves closing the inlet valve 100 s after pool fire ignition, and then closing the
2158
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Table 1 FLIP test features Test number
FLIP1
FLIP10.1
FLIP10.2
FLIP10.3
Procedure m ˙ c (kg/(sm2 )) HRR (kW) Fire duration (s)
– 0.0251 360 1560
IVC 0.0168 242 1203
IEVC 0.0164 236 1272
IEVC 0.0177 256 1200
exhaust valve after 210 s. The 110 s, allowed to lapse between inlet closure and exhaust closure, is an averaged estimate based on practical considerations such as the time required for an operator in a control room to close the exhaust valve. In the PLUTON facility, the valve is entirely closed after 247 s due to the time the valve needs to move from the open to the closed position (i.e. 37 s). The enclosure then remains nearly sealed until fire extinction (the only gas leakage possible was attributable to the enclosure air-tightness). The fan is still operating and the entire airflow passes through the dilution branch (cf. Fig. 1). 2.4. Typical test procedure Four large-scale tests are considered in this study, their main characteristics are presented in Table 1. For the four tests, 0.02 m3 of fuel (approximately 16 kg) is poured into a 0.4 m2 square shaped container. The ventilation system is adjusted to achieve the required vessel pressure (about −800 Pa) and ventilation rate (about 3 h−1 ). Once the ventilation rate is stabilized, the pool fire is ignited by a propane gas burner that is shut off as soon as the propagation phase on the pool surface is initiated. The flame spreads over the surface until the entire pool area is burning and a fully developed fire is observed. During this period, the vessel is gradually filled with soot particles and visual access to the fire is prevented as soon as the soot particles reach the ground. The fire goes out without any external intervention; the pool fire is extinguished either due to a lack of fuel within the pool or because of oxygen depletion within the compartment. The analysis proposed in the next sections describes the effect of the two ventilation procedures on the fire scenario by examining the airflow rate, the mass fuel rate, gas pressure and gas temperature. The first test, FLIP1, is the reference test. No ventilation procedure is applied. The characteristics of the second test,
FLIP10.1, are identical to those of FLIP1 except that the IVC sequence is applied. The third and fourth tests, FLIP10.2 and FLIP10.3, are conducted following the IEVC sequence.
3. Analysis of the reference test 3.1. Theoretical considerations Before describing the behaviour of the reference test, the physical mechanisms contributing to modification of the vessel pressure, the gas temperature and the oxygen concentration are recalled through classical energy and mass balance equations commonly defined in compartment fires (Quintiere, 1989; Karlsson and Quintiere, 2001). The system examined, sketched in Fig. 2, is the volume of gas located in the vessel. According to the total mass conservation: d m= dt
m ˙f Fuel mass loss rate from combustion
+
m ˙ v,in − m ˙ v,ex gain and loss from the ventilation branches
+
m ˙ leak
(1)
gain or loss from leakages
Fig. 2. Schematic representation of the physical mechanisms involved during the fire scenario.
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2159
The time variation of the mass of gas in the vessel is the result of competition between the mass fuel rate contributed to the system (gas resulting from the evaporation of fuel that then undergoes the combustion reaction) m ˙ f , the inlet and exhaust airflows m ˙ v,in − m ˙ v,ex mass flow rate from gas leakages mleak . This last term, driven by the pressure within the vessel, can be positive or negative (the airflow enters or leaves the system) depending on the vessel gauge pressure. According to total energy conservation: d V d (mcv T ) = P dt γ − 1 dt internal energy variation
=
˙c Q
+
HRR from the fire
−
˙ v,in − Q ˙ v,ex Q gain or loss from the ventilation branches
˙w Q thermal losses through the walls
+
˙ leak Q gain/loss from leakages
(2) The time variation of the gas internal energy is the re˙ c, sult of a balance between the fire heat release rate, Q the energy gained or lost by the ventilation branches, ˙ v,in − Q ˙ v,ex , the thermal losses through the walls Q ˙w Q and the energy gained or lost (depending on the vessel ˙ leak . Considering the pressure) through gas leakages Q perfect gas law, the internal energy (mcv T) can also be expressed in terms of pressure as PV/(γ − 1) (γ being the isentropic gas coefficient), thus, the energy balance leads to the pressure equation (Rehm and Forney, 1994) as formulated with Eq. (2). These conservation equations clearly illustrate that a ventilation procedure, by setting the energy and mass fluxes passing through one or both ventilation branches to zero, modifies the mass, temperature and pressure within the vessel. Analysis of the test data will quantify the amplitude of these variations. 3.2. Description of the reference test The combustion of the TPH/TBP shows a typical behaviour illustrated on Fig. 3. The fuel mass loss rate increases over the first 160 s until it reaches a maximum value and then follows a slow decay until 1380 s, a few instants before extinction. At this time, just before extinction, a sudden
Fig. 3. Fuel mass loss rate and oxygen yield during the reference test FLIP1.
increase in the mass loss rate is observed. This phenomenon (also reproduced at a small scale with Cone Calorimeter tests) is attributed to the behaviour of the TBP fuel alone that experiences a very high evaporation rate at extinction when it occurs by lack of fuel. Concerning oxygen concentration, as combustion continues, the oxygen concentration within the enclosure decreases gradually until it stabilizes at a value of 18.5% after 1000 s of fire (cf. Fig. 3). Upon extinction because of the sudden increase of fuel mass loss rate, the oxygen concentration also dropped abruptly and then increased after extinction due to the fresh air introduced by ventilation. Pressure evolution within the enclosure follows also a typical behaviour observed in a confined and ventilated enclosure (cf. Fig. 4). Before ignition, the pressure within the vessel is constant and equal to the negative relative pressure level induced by the fan (−800 Pa for these tests). Starting as of ignition and during flame propagation on the pool, the room pressure rises gradually to a peak, and then falls back to the value set before ignition. After this peak and during the fully developed stage of the fire, the pressure remains almost constant. Upon extinction, an abrupt fall in pressure is observed. The over-pressure peak upon ignition and the lowpressure peak upon extinction are explained through the pressure equation (Eq. (2)). The over-pressure peak is the result of the gas expansion mechanism caused by
2160
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Fig. 5. Airflow rates in the ventilation branches during the test FLIP1. Fig. 4. Pressure within the vessel during the reference test FLBP1.
the rapid release of heat within the vessel gradually compensated, firstly by the net energy lost through the ventilation branches (net balance between inlet and exhaust), and secondly, by heat losses through the enclosure walls. The low-pressure peak is the results of the gas contraction mechanism caused by the sudden interruption of the fire heat release gradually compensated firstly, by the net energy gained through the ventilation branches, and secondly, by the energy contributed by gas leakages (entrance of mass because the vessel pressure is negative at this time). The amplitude of the over-pressure peak upon ignition depends on the heat release of the fire, the thermal losses though the walls, the features of the ventilation branches and compartment air-tightness. The airflows in the inlet and exhaust branches follow the same trend as the vessel pressure (cf. Fig. 5). Before ignition, the airflow rate is roughly equal to 0.33 m3 /s (or 1200 m3 /h), which corresponds to the initial conditions imposed. The discrepancy between the flow rates in the two branches (that theoretically should be equal because of the mass conservation) is the result of measurement uncertainties in the exhaust branch (only observed in the test FLIP1). Once the pool fire is ignited, a sudden increase in the flow rate is observed in the exhaust branch, this was the consequence of an overpressure peak in the vessel. For these FLIP tests, the pressure was not observed to have any effect on the
inlet flow rate, this was because the airflow resistance of the inlet branch was greater than that of the exhaust branch. The reference test is also characterised by gas temperature variations within the vessel that are typical of compartment fire (Pr´etrel et al., 2001). The gas temperature variation over time is illustrated in Fig. 6 for four locations (within the fire plume, the ceiling jet, the hot layer and the cold layer). The evolution of the gas temperature within the enclosure follows the flow motion. From the fire plume to the ceiling jet and then within the hot layer and the cold layer, the gas temperature decreases gradually. The gas reaches a temperature of
Fig. 6. Gas temperature during the test FLIP1.
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2161
350 ◦ C within the plume and 170 ◦ C in the ceiling jet. The gas temperature variation over time follows the same trend as the mass fuel rate, and thus the rate of heat release, and is typical of compartment fire. After ignition, the temperature rises rapidly because of the sudden release of heat, and then stabilizes gradually due to the continuous increase of the thermal losses through the walls. Based on this description of the reference test, the next sections will analyse how ventilation procedures can modify this behaviour.
4. Description of the ventilation procedures through the airflow rate The analysis of the airflow rates through the ventilation branches illustrates the two ventilation procedures chosen for this study. Figs. 7 and 8 show the time history of the airflow rates in the inlet and exhaust branches for the sequence with inlet closure only (IVC, test FLIP10.1) and both inlet and exhaust closures (IEVC, test FLIP10.3), respectively. The reference time (t = 0) corresponds to the pool fire ignition. 4.1. Sequence IVC The closure of the inlet valve is clearly identified at t = 100 s leading to zero flow rate in the inlet branch. As a result, the exhaust flow rate experiences a dramatic
Fig. 8. Airflow rates in the ventilation branches during the test FLIP10.3.
initial drop followed by a gradual decay towards zero at t = 420 s. This behaviour is analysed with the mass conservation (Eq. (1)). Once the inlet valve is closed, the exhaust airflow is generated by the balance between gas mass variation over time (due to gas expansion) minus the fuel mass loss rate and the gas leakage mass rate. Between 100 and 420 s, gas temperature continues to rise thus inducing airflow at the exhaust (due to thermal gas expansion). Gas temperature gradually achieves a steady state value and the exhaust flow rate is then attributable to the fuel mass loss rate and leaks only. For these tests, this airflow rate remain within the sensitivity range of the exhaust branch measurement devices (about 0.01 kg/s) as observed after t = 420 s. The practical conclusion that may be drawn from this result is that even when the exhaust branch is open, the amount of mass that leaves the compartment is very small. 4.2. Sequence IEVC
Fig. 7. Airflow rates in the ventilation branches during the test FLIP10.1.
The variation of airflow over time during the IEVC sequence is similar to that observed for the IVC sequence. The closure of the inlet valve results in a drastic drop of the exhaust flow rate, which is followed by a slower decay. The exhaust valve closure process starts at 210 s and leads to complete shut down of the exhaust flow at 247 s (37 s needed for the valve to close). At this time, the gas mass variation within the vessel becomes
2162
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
the net balance between the fuel mass loss rate brought to the system and the mass rate of gas leakages lost by the system. This net balance can be positive or negative depending on the fire HRR and on leakage characteristics. As an example, for the test FLIP10.3, the total net balance is determined using the following relation (deduced from Eq. (1)): text d m(t = 247) − n(text ) = m dt t=247 dt text = (m ˙c−m ˙ leak ) dt (3)
5. Effect of the ventilation procedures on fire characteristics
directed towards the pool fire. Both effects reduce the fuel mass rate significantly. On the contrary, closure of the exhaust valve has no noticeable effect on the fuel mass loss rate. Once the inlet valve is closed, the airflow rate in the exhaust branch is so low that the configuration of the exhaust valve (open or closed) has no immediate effect on the combustion process. The analysis is confirmed looking at the average fuel mass loss rate (indicated in Table 1). Without the ventilation procedure, the total average fuel mass loss rate is about 0.0251 kg/(s m2 ). With the ventilation procedure, it drops down to 0.0168 kg/(s m2 ) for FLIP10.1 and 0.0177 kg/(s m2 ) for FLIP10.3. These conclusions are mostly qualitative since the reduction of the burning rate depends also on the position of the inlet with respect to the pool fire. The configuration adopted for FLIP involved blowing the inlet airflow directly on the pool fire, in others configurations, the effect of the ventilation procedures on the fuel burning rate can be amplified or weakened.
5.1. Fuel mass loss rate
5.2. Extinction mode
Closing the inlet valve during the fire scenario lead to a sudden decrease (approximately 25%) of the fuel mass loss rate (cf. Fig. 9 presenting the variation of mass loss rate over time for all three tests). This is the consequence of two mechanisms induced by inlet valve closure, first, cut-off of air feeding the lower part of the vessel, and second, shut off of dynamic fresh airflow
According to the FLIP tests, ventilation procedures can also alter the mode of extinction. In the reference test FLIP1, extinction occurs when all the fuel is burnt, but for the tests FLIP 10.1 and FLIP10.3 (with inlet and exhaust closures, respectively), extinction is induced by lack of oxygen. These two extinction modes are illustrated in Fig. 10 that presents oxygen yield over time within the vessel. In the absence of a ventilation procedure, at extinction the molar oxygen yield is around 18% (certainly above the limiting oxygen index, LOI, that characterizing the extinction limit) and extinction occurs through lack of fuel. With the ventilation procedures, extinction occurs by lack of oxygen indicating that the oxygen yield within the vessel definitely reached the LOI before the entire mass of fuel is burnt. The molar oxygen concentration measured at extinction is approximately 15 vol%. However, the change in the extinction mode is specific to the FLIP tests and cannot be generalized as a systematic effect of the use of ventilation procedures. Indeed, the use of the inlet closure procedure, modifying the oxygen feeding, tends to move the fire scenario from an over-ventilated situation to an under-ventilated one and then changes the extinction mode. However, the combustion regime (that depends on the initial ven-
t=247
The integral evaluated from the measurements gives a loss of mass equal to 5.3 kg that corresponds to about 1% of the initial mass of gas in the vessel.
Fig. 9. Variation of fuel mass loss rate with time.
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2163
considering a vessel equipped with an exhaust only (no inlet flow, m ˙ v,in = 0 and no leaks). Based on these assumptions, the theoretical fire duration deduced from the oxygen mass conservation may be written: d ˙ f − Y O2 m YO2 m = −r m ˙ v,ex dt
(4)
From the previous equation and using the total mass conservation (Eq. (1)), this becomes: m
dYO2 − (YO2 + r)m ˙f dt 1 ⇒ tth = ρg V ln m ˙f
Fig. 10. Molar oxygen yield (XO2 ) in the lower part of the vessel.
tilation rate, the fuel mass loss rate and the fire scenario) can still remain over-ventilated when the ventilation procedure is applied or can be under-ventilated without the ventilation procedure. In these two cases, no change in extinction mode would be induced by applying ventilation procedures. A general conclusion can be drawn, the ventilation procedures tested help reduce the amount of fresh air arriving at the fire and hence move the combustion regime from over-ventilated towards under-ventilated. 5.3. Fire duration The ventilation procedures tested shorten the duration of the fire, from approximately 26 min for the reference test to 20 min for both FLIP10.1 and FLIP10.3 tests. This result is explained as follows. The fire duration is mainly dependant on the fuel mass loss rate and the amount of oxygen available to the combustion reaction. Inlet valve closure induces two mechanisms: it contributes to interrupting the supply of oxygen (only the oxygen in the vessel is available for combustion), and it brings about a decrease of the fuel mass loss rate, and thus, the speed of oxygen consumption required for the combustion. As the interruption of the oxygen supply is the most influencing mechanism, the net result is decreased fire duration. The fire duration obtained with the ventilation procedures is compared to a theoretical fire duration tth
YOo 2 (to ) + r YOext2 (text ) + r
(5)
Although this equation corresponds to an idealistic approach (the parameter r, the fuel mass loss rate and gas density are assumed constant), comparison with the FLIP test durations is interesting for practical design purposes. The following values are chosen for this comparison: 0.23 (=0.21 × 32/29) for YOo 2 , the average oxygen mass yield in the vessel before ignition (cf. Fig. 10), 0.165 (=0.15 × 32/29) for YOext2 , the average oxygen mass yield in the vessel at extinction (cf. Fig. 10), 400 m3 for V the vessel volume, 1 kg/m3 for ρg the density of the gas at 75 ◦ C, 3 for r the “oxygen to fuel” stoichiometric ratio of the THP/TBP mixture and (17 × 0.4)g/s for the average fuel mass loss rate. The fire duration calculated using Eq. (5) is about 20 min (1186 s), and thus correlates well with the measured values (1203 and 1200 s). The ventilation procedures studied contribute to reducing the fire duration towards a theoretical limit defined by Eq. (5). The previous conclusions should be considered in the light of two parameters: the initial ventilation airflow rate and the instant the inlet is closed. For example for scenarios for which the initial ventilation air flow rates is very low, the fire scenario will be closer to the theoretical one used to elaborate relation (5) and considering no flow at the inlet. The use of the ventilation procedure in this case will not reduce the fire duration significantly; the fire duration being already close to the minimum one. Concerning the inlet closure time influence, the earlier inlet closure happens, the more significant the inlet closure effect on the fire duration. During the FLIP tests, inlet closure occurs quite early, 100 s after ignition-period corresponding to about 5% of the total fire duration without any ventilation
2164
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
procedure. The fire scenario then becomes very similar to a configuration involving no ventilation at all (vessel closed before ignition), and therefore, the effect of inlet closure on the fire duration is clearly enhanced. If inlet closure occurs very late during the fire, its effect would be lessened. 5.4. Discussion concerning the effect of the ventilation procedure on the combustion products Closure of the inlet tends to reduce the fire duration and the fuel mass loss rate, and hence, leads to a reduction of the total heat released within the vessel and the total amount of products generated by combustion. Note that carbon monoxide or unburned product concentrations do not necessarily follow the same trend since closing the inlet valve, and thus, reducing the oxygen concentration, may also modify the chemical reaction toward higher carbon monoxide or unburned yields. Gas concentration measurements were taken during these tests but no clear conclusions could be drawn regarding this matter.
6. Effect of the ventilation procedures on vessel pressure As vessel pressure is key to the level of confinement, it is always examined with interest from a nuclear safety viewpoint and especially when ventilation procedures are applied. 6.1. With inlet valve closure sequence The pressure history of the test FLIP10.1 is presented in Fig. 11. The initial pressure is slightly higher, as for the reference test (approximately −600 Pa instead of −800 Pa), due to a slight change in the setting of the admission valve and in the air-tightness of the vessel (this change has no consequence on the conclusions drawn). The inlet valve closes during the overpressure peak and induces a rapid fall of the pressure to a new equilibrium level (approximately −700 Pa) corresponding to the pressure level at the dilution nodes, which is 100 Pa below the initial pressure level within the vessel (approximately −600 Pa). The pressure then remains stable during the fire until extinction when the typical low-pressure peak is observed.
Fig. 11. Pressure history with inlet closure ventilation control mode of test FLDP10.1.
By modifying the ventilation system configuration, inlet valve closure changes the pressure level in the vessel. The vessel behaves as a ‘dead volume’ and the entire airflow passes through the dilution branch. When the inlet branch is closed, the pressure within the vessel and that at the dilution node equalize. With the PLUTON facility configuration (especially with a fan installed on the exhaust branch), the pressure change corresponds to a slight decrease (−100 Pa) but the vessel relative pressure remains negative (−700 Pa) and confinement is maintained. The amplitude of the decrease (about 100 Pa) depends on the ventilation characteristics of the flow passing through the vessel (airflow resistances and vessel air leakage) and the flow rate within the vessel compared to the one in the dilution branch. Since airflow in the dilution branch is much higher (10 times) than that within the inlet branch before closure, the pressure decrease due to the inlet valve closure is small. 6.2. With inlet and exhaust closures sequence On the contrary of the sequence with inlet closure only, the procedure with both inlet and exhaust closure shows a totally different behaviour. The FLIP10.3 test pressure history is presented in Fig. 12. Up until inlet valve closure, it is similar to that observed for the FLIP10.1 test. Once the exhaust valve is fully closed (t = 247 s), vessel pressure increases rapidly. The
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2165
and Po is the pressure outside the vessel (conventionally, atmospheric pressure). Assuming that gas density varies only with temperature (pressure variation are too small to affect the density) so that ρ = (PM)/(RT) = (Po M)/(RT), the quasi-constant pressure level reached after exhaust valve closure can be expressed as:
1/n ˙c−Q ˙w Q P − Po = γ C Po γ−1
Fig. 12. Pressure history with both inlet and exhaust closure ventilation mode.
pressure rise at thus point is the system’s response to the fact that the exhaust branch can no longer release gas energy. It is the result of the net balance between the fire heat release, thermal losses through the walls and gas leakages as illustrated by the equation deduced from the energy balance (relation (2)): V d ˙c−Q ˙w−Q ˙ leak P =Q γ − 1 dt
(6)
For a given fire heat release rate, the pressure level achieved depends on thermal losses through the walls and on the vessel air-tightness (through the C value). However, for the FLIP tests, after reaching a quasiconstant plateau, the pressure continues to increase slightly. This effect is attributed to the continuous rise in temperature such that the heat release rate remains greater than the losses and the leaks and, during the last 200 s, to the small increase of the heat release rate. Upon extinction, pressure falls rapidly due to the sudden interruption of heat release that causes a negative value of dP/dt (cf. Eq. (5)). The pressure drops until a low-pressure limit is reached, a level characterized at this point by equilibrium between the thermal losses through the walls and the mass leaks. The amplitude of the low-pressure peak can thus be written:
Later, vessel pressure stabilises at about 3200 Pa. This quasi-constant level is brought about by an equilibrium reached between fire heat release, gas leakages and wall heat losses, term that increases gradually due to the increasing vessel gas temperature (as illustrated by the grey line in Fig. 12). The energy balance is then: ˙c=Q ˙w+Q ˙ leak Q
(9)
(7)
In order to formulate an expression of the pressure amplitude of the plateau, the pressure compartment is explicitly introduced into the previous relation. Only the term corresponding to energy lost through gas leakages depends on the pressure and is expressed as: ˙ leak = v˙ leak ρg cp Tg = v˙ leak Po γ Q (8) γ −1 with v˙ leak = C(P − Po )n , C is the flow coefficient (0.05 m3 /(hPan )), n an exponent (found experimentally to be close to 1 for the pressure range considered)
˙w+Q ˙ leak Q
˙w −Q = 0 ⇒ P − Po = γ CPo γ−1
1/n (10)
The influence of the fire heat release is implicitly included in the term corresponding to thermal losses through the walls. Indeed, the higher the fire heat release rate, the higher the gas temperature and thus the higher the thermal losses. In the FLIP10.3 test, to avoid exceeding the vessel’s low pressure limit (about −2500 Pa), nitrogen is injected into the vessel several times just after extinction at approximately 400 m3 /h (cf. two successive pressure increases after extinction on Fig. 12). Without these nitrogen injections, it has been estimated that the lowpressure peak would have been around −5000 Pa assuming a value of 25 kW for total thermal losses. This last value is calculated from the test data and corresponds to an average heat flux of 80 W/m2 through the 323 m2 of the vessel wall.
2166
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Analysis of the FLIP10.3 test pressure history clearly shows that closure of both inlet and exhaust valves has a significant impact on pressure within the vessel, firstly inducing a pressure rise after exhaust closure, and secondly, a low-pressure peak just after extinction. According to Eq. (9), the amplitude of the over-pressure peak depends on the relative weight of the fire heat release, the leaks and the thermal losses through the walls. In some cases, the pressure gauge can become positive (as in the FLIP10.3 test), and thus lead to loss of vessel confinement normally ensured by a negative pressure. This consequence demonstrates the practical value of studying pressure variations for fire safety assessment. 6.3. Influence of the exhaust valve closure time During the FLIP test campaign, two tests (named FLIP10.2 and FLIP10.3) involving the closure of both inlet and exhaust valves were performed. For the test FLIP10.2, closure of the inlet and exhaust valves occurs 35 s earlier than the FLIP10.3 test (inlet closure at 65 s instead of 100 s and exhaust closure at 175 s instead of 210 s, cf. Fig. 13). In addition, the exhaust valve is re-opened for few seconds at t = 290 s after ignition (because of safety reason). Comparing these two tests can highlight the effect of the exhaust closure time on the pressure level reached
Fig. 13. Effect of the exhaust valve closure time on the pressure history.
in the vessel once both valves have been closed. Indeed, the two tests allow comparing three exhaust valves closures at three instants in time (cf. Fig. 13). The results indicate that the sooner the exhaust valves is closed the stiffer the pressure rise and the higher the amplitude of the stabilized pressure. This behaviour depends on the amplitude of the gas temperature within the vessel when the exhaust valve is closed. As the temperature difference (between the vessel and the outside) is proportional to the heat flux through the walls, the higher the temperature within the vessel and the higher the energy dissipated through the walls, the weaker the pressure increase within the vessel. Application of the perfect gas law confirms this analysis in another way. The pressure change between the instant of exhaust closure (subscript 1) and the instant where the pressure stabilizes (subscript 2) is expressed as: P2 − P 1 =
cv (γ − 1) [m1 (T2 − T1 ) + T2 (m2 − m1 )] V (11)
The pressure increase is the result of two effects: gas temperature increase (due to the fire heat release) weighted by the mass variation due to the net balance between the mass of fuel and the gas leakages. Eq. (11) illustrates that a high level of gas temperature T1 (when the exhaust valve is closed) decreases the change of temperature (T2 − T1 ), and thus, the pressure increase, thus confirming experimental observations. A similar analysis is proposed to establish how the closure time could also affect the low-pressure peak amplitude at extinction. Depending on the time the vessel is closed, the gas temperature (and thus the gas mass within the vessel) will vary: the later the vessel is closed the smaller the amount of gas enclosed within the vessel. Considering that thermal losses are the same in both cases, the pressure drop at extinction will be higher if vessel closure occurs later during the fire (i.e. small amount of gas enclosed within the vessel). On the contrary, vessel closure occurring very early during the fire will lead to a less severe pressure at extinction. No data being available from the FLIP test (because of nitrogen injections), this explanation needs to be confirmed experimentally. This examination of vessel closure time can be schematically summarised as in Fig. 14. Earlier exhaust closure (after inlet closure) leads to a high-pressure increase but a small pressure drop upon extinction. On the
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2167
contrary, later exhaust closure leads to a smaller pressure increase but a large pressure drop at extinction. 7. Effect of the ventilation procedures on gas temperature The effect of the ventilation procedures on gas temperature within the vessel is investigated by comparing gas temperature at several locations within the vessel (in the fire plume, in the ceiling jet and outside these two specific flows) for the three tests. Gas temperature variations 5 m from the ground outside both the fire plume and the ceiling jet is illustrated in Fig. 15 for the three tests. No significant change in gas temperature is observed when the inlet valve is closed. Moreover, the temperature variations noted during FLIP10.1 and FLIP10.3 are almost identical indicating that exhaust valve closure has no impact on gas temperature. Differences observed between the reference test FLIP1, and the other two tests are attributed to the variation of the fuel mass loss rate mentioned in the previous section rather than any direct effect of the inlet or exhaust valve closure. Analysis of the vertical gas temperature profiles leads to the same conclusion. A typical example is presented on Fig. 16 illustrating vertical temperature profiles 500 s after ignition. The temperature profiles of the FLIP10.1 and FLIP10.3
Fig. 14. Illustration of the effect on the pressure of the time the exhaust is closed.
Fig. 15. Gas temperature within the vessel 5 m over the ground (TG1 500 is the name of the thermocouple).
tests are remarkably similar although one is measured with the exhaust valve open and the other with the exhaust valve closed. Gas temperatures within the fire plume and the ceiling jet are also analysed and again, no significant changes in gas temperature are observed (Figs. 17 and 18). As concluded previously, the implementation of ventilation procedures had no significant impact on gas temperature in the configuration studied.
Fig. 16. Vertical gas temperature profile at 500 s after ignition.
2168
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Fig. 17. Gas temperature within the fire plume (TF16 010 is the name of the thermocouple located at 3.75 m above the pool and 0.1 m from the wall).
This result can be analysed with the internal energy balance (Eq. (2)) that is reformulated as: d d T + Tcv m dt dt ˙c+Q ˙ v,in − Q ˙ v,ex − Q ˙w+Q ˙ leak =Q
mcv
(12)
Once the inlet valve is closed, the flow rate in the exhaust branch falls to almost zero (cf. Fig. 7) and the energy lost by the ventilation branches is small compared to the other terms. Closure of the exhaust branch has no effect on the amplitude of the left hand side terms of the energy equation, and hence, on the amplitude of the temperature variation (mcv dT/dt) thus explaining experimental observations. From this analysis, it can be deduced that whatever the ventilation procedure applied, the temperature variation is mainly the result of the net balance between only the fire heat release rate and the thermal losses through walls. The variation of the ventilation terms ˙ v,in − Q ˙ v,ex + Q ˙ leak ) mostly plus the air leakages (Q balanced by the mass variation term (Tcv (dm/dt)), and thus, has a negligible effect on the temperature variation term (mcv (dT/dt)). It should also be noted that although the closure of ventilation branches has a strong impact on the pressure (i.e. on the total internal energy mcv T), it has almost no significant effect on the gas temperature. In fact, this surprising statement may be clearly understood in the light of the definition of “significant effect”. Indeed, considering a closed volume of air and using the perfect gas law, a gas temperature variation of 10 ◦ C, that can be considered low in terms of fire safety, leads to pressure variations of thousands of Pascal (roughly 20 hPa), a value considered significant when dealing with ventilation networks.
8. Conclusions
Fig. 18. Gas temperature within the ceiling jet (TG5 010 is the name of the thermocouple located at 0.1 m below the ceiling and 2.5 m from the wall).
This document presents the effect of two ventilation procedures on the behaviour of a 300 kW pool fire in a ventilated enclosure. The first procedure involves closing the inlet branch at a given time after ignition. The second involves closing first the inlet branch and then the exhaust branch, the enclosure thus being entirely sealed. The experimental results show that closure of the inlet branch contributes to a reduction of the fuel mass loss rate and of the fire duration. Yet, this closure has no significant effect on the pressure within the vessel and on the gas temperature which both remain similar to the scenario without a ventilation control procedure. In addition, vessel relative pressure remains negative thus maintaining vessel confinement. It is also noted that even if the exhaust branch is left open, closure of
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
the inlet branch significantly reduces the airflow rate at the exhaust. The procedure involving exhaust branch closure give rise to the same conclusions regarding the fuel mass loss rate, the fire duration and the gas temperature. However as regards pressure, this ventilation strategy leads to significant pressure variations within the vessel. Pressure increases significantly after the closure and features an important low-pressure peak after the fire extinction. The amplitudes of these two peaks depend on the fire heat release rate, thermal losses through walls and leakages related to the air tightness of the enclosure. Positive gauge pressure in the vessel can, therefore, be expected with this ventilation control procedure, which then may induce a loss of vessel confinement. In addition, the experimental results show that the time at which the exhaust valve is closed impacts the pressure increase, and that the later the exhaust valve is closed, the lower the pressure increase in the vessel but the larger the pressure drop at extinction. Based on the experiments performed, some important features of the two procedures investigated can be identified that could guide safety engineers in selecting suitable ventilation procedures. The procedure involving closing both the inlet and exhaust allows the fire compartment to be isolated entirely but should be applied with caution in terms of pressure variations induced. The inlet closure procedure is an alternative that features similar behaviour but does not involve significant pressure variations. As regards the mass leaving the compartment, the strategy where both inlet and exhaust are closed is not necessarily more efficient. Indeed, upon applying the inlet closure procedure, gas can leave the vessel through the exhaust but at a very low rate; when both the inlet and exhaust are closed, mass still leaves the compartment through gas leakages enhanced by the over-pressure within the vessel. Although the choice of ventilation procedure may depend on other considerations not included in this study (the mechanical resistance of the vessel or valves to
2169
pressure, the type of fuel, . . .), conclusions drawn illustrate that both procedures investigated could help significantly reducing fire hazards. Suitable procedures may be determined by bringing together these conclusions and the specific features of the installation.
Acknowledgements The authors would like to thank all the members of the IRSN Fire Experimentation Laboratory for their contributions to the FLIP test campaign, J. Torero from the University of Edinburgh for his fruitful advice and the COGEMA Company, for its involvement in the project.
References Audouin, L., Tourniaire, B., 1999. New estimation of the thermal interface height in forced-ventilation enclosure fires. In: Proceeding of the Sixth International Symposium of Fire Safety Science, Poitiers. Feng Jiang, Sui-sheng Ye, Jun Yang, 2002. Design of the exhaust system in the confinement of the HTR-10. Nucl. Eng. Des. 218, 209–214. Karlsson, B., Quintiere, J.G., 2001. Enclosure Fire Dynamics. CRC Press. Klote, J.H., 1993. Air Moving Systems and Fire Protection. NISTIR 5227 Report. National Institute of Standards and Technology. Nuclear Air Cleaning Handbook, Fire Protection, US Department of Energy, DOE HDBK 1169.2003 (Chapter 10). Pr´etrel, H., Coutreau, J., Such, J.M., 11–15 June 2001. Thermal plume of large pool fires in confined and forced ventilation enclosures. In: Proceeding of the International Conference on Engineered Fire Protection Design, San Francisco, USA. Quintiere, J.G., 1989. Fundamentals of enclosure fire “zone” models. J. Fire Prot. Eng. 1 (3), 99–119. Rehm, R.G., Forney, G.P., 1994. The pressure equation in zone-fire modelling. Fire Sci. Technol. 14 (1–2), 61–73. Svensson, S., 2002. A study of tactical patterns during fire fighting operations. Fire Saf. J. 37, 673–695. Vaari, J., Hietaniemi, J., 2000. Smoke Ventilation in Operational Fire Fighting, VTT Technical Report No. 419.
Effect of ventilation procedures on the behaviour of a fire compartment scenario H. Pr´etrel ∗ , J.M. Such Institut de Radioprotection et de Sˆuret´e Nucl´eaire (IRSN), Service d’Etude et de Recherches Exp´erimentales sur les Accidents (SEREA), Laboratoire d’Exp´erimentation des Feux (LEF), Centre de Cadarache, 13108 C´edex Saint Paul-lez-Durance, France Received 29 July 2004; received in revised form 23 February 2005; accepted 3 March 2005
Abstract This contribution presents a study on the consequences of applying ventilation procedures during a fire scenario involving a TPH/TBP pool fire in a ventilated enclosure. This research is addressed to fire safety in the nuclear industry in which ventilated enclosures remain a configuration frequently encountered. This work presents experiments comprising a 300 kW liquid pool fire in a 400 m3 vessel connected to an industrial ventilation system featuring one inlet and one exhaust branch. The investigated ventilation procedures consist in closing the inlet branch only or closing both inlet and exhaust branches. The analysis compares fire behaviour with and without the implementation of a ventilation procedure and points out the effects of said procedures on the combustion rate, fire duration and gas temperature within the vessel. It highlights pressure variations within the vessel when both the inlet and exhaust ventilation branches are closed. Conclusions provide practical answers that would be useful when designing appropriate ventilation strategies limiting fire hazards. © 2005 Elsevier B.V. All rights reserved.
1. Introduction The scenario comprising a fire within a ventilated enclosure remains one of the key issues for fire safety assessment in the nuclear industry. Indeed, partitioning is frequently encountered irrespective of the building’s designated use, building reactor, fuel processing or reprocessing plants. Moreover, to ensure confinement, these enclosures are equipped with a ventilation ∗ Corresponding author. Tel.: +33 4 42 25 26 67; fax: +33 4 42 25 48 74. E-mail address: [email protected] (H. Pr´etrel).
0029-5493/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.nucengdes.2005.03.003
air cleaning system that allows negative pressure to be created in the compartments and thus prevent radioactive materials or harmful products from escaping out of the compartment (Jiang et al., 2002; Nuclear Air Cleaning Handbook, 2003). It is for this reason that configurations involving a fire in a mechanically ventilated enclosure are extensively studied in the framework of nuclear fire safety research (Audouin and Tourniaire, 1999; Pr´etrel et al., 2001). In such fire scenarios, the oxygen feeding into the compartment is mainly controlled by the ventilation system; consequently, one procedure that could limit the fire hazards would be to switch the ventilation
2156
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Nomenclature cv Cp C m m ˙ n P ˙ Q r t T V v˙ X Y
heat capacity at constant volume heat capacity at constant pressure coefficient mass rate of mass flow exponent pressure rate of heat release oxygen to fuel stoichiometric ratio time temperature volume rate of volumetric flow molar fraction mass fraction
Indices c ex ext f g in leak o O2 th v w
combustion exhaust extinction fuel gas inlet leakages reference oxygen theoretical ventilation losses
Greek symbols γ isentropic coefficient ρ density system to a special control mode. There are many possible approaches; these depend on the particularities of the enclosure and the expected efficiency (Klote, 1993; Vaari and Hietaniemi, 2000). A typical example is the rapid extraction of smoke (using positive pressure ventilation (PPV) concept) by mechanical fans. This approach limits the increase in gas toxicity and gas temperature (Svensson, 2002). In the case of one ventilated compartment equipped with inlet and exhaust branches, a possible ventilation procedure would be to close the valves situated either upstream or down-
Fig. 1. Schematic representation of the investigated fire scenario and the PLUTON facility.
stream of the fire compartment. In the nuclear industry, this procedure is of particular interest because it allows the compartment to be isolated and thus protects the rest of the installation against the consequences of fire, such as the plugging of filters with smoke or the release of radioactive dust. In order to determine the efficiency of such procedures and to identify the most appropriate one, the behaviour of the compartment and of the pool fire has to be investigated. To address this issue, an experimental study was undertaken within the framework of the FLIP research programme developed by IRSN in collaboration with COGEMA. Fig. 1 illustrates a fire scenario representing actual situations in reprocessing plants where some enclosures are dedicated to chemical processes using flammable liquids. The fire scenario is a pool fire located against one side of a rectangular enclosure. The ventilation network comprises an inlet branch and an exhaust branch each equipped with a closing valve. A fan located on the exhaust branch is used to create negative pressure in the compartment thus ensuring confinement. The location of the pool fire next to the wall, its shape and the ventilation characteristics, are selected in order to be representative of a typical scenario found in the nuclear industry. The research programme, performed at IRSN’s fire test laboratory at Cadarache (France), involved large-scale experiments designed to mimic the real situation as closely as
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
possible. The aims of these tests are to analyse the fire scenario under two ventilation procedures and to point out the consequences on parameters key to fire safety: the rate of combustion, the fire duration, the gas temperature and the pressure within the enclosure.
2. Description of the tests 2.1. Test set-up Tests are carried out in the IRSN PLUTON facility (cf. Fig. 1), which is a 400 m3 reinforced concrete vessel with 0.25 m thick walls. The vessel dimensions are 7.5 m in height and 54 m2 in area. The vessel is connected to an industrial ventilation network made of an inlet and an exhaust branch. The inlet branch is connected to the atmosphere. The flow inlet is located in the lower part of the enclosure (1.5 m from the ground) and is directed towards the pool. The inlet cross-section is 0.4 m2 . The decision regarding the position of the inlet, right in front of the pool, is made to ensure the symmetry of the inlet airflow entering the vessel. A butterfly type valve (from AMRI) is installed 1 m upstream from the vessel inlet. Equipped with an air control actuator, the valve allows the opening to be varied from a control room. The exhaust branch is connected in the upper part of the enclosure, and is equipped with a valve positioned flush with the vessel ceiling that acts as the valve seat. The valve operates vertically via an ac motor enabling the opening to be varied. Downstream, a dilution branch is connected to the exhaust branch thus allowing a wider range of airflow operating conditions. The exhaust branch ends with a 55 kW fan upstream where eight high efficiency particles air (HEPA) filters are installed to collect the soot particles produced by the fire. The pool consists of a rectangular container filled with 50 mm of a liquid fuel mixture named TPH/TBP, proportions 70% of hydrogenated tetra-propylene (C12 H26 ) and 30% of tri-butyl phosphate (C12 H27 O4 P) by volume. It is positioned against a wall at an equal distance from the two opposite sides as shown in Fig. 1. The fire is ignited with a propane burner directed toward the fuel surface.
2157
2.2. Instrumentation The mass fuel loss rate is determined based on pool mass loss measurements taken by the weighing system, SARTORIUS IS-300, located under the pool. The fuel mass loss rate noted mc is calculated based on the pool weight measured at each time interval (approximately 0.5 s). The fire heat release rate, HRR, is considered to be the product of the fuel mass loss rate and the total heat of combustion taken to be 36 MJ/kg for the mixture TPH/TBP. Gas temperature measurements are performed by means of thermocouples trees set in various places within the enclosure in order to characterize the fire plume, the ceiling jet and the vertical gas temperature stratification. Gas temperatures are also measured in the ventilation branches. K type thermocouples, diameter 1.5 mm, are used. The pressure within the compartment is measured with differential pressure transducers mounted flush with the wall enclosure. The airflow rates within the ventilation branches are measured with differential pressure probes (ANNUBAR type) connected to a differential pressure transducer. The air temperature is also measured at the probe location to take the density change into consideration in the airflow calculation. Oxygen concentration is measured with SERVOMEX4100 gas analysers connected to gas sampling systems. 2.3. Ventilation control sequence Two ventilation control sequences, based on actual operating procedures applied in the nuclear industry, are studied. The first sequence, referred to in this paper as inlet valve closure (IVC), involves closing the inlet valve 100 s after pool fire ignition without modifying any of the other ventilation system devices; the fan on the exhaust continues to operate. The 100 s time period corresponds to an estimate of the time that would lapse between fire ignition and fire detection (by detectors that automatically trigger inlet valve closure). The time taken for the inlet valve to close is approximately 5 s in the PLUTON facility, this is realistic as regards valves used in nuclear industry. The second sequence, referred to as inlet and exhaust valves closure (IEVC), involves closing the inlet valve 100 s after pool fire ignition, and then closing the
2158
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Table 1 FLIP test features Test number
FLIP1
FLIP10.1
FLIP10.2
FLIP10.3
Procedure m ˙ c (kg/(sm2 )) HRR (kW) Fire duration (s)
– 0.0251 360 1560
IVC 0.0168 242 1203
IEVC 0.0164 236 1272
IEVC 0.0177 256 1200
exhaust valve after 210 s. The 110 s, allowed to lapse between inlet closure and exhaust closure, is an averaged estimate based on practical considerations such as the time required for an operator in a control room to close the exhaust valve. In the PLUTON facility, the valve is entirely closed after 247 s due to the time the valve needs to move from the open to the closed position (i.e. 37 s). The enclosure then remains nearly sealed until fire extinction (the only gas leakage possible was attributable to the enclosure air-tightness). The fan is still operating and the entire airflow passes through the dilution branch (cf. Fig. 1). 2.4. Typical test procedure Four large-scale tests are considered in this study, their main characteristics are presented in Table 1. For the four tests, 0.02 m3 of fuel (approximately 16 kg) is poured into a 0.4 m2 square shaped container. The ventilation system is adjusted to achieve the required vessel pressure (about −800 Pa) and ventilation rate (about 3 h−1 ). Once the ventilation rate is stabilized, the pool fire is ignited by a propane gas burner that is shut off as soon as the propagation phase on the pool surface is initiated. The flame spreads over the surface until the entire pool area is burning and a fully developed fire is observed. During this period, the vessel is gradually filled with soot particles and visual access to the fire is prevented as soon as the soot particles reach the ground. The fire goes out without any external intervention; the pool fire is extinguished either due to a lack of fuel within the pool or because of oxygen depletion within the compartment. The analysis proposed in the next sections describes the effect of the two ventilation procedures on the fire scenario by examining the airflow rate, the mass fuel rate, gas pressure and gas temperature. The first test, FLIP1, is the reference test. No ventilation procedure is applied. The characteristics of the second test,
FLIP10.1, are identical to those of FLIP1 except that the IVC sequence is applied. The third and fourth tests, FLIP10.2 and FLIP10.3, are conducted following the IEVC sequence.
3. Analysis of the reference test 3.1. Theoretical considerations Before describing the behaviour of the reference test, the physical mechanisms contributing to modification of the vessel pressure, the gas temperature and the oxygen concentration are recalled through classical energy and mass balance equations commonly defined in compartment fires (Quintiere, 1989; Karlsson and Quintiere, 2001). The system examined, sketched in Fig. 2, is the volume of gas located in the vessel. According to the total mass conservation: d m= dt
m ˙f Fuel mass loss rate from combustion
+
m ˙ v,in − m ˙ v,ex gain and loss from the ventilation branches
+
m ˙ leak
(1)
gain or loss from leakages
Fig. 2. Schematic representation of the physical mechanisms involved during the fire scenario.
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2159
The time variation of the mass of gas in the vessel is the result of competition between the mass fuel rate contributed to the system (gas resulting from the evaporation of fuel that then undergoes the combustion reaction) m ˙ f , the inlet and exhaust airflows m ˙ v,in − m ˙ v,ex mass flow rate from gas leakages mleak . This last term, driven by the pressure within the vessel, can be positive or negative (the airflow enters or leaves the system) depending on the vessel gauge pressure. According to total energy conservation: d V d (mcv T ) = P dt γ − 1 dt internal energy variation
=
˙c Q
+
HRR from the fire
−
˙ v,in − Q ˙ v,ex Q gain or loss from the ventilation branches
˙w Q thermal losses through the walls
+
˙ leak Q gain/loss from leakages
(2) The time variation of the gas internal energy is the re˙ c, sult of a balance between the fire heat release rate, Q the energy gained or lost by the ventilation branches, ˙ v,in − Q ˙ v,ex , the thermal losses through the walls Q ˙w Q and the energy gained or lost (depending on the vessel ˙ leak . Considering the pressure) through gas leakages Q perfect gas law, the internal energy (mcv T) can also be expressed in terms of pressure as PV/(γ − 1) (γ being the isentropic gas coefficient), thus, the energy balance leads to the pressure equation (Rehm and Forney, 1994) as formulated with Eq. (2). These conservation equations clearly illustrate that a ventilation procedure, by setting the energy and mass fluxes passing through one or both ventilation branches to zero, modifies the mass, temperature and pressure within the vessel. Analysis of the test data will quantify the amplitude of these variations. 3.2. Description of the reference test The combustion of the TPH/TBP shows a typical behaviour illustrated on Fig. 3. The fuel mass loss rate increases over the first 160 s until it reaches a maximum value and then follows a slow decay until 1380 s, a few instants before extinction. At this time, just before extinction, a sudden
Fig. 3. Fuel mass loss rate and oxygen yield during the reference test FLIP1.
increase in the mass loss rate is observed. This phenomenon (also reproduced at a small scale with Cone Calorimeter tests) is attributed to the behaviour of the TBP fuel alone that experiences a very high evaporation rate at extinction when it occurs by lack of fuel. Concerning oxygen concentration, as combustion continues, the oxygen concentration within the enclosure decreases gradually until it stabilizes at a value of 18.5% after 1000 s of fire (cf. Fig. 3). Upon extinction because of the sudden increase of fuel mass loss rate, the oxygen concentration also dropped abruptly and then increased after extinction due to the fresh air introduced by ventilation. Pressure evolution within the enclosure follows also a typical behaviour observed in a confined and ventilated enclosure (cf. Fig. 4). Before ignition, the pressure within the vessel is constant and equal to the negative relative pressure level induced by the fan (−800 Pa for these tests). Starting as of ignition and during flame propagation on the pool, the room pressure rises gradually to a peak, and then falls back to the value set before ignition. After this peak and during the fully developed stage of the fire, the pressure remains almost constant. Upon extinction, an abrupt fall in pressure is observed. The over-pressure peak upon ignition and the lowpressure peak upon extinction are explained through the pressure equation (Eq. (2)). The over-pressure peak is the result of the gas expansion mechanism caused by
2160
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Fig. 5. Airflow rates in the ventilation branches during the test FLIP1. Fig. 4. Pressure within the vessel during the reference test FLBP1.
the rapid release of heat within the vessel gradually compensated, firstly by the net energy lost through the ventilation branches (net balance between inlet and exhaust), and secondly, by heat losses through the enclosure walls. The low-pressure peak is the results of the gas contraction mechanism caused by the sudden interruption of the fire heat release gradually compensated firstly, by the net energy gained through the ventilation branches, and secondly, by the energy contributed by gas leakages (entrance of mass because the vessel pressure is negative at this time). The amplitude of the over-pressure peak upon ignition depends on the heat release of the fire, the thermal losses though the walls, the features of the ventilation branches and compartment air-tightness. The airflows in the inlet and exhaust branches follow the same trend as the vessel pressure (cf. Fig. 5). Before ignition, the airflow rate is roughly equal to 0.33 m3 /s (or 1200 m3 /h), which corresponds to the initial conditions imposed. The discrepancy between the flow rates in the two branches (that theoretically should be equal because of the mass conservation) is the result of measurement uncertainties in the exhaust branch (only observed in the test FLIP1). Once the pool fire is ignited, a sudden increase in the flow rate is observed in the exhaust branch, this was the consequence of an overpressure peak in the vessel. For these FLIP tests, the pressure was not observed to have any effect on the
inlet flow rate, this was because the airflow resistance of the inlet branch was greater than that of the exhaust branch. The reference test is also characterised by gas temperature variations within the vessel that are typical of compartment fire (Pr´etrel et al., 2001). The gas temperature variation over time is illustrated in Fig. 6 for four locations (within the fire plume, the ceiling jet, the hot layer and the cold layer). The evolution of the gas temperature within the enclosure follows the flow motion. From the fire plume to the ceiling jet and then within the hot layer and the cold layer, the gas temperature decreases gradually. The gas reaches a temperature of
Fig. 6. Gas temperature during the test FLIP1.
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2161
350 ◦ C within the plume and 170 ◦ C in the ceiling jet. The gas temperature variation over time follows the same trend as the mass fuel rate, and thus the rate of heat release, and is typical of compartment fire. After ignition, the temperature rises rapidly because of the sudden release of heat, and then stabilizes gradually due to the continuous increase of the thermal losses through the walls. Based on this description of the reference test, the next sections will analyse how ventilation procedures can modify this behaviour.
4. Description of the ventilation procedures through the airflow rate The analysis of the airflow rates through the ventilation branches illustrates the two ventilation procedures chosen for this study. Figs. 7 and 8 show the time history of the airflow rates in the inlet and exhaust branches for the sequence with inlet closure only (IVC, test FLIP10.1) and both inlet and exhaust closures (IEVC, test FLIP10.3), respectively. The reference time (t = 0) corresponds to the pool fire ignition. 4.1. Sequence IVC The closure of the inlet valve is clearly identified at t = 100 s leading to zero flow rate in the inlet branch. As a result, the exhaust flow rate experiences a dramatic
Fig. 8. Airflow rates in the ventilation branches during the test FLIP10.3.
initial drop followed by a gradual decay towards zero at t = 420 s. This behaviour is analysed with the mass conservation (Eq. (1)). Once the inlet valve is closed, the exhaust airflow is generated by the balance between gas mass variation over time (due to gas expansion) minus the fuel mass loss rate and the gas leakage mass rate. Between 100 and 420 s, gas temperature continues to rise thus inducing airflow at the exhaust (due to thermal gas expansion). Gas temperature gradually achieves a steady state value and the exhaust flow rate is then attributable to the fuel mass loss rate and leaks only. For these tests, this airflow rate remain within the sensitivity range of the exhaust branch measurement devices (about 0.01 kg/s) as observed after t = 420 s. The practical conclusion that may be drawn from this result is that even when the exhaust branch is open, the amount of mass that leaves the compartment is very small. 4.2. Sequence IEVC
Fig. 7. Airflow rates in the ventilation branches during the test FLIP10.1.
The variation of airflow over time during the IEVC sequence is similar to that observed for the IVC sequence. The closure of the inlet valve results in a drastic drop of the exhaust flow rate, which is followed by a slower decay. The exhaust valve closure process starts at 210 s and leads to complete shut down of the exhaust flow at 247 s (37 s needed for the valve to close). At this time, the gas mass variation within the vessel becomes
2162
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
the net balance between the fuel mass loss rate brought to the system and the mass rate of gas leakages lost by the system. This net balance can be positive or negative depending on the fire HRR and on leakage characteristics. As an example, for the test FLIP10.3, the total net balance is determined using the following relation (deduced from Eq. (1)): text d m(t = 247) − n(text ) = m dt t=247 dt text = (m ˙c−m ˙ leak ) dt (3)
5. Effect of the ventilation procedures on fire characteristics
directed towards the pool fire. Both effects reduce the fuel mass rate significantly. On the contrary, closure of the exhaust valve has no noticeable effect on the fuel mass loss rate. Once the inlet valve is closed, the airflow rate in the exhaust branch is so low that the configuration of the exhaust valve (open or closed) has no immediate effect on the combustion process. The analysis is confirmed looking at the average fuel mass loss rate (indicated in Table 1). Without the ventilation procedure, the total average fuel mass loss rate is about 0.0251 kg/(s m2 ). With the ventilation procedure, it drops down to 0.0168 kg/(s m2 ) for FLIP10.1 and 0.0177 kg/(s m2 ) for FLIP10.3. These conclusions are mostly qualitative since the reduction of the burning rate depends also on the position of the inlet with respect to the pool fire. The configuration adopted for FLIP involved blowing the inlet airflow directly on the pool fire, in others configurations, the effect of the ventilation procedures on the fuel burning rate can be amplified or weakened.
5.1. Fuel mass loss rate
5.2. Extinction mode
Closing the inlet valve during the fire scenario lead to a sudden decrease (approximately 25%) of the fuel mass loss rate (cf. Fig. 9 presenting the variation of mass loss rate over time for all three tests). This is the consequence of two mechanisms induced by inlet valve closure, first, cut-off of air feeding the lower part of the vessel, and second, shut off of dynamic fresh airflow
According to the FLIP tests, ventilation procedures can also alter the mode of extinction. In the reference test FLIP1, extinction occurs when all the fuel is burnt, but for the tests FLIP 10.1 and FLIP10.3 (with inlet and exhaust closures, respectively), extinction is induced by lack of oxygen. These two extinction modes are illustrated in Fig. 10 that presents oxygen yield over time within the vessel. In the absence of a ventilation procedure, at extinction the molar oxygen yield is around 18% (certainly above the limiting oxygen index, LOI, that characterizing the extinction limit) and extinction occurs through lack of fuel. With the ventilation procedures, extinction occurs by lack of oxygen indicating that the oxygen yield within the vessel definitely reached the LOI before the entire mass of fuel is burnt. The molar oxygen concentration measured at extinction is approximately 15 vol%. However, the change in the extinction mode is specific to the FLIP tests and cannot be generalized as a systematic effect of the use of ventilation procedures. Indeed, the use of the inlet closure procedure, modifying the oxygen feeding, tends to move the fire scenario from an over-ventilated situation to an under-ventilated one and then changes the extinction mode. However, the combustion regime (that depends on the initial ven-
t=247
The integral evaluated from the measurements gives a loss of mass equal to 5.3 kg that corresponds to about 1% of the initial mass of gas in the vessel.
Fig. 9. Variation of fuel mass loss rate with time.
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2163
considering a vessel equipped with an exhaust only (no inlet flow, m ˙ v,in = 0 and no leaks). Based on these assumptions, the theoretical fire duration deduced from the oxygen mass conservation may be written: d ˙ f − Y O2 m YO2 m = −r m ˙ v,ex dt
(4)
From the previous equation and using the total mass conservation (Eq. (1)), this becomes: m
dYO2 − (YO2 + r)m ˙f dt 1 ⇒ tth = ρg V ln m ˙f
Fig. 10. Molar oxygen yield (XO2 ) in the lower part of the vessel.
tilation rate, the fuel mass loss rate and the fire scenario) can still remain over-ventilated when the ventilation procedure is applied or can be under-ventilated without the ventilation procedure. In these two cases, no change in extinction mode would be induced by applying ventilation procedures. A general conclusion can be drawn, the ventilation procedures tested help reduce the amount of fresh air arriving at the fire and hence move the combustion regime from over-ventilated towards under-ventilated. 5.3. Fire duration The ventilation procedures tested shorten the duration of the fire, from approximately 26 min for the reference test to 20 min for both FLIP10.1 and FLIP10.3 tests. This result is explained as follows. The fire duration is mainly dependant on the fuel mass loss rate and the amount of oxygen available to the combustion reaction. Inlet valve closure induces two mechanisms: it contributes to interrupting the supply of oxygen (only the oxygen in the vessel is available for combustion), and it brings about a decrease of the fuel mass loss rate, and thus, the speed of oxygen consumption required for the combustion. As the interruption of the oxygen supply is the most influencing mechanism, the net result is decreased fire duration. The fire duration obtained with the ventilation procedures is compared to a theoretical fire duration tth
YOo 2 (to ) + r YOext2 (text ) + r
(5)
Although this equation corresponds to an idealistic approach (the parameter r, the fuel mass loss rate and gas density are assumed constant), comparison with the FLIP test durations is interesting for practical design purposes. The following values are chosen for this comparison: 0.23 (=0.21 × 32/29) for YOo 2 , the average oxygen mass yield in the vessel before ignition (cf. Fig. 10), 0.165 (=0.15 × 32/29) for YOext2 , the average oxygen mass yield in the vessel at extinction (cf. Fig. 10), 400 m3 for V the vessel volume, 1 kg/m3 for ρg the density of the gas at 75 ◦ C, 3 for r the “oxygen to fuel” stoichiometric ratio of the THP/TBP mixture and (17 × 0.4)g/s for the average fuel mass loss rate. The fire duration calculated using Eq. (5) is about 20 min (1186 s), and thus correlates well with the measured values (1203 and 1200 s). The ventilation procedures studied contribute to reducing the fire duration towards a theoretical limit defined by Eq. (5). The previous conclusions should be considered in the light of two parameters: the initial ventilation airflow rate and the instant the inlet is closed. For example for scenarios for which the initial ventilation air flow rates is very low, the fire scenario will be closer to the theoretical one used to elaborate relation (5) and considering no flow at the inlet. The use of the ventilation procedure in this case will not reduce the fire duration significantly; the fire duration being already close to the minimum one. Concerning the inlet closure time influence, the earlier inlet closure happens, the more significant the inlet closure effect on the fire duration. During the FLIP tests, inlet closure occurs quite early, 100 s after ignition-period corresponding to about 5% of the total fire duration without any ventilation
2164
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
procedure. The fire scenario then becomes very similar to a configuration involving no ventilation at all (vessel closed before ignition), and therefore, the effect of inlet closure on the fire duration is clearly enhanced. If inlet closure occurs very late during the fire, its effect would be lessened. 5.4. Discussion concerning the effect of the ventilation procedure on the combustion products Closure of the inlet tends to reduce the fire duration and the fuel mass loss rate, and hence, leads to a reduction of the total heat released within the vessel and the total amount of products generated by combustion. Note that carbon monoxide or unburned product concentrations do not necessarily follow the same trend since closing the inlet valve, and thus, reducing the oxygen concentration, may also modify the chemical reaction toward higher carbon monoxide or unburned yields. Gas concentration measurements were taken during these tests but no clear conclusions could be drawn regarding this matter.
6. Effect of the ventilation procedures on vessel pressure As vessel pressure is key to the level of confinement, it is always examined with interest from a nuclear safety viewpoint and especially when ventilation procedures are applied. 6.1. With inlet valve closure sequence The pressure history of the test FLIP10.1 is presented in Fig. 11. The initial pressure is slightly higher, as for the reference test (approximately −600 Pa instead of −800 Pa), due to a slight change in the setting of the admission valve and in the air-tightness of the vessel (this change has no consequence on the conclusions drawn). The inlet valve closes during the overpressure peak and induces a rapid fall of the pressure to a new equilibrium level (approximately −700 Pa) corresponding to the pressure level at the dilution nodes, which is 100 Pa below the initial pressure level within the vessel (approximately −600 Pa). The pressure then remains stable during the fire until extinction when the typical low-pressure peak is observed.
Fig. 11. Pressure history with inlet closure ventilation control mode of test FLDP10.1.
By modifying the ventilation system configuration, inlet valve closure changes the pressure level in the vessel. The vessel behaves as a ‘dead volume’ and the entire airflow passes through the dilution branch. When the inlet branch is closed, the pressure within the vessel and that at the dilution node equalize. With the PLUTON facility configuration (especially with a fan installed on the exhaust branch), the pressure change corresponds to a slight decrease (−100 Pa) but the vessel relative pressure remains negative (−700 Pa) and confinement is maintained. The amplitude of the decrease (about 100 Pa) depends on the ventilation characteristics of the flow passing through the vessel (airflow resistances and vessel air leakage) and the flow rate within the vessel compared to the one in the dilution branch. Since airflow in the dilution branch is much higher (10 times) than that within the inlet branch before closure, the pressure decrease due to the inlet valve closure is small. 6.2. With inlet and exhaust closures sequence On the contrary of the sequence with inlet closure only, the procedure with both inlet and exhaust closure shows a totally different behaviour. The FLIP10.3 test pressure history is presented in Fig. 12. Up until inlet valve closure, it is similar to that observed for the FLIP10.1 test. Once the exhaust valve is fully closed (t = 247 s), vessel pressure increases rapidly. The
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2165
and Po is the pressure outside the vessel (conventionally, atmospheric pressure). Assuming that gas density varies only with temperature (pressure variation are too small to affect the density) so that ρ = (PM)/(RT) = (Po M)/(RT), the quasi-constant pressure level reached after exhaust valve closure can be expressed as:
1/n ˙c−Q ˙w Q P − Po = γ C Po γ−1
Fig. 12. Pressure history with both inlet and exhaust closure ventilation mode.
pressure rise at thus point is the system’s response to the fact that the exhaust branch can no longer release gas energy. It is the result of the net balance between the fire heat release, thermal losses through the walls and gas leakages as illustrated by the equation deduced from the energy balance (relation (2)): V d ˙c−Q ˙w−Q ˙ leak P =Q γ − 1 dt
(6)
For a given fire heat release rate, the pressure level achieved depends on thermal losses through the walls and on the vessel air-tightness (through the C value). However, for the FLIP tests, after reaching a quasiconstant plateau, the pressure continues to increase slightly. This effect is attributed to the continuous rise in temperature such that the heat release rate remains greater than the losses and the leaks and, during the last 200 s, to the small increase of the heat release rate. Upon extinction, pressure falls rapidly due to the sudden interruption of heat release that causes a negative value of dP/dt (cf. Eq. (5)). The pressure drops until a low-pressure limit is reached, a level characterized at this point by equilibrium between the thermal losses through the walls and the mass leaks. The amplitude of the low-pressure peak can thus be written:
Later, vessel pressure stabilises at about 3200 Pa. This quasi-constant level is brought about by an equilibrium reached between fire heat release, gas leakages and wall heat losses, term that increases gradually due to the increasing vessel gas temperature (as illustrated by the grey line in Fig. 12). The energy balance is then: ˙c=Q ˙w+Q ˙ leak Q
(9)
(7)
In order to formulate an expression of the pressure amplitude of the plateau, the pressure compartment is explicitly introduced into the previous relation. Only the term corresponding to energy lost through gas leakages depends on the pressure and is expressed as: ˙ leak = v˙ leak ρg cp Tg = v˙ leak Po γ Q (8) γ −1 with v˙ leak = C(P − Po )n , C is the flow coefficient (0.05 m3 /(hPan )), n an exponent (found experimentally to be close to 1 for the pressure range considered)
˙w+Q ˙ leak Q
˙w −Q = 0 ⇒ P − Po = γ CPo γ−1
1/n (10)
The influence of the fire heat release is implicitly included in the term corresponding to thermal losses through the walls. Indeed, the higher the fire heat release rate, the higher the gas temperature and thus the higher the thermal losses. In the FLIP10.3 test, to avoid exceeding the vessel’s low pressure limit (about −2500 Pa), nitrogen is injected into the vessel several times just after extinction at approximately 400 m3 /h (cf. two successive pressure increases after extinction on Fig. 12). Without these nitrogen injections, it has been estimated that the lowpressure peak would have been around −5000 Pa assuming a value of 25 kW for total thermal losses. This last value is calculated from the test data and corresponds to an average heat flux of 80 W/m2 through the 323 m2 of the vessel wall.
2166
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Analysis of the FLIP10.3 test pressure history clearly shows that closure of both inlet and exhaust valves has a significant impact on pressure within the vessel, firstly inducing a pressure rise after exhaust closure, and secondly, a low-pressure peak just after extinction. According to Eq. (9), the amplitude of the over-pressure peak depends on the relative weight of the fire heat release, the leaks and the thermal losses through the walls. In some cases, the pressure gauge can become positive (as in the FLIP10.3 test), and thus lead to loss of vessel confinement normally ensured by a negative pressure. This consequence demonstrates the practical value of studying pressure variations for fire safety assessment. 6.3. Influence of the exhaust valve closure time During the FLIP test campaign, two tests (named FLIP10.2 and FLIP10.3) involving the closure of both inlet and exhaust valves were performed. For the test FLIP10.2, closure of the inlet and exhaust valves occurs 35 s earlier than the FLIP10.3 test (inlet closure at 65 s instead of 100 s and exhaust closure at 175 s instead of 210 s, cf. Fig. 13). In addition, the exhaust valve is re-opened for few seconds at t = 290 s after ignition (because of safety reason). Comparing these two tests can highlight the effect of the exhaust closure time on the pressure level reached
Fig. 13. Effect of the exhaust valve closure time on the pressure history.
in the vessel once both valves have been closed. Indeed, the two tests allow comparing three exhaust valves closures at three instants in time (cf. Fig. 13). The results indicate that the sooner the exhaust valves is closed the stiffer the pressure rise and the higher the amplitude of the stabilized pressure. This behaviour depends on the amplitude of the gas temperature within the vessel when the exhaust valve is closed. As the temperature difference (between the vessel and the outside) is proportional to the heat flux through the walls, the higher the temperature within the vessel and the higher the energy dissipated through the walls, the weaker the pressure increase within the vessel. Application of the perfect gas law confirms this analysis in another way. The pressure change between the instant of exhaust closure (subscript 1) and the instant where the pressure stabilizes (subscript 2) is expressed as: P2 − P 1 =
cv (γ − 1) [m1 (T2 − T1 ) + T2 (m2 − m1 )] V (11)
The pressure increase is the result of two effects: gas temperature increase (due to the fire heat release) weighted by the mass variation due to the net balance between the mass of fuel and the gas leakages. Eq. (11) illustrates that a high level of gas temperature T1 (when the exhaust valve is closed) decreases the change of temperature (T2 − T1 ), and thus, the pressure increase, thus confirming experimental observations. A similar analysis is proposed to establish how the closure time could also affect the low-pressure peak amplitude at extinction. Depending on the time the vessel is closed, the gas temperature (and thus the gas mass within the vessel) will vary: the later the vessel is closed the smaller the amount of gas enclosed within the vessel. Considering that thermal losses are the same in both cases, the pressure drop at extinction will be higher if vessel closure occurs later during the fire (i.e. small amount of gas enclosed within the vessel). On the contrary, vessel closure occurring very early during the fire will lead to a less severe pressure at extinction. No data being available from the FLIP test (because of nitrogen injections), this explanation needs to be confirmed experimentally. This examination of vessel closure time can be schematically summarised as in Fig. 14. Earlier exhaust closure (after inlet closure) leads to a high-pressure increase but a small pressure drop upon extinction. On the
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
2167
contrary, later exhaust closure leads to a smaller pressure increase but a large pressure drop at extinction. 7. Effect of the ventilation procedures on gas temperature The effect of the ventilation procedures on gas temperature within the vessel is investigated by comparing gas temperature at several locations within the vessel (in the fire plume, in the ceiling jet and outside these two specific flows) for the three tests. Gas temperature variations 5 m from the ground outside both the fire plume and the ceiling jet is illustrated in Fig. 15 for the three tests. No significant change in gas temperature is observed when the inlet valve is closed. Moreover, the temperature variations noted during FLIP10.1 and FLIP10.3 are almost identical indicating that exhaust valve closure has no impact on gas temperature. Differences observed between the reference test FLIP1, and the other two tests are attributed to the variation of the fuel mass loss rate mentioned in the previous section rather than any direct effect of the inlet or exhaust valve closure. Analysis of the vertical gas temperature profiles leads to the same conclusion. A typical example is presented on Fig. 16 illustrating vertical temperature profiles 500 s after ignition. The temperature profiles of the FLIP10.1 and FLIP10.3
Fig. 14. Illustration of the effect on the pressure of the time the exhaust is closed.
Fig. 15. Gas temperature within the vessel 5 m over the ground (TG1 500 is the name of the thermocouple).
tests are remarkably similar although one is measured with the exhaust valve open and the other with the exhaust valve closed. Gas temperatures within the fire plume and the ceiling jet are also analysed and again, no significant changes in gas temperature are observed (Figs. 17 and 18). As concluded previously, the implementation of ventilation procedures had no significant impact on gas temperature in the configuration studied.
Fig. 16. Vertical gas temperature profile at 500 s after ignition.
2168
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
Fig. 17. Gas temperature within the fire plume (TF16 010 is the name of the thermocouple located at 3.75 m above the pool and 0.1 m from the wall).
This result can be analysed with the internal energy balance (Eq. (2)) that is reformulated as: d d T + Tcv m dt dt ˙c+Q ˙ v,in − Q ˙ v,ex − Q ˙w+Q ˙ leak =Q
mcv
(12)
Once the inlet valve is closed, the flow rate in the exhaust branch falls to almost zero (cf. Fig. 7) and the energy lost by the ventilation branches is small compared to the other terms. Closure of the exhaust branch has no effect on the amplitude of the left hand side terms of the energy equation, and hence, on the amplitude of the temperature variation (mcv dT/dt) thus explaining experimental observations. From this analysis, it can be deduced that whatever the ventilation procedure applied, the temperature variation is mainly the result of the net balance between only the fire heat release rate and the thermal losses through walls. The variation of the ventilation terms ˙ v,in − Q ˙ v,ex + Q ˙ leak ) mostly plus the air leakages (Q balanced by the mass variation term (Tcv (dm/dt)), and thus, has a negligible effect on the temperature variation term (mcv (dT/dt)). It should also be noted that although the closure of ventilation branches has a strong impact on the pressure (i.e. on the total internal energy mcv T), it has almost no significant effect on the gas temperature. In fact, this surprising statement may be clearly understood in the light of the definition of “significant effect”. Indeed, considering a closed volume of air and using the perfect gas law, a gas temperature variation of 10 ◦ C, that can be considered low in terms of fire safety, leads to pressure variations of thousands of Pascal (roughly 20 hPa), a value considered significant when dealing with ventilation networks.
8. Conclusions
Fig. 18. Gas temperature within the ceiling jet (TG5 010 is the name of the thermocouple located at 0.1 m below the ceiling and 2.5 m from the wall).
This document presents the effect of two ventilation procedures on the behaviour of a 300 kW pool fire in a ventilated enclosure. The first procedure involves closing the inlet branch at a given time after ignition. The second involves closing first the inlet branch and then the exhaust branch, the enclosure thus being entirely sealed. The experimental results show that closure of the inlet branch contributes to a reduction of the fuel mass loss rate and of the fire duration. Yet, this closure has no significant effect on the pressure within the vessel and on the gas temperature which both remain similar to the scenario without a ventilation control procedure. In addition, vessel relative pressure remains negative thus maintaining vessel confinement. It is also noted that even if the exhaust branch is left open, closure of
H. Pr´etrel, J.M. Such / Nuclear Engineering and Design 235 (2005) 2155–2169
the inlet branch significantly reduces the airflow rate at the exhaust. The procedure involving exhaust branch closure give rise to the same conclusions regarding the fuel mass loss rate, the fire duration and the gas temperature. However as regards pressure, this ventilation strategy leads to significant pressure variations within the vessel. Pressure increases significantly after the closure and features an important low-pressure peak after the fire extinction. The amplitudes of these two peaks depend on the fire heat release rate, thermal losses through walls and leakages related to the air tightness of the enclosure. Positive gauge pressure in the vessel can, therefore, be expected with this ventilation control procedure, which then may induce a loss of vessel confinement. In addition, the experimental results show that the time at which the exhaust valve is closed impacts the pressure increase, and that the later the exhaust valve is closed, the lower the pressure increase in the vessel but the larger the pressure drop at extinction. Based on the experiments performed, some important features of the two procedures investigated can be identified that could guide safety engineers in selecting suitable ventilation procedures. The procedure involving closing both the inlet and exhaust allows the fire compartment to be isolated entirely but should be applied with caution in terms of pressure variations induced. The inlet closure procedure is an alternative that features similar behaviour but does not involve significant pressure variations. As regards the mass leaving the compartment, the strategy where both inlet and exhaust are closed is not necessarily more efficient. Indeed, upon applying the inlet closure procedure, gas can leave the vessel through the exhaust but at a very low rate; when both the inlet and exhaust are closed, mass still leaves the compartment through gas leakages enhanced by the over-pressure within the vessel. Although the choice of ventilation procedure may depend on other considerations not included in this study (the mechanical resistance of the vessel or valves to
2169
pressure, the type of fuel, . . .), conclusions drawn illustrate that both procedures investigated could help significantly reducing fire hazards. Suitable procedures may be determined by bringing together these conclusions and the specific features of the installation.
Acknowledgements The authors would like to thank all the members of the IRSN Fire Experimentation Laboratory for their contributions to the FLIP test campaign, J. Torero from the University of Edinburgh for his fruitful advice and the COGEMA Company, for its involvement in the project.
References Audouin, L., Tourniaire, B., 1999. New estimation of the thermal interface height in forced-ventilation enclosure fires. In: Proceeding of the Sixth International Symposium of Fire Safety Science, Poitiers. Feng Jiang, Sui-sheng Ye, Jun Yang, 2002. Design of the exhaust system in the confinement of the HTR-10. Nucl. Eng. Des. 218, 209–214. Karlsson, B., Quintiere, J.G., 2001. Enclosure Fire Dynamics. CRC Press. Klote, J.H., 1993. Air Moving Systems and Fire Protection. NISTIR 5227 Report. National Institute of Standards and Technology. Nuclear Air Cleaning Handbook, Fire Protection, US Department of Energy, DOE HDBK 1169.2003 (Chapter 10). Pr´etrel, H., Coutreau, J., Such, J.M., 11–15 June 2001. Thermal plume of large pool fires in confined and forced ventilation enclosures. In: Proceeding of the International Conference on Engineered Fire Protection Design, San Francisco, USA. Quintiere, J.G., 1989. Fundamentals of enclosure fire “zone” models. J. Fire Prot. Eng. 1 (3), 99–119. Rehm, R.G., Forney, G.P., 1994. The pressure equation in zone-fire modelling. Fire Sci. Technol. 14 (1–2), 61–73. Svensson, S., 2002. A study of tactical patterns during fire fighting operations. Fire Saf. J. 37, 673–695. Vaari, J., Hietaniemi, J., 2000. Smoke Ventilation in Operational Fire Fighting, VTT Technical Report No. 419.