The ongoing experimental programs at IRSN on

2 1 Recent results from experimental research on fire J.-M. SUCH (IRSN) C. CASSELMAN (IRSN) L. RIGOLLET (IRSN) H. PRETREL (IRSN) L. AUDOUIN (IRSN) The ongoing experimental programs at IRSN on organic compound fires are aimed at improving knowledge on two topics that are major concerns in assessing nuclear facility safety with regard to fire risk: solvent fires, in particular when they are located close to a wall in closed and ventilated premises, and electrical cabinet fires. These programs are conducted in IRSN s test facilities, grouped together under the title of GALAXIE, at the Cadarache center. This set of facilities is unique in Europe by virtue of its diversity and the capability of its basic devices (containments, vessels and associated ventilation systems). All the experimental results help in understanding physical phenomena and in developing and qualifying IRSN s fire/ventilation computer codes. The FLIP Experimental Program The FLIP test program (liquid fires interacting with a wall), conducted by IRSN since 1997, forms part of a PIC (Common Interest Program) including IRSN and COGEMA. It is aimed at improving knowledge of a solvent pool fire s behavior in configurations representative of those of the TPH/TBP solvent storage cells in the La Hague fuel reprocessing plant. The experimental results are further used for qualifying IRSN s FLAMME_S - SIMEVENT code in configurations where the seat of the fire is close to a wall (corner fires). The pools used in the tests are square, with a free surface area varying from 0.4 to 5 m 2 and a thickness of 5 cm; a fire on a 36 cm wide, channel-shaped pool was also created for a surface area of 3.2 m 2 in order to study the effect of the pool s geometry. Figure 1 shows the configuration studied: the tank containing the fuel is situated close to one of the experimental vessel walls (400 m 3 PLUTON containment), mechanically ventilated at a rate of three renewals per hour (1200 m 3 /h). Figure 1 General view of the FLIP test system (PLUTON containment). TEST AND INSTRUMENTATION PROCEDURE Once the vessel is closed and the ventilation conditions established, the pool is ignited with a gas burner. After a flame propagation phase over the fuel pool, the flames spread out fully. They may even reach the ceiling for quite large fire powers (fire power is proportional to the surface area of the pool). The fire then behaves as a disruptive element of the containment-ventilation system, as a producer of material and heat. The instrumentation is used for on-line monitoring of the changes over time of values like the loss of mass of the fuel pool, the temperature of the gases and the walls, the gas pressure and concentration (O 2, CO 2 and CO), the velocity of the gases in the plume, the heat flows at the walls, and the gas flow rates in the ventilation ducts. Sequential samples of aerosols are also taken in order to determine their concentration and grain size distribution. 72 INSTITUT DE RADIOPROTECTION ET DE SÛRETÉ NUCLÉAIRE

2 The FLIP program has supplied elements for quantifying the main phenomena characterizing a fire in a closed, ventilated compartment. ( Safety of nuclear facilities Numerous temperature measurements made in the region of the fire, in particular in the areas of the flame and the plume, are used to subsequently display the temperature fields over time. Videography of these temperature fields provides valuable information on the behavior of gases in the compartment during the fire and on that of the flames. This imaging approach forms an aid to the qualitative understanding of the complex phenomenology of fires in a confined, ventilated environment. OBSERVATIONS ON FIRE IN A CONFINED, VENTILATED ENVIRONMENT The FLIP program has supplied elements for quantifying the main phenomena characterizing a fire in a closed, ventilated compartment: a layer of hot gases and smoke forms below the ceiling of the compartment. It can sweep through the whole compartment if, for the intensity of fire considered, the ventilation is not capable of extracting the smoke produced; the heating of the gases in the compartment where the fire is spreading causes a rapid increase in pressure (figure 2). The extinction of the fire, through consumption of the fuel or lack of oxygen, is accompanied by a sharp drop in pressure following the cooling of the gases in the compartment. These sudden pressure variations lead to studying the risk of possible damage to equipment such as the valves or filters of the ventilation system; the intensity of the fire, proportional to the surface area of fuel on fire, strongly affects the progress and consequences of the fire. The thermal and mechanical stresses generated by a fire increase with its intensity, whereas the duration of the fire decreases with the intensity, since it rapidly consumes the available oxygen. The order of magnitude of the average power of a fire in the configurations studied is approximately 1 MW for a pool of 1 m 2 ; the fire goes out when the oxygen concentration of the compartment reaches a threshold value of the order of 12%. After extinction, while the fuel is still hot and evaporating, an inflow of fresh air into the compartment can lead to a sudden re-ignition of the fire. The phases of flame propagation over the fuel pool and progressive darkening of the atmosphere of the enclosure by combustion products, a characteristic phenomenon of fires in a Figure 2 Change in pressure during a fire in a confined, ventilated compartment. SCIENTIFIC AND TECHNICAL REPORT 2002 73

Figure 3 Fuel ignition (a) and flame propagation (b) phases on a pool during the FLIP 6 test. confined environment, are illustrated in figure 3 to figure 5 for the FLIP 6 test, which involves a pool of 3.2 m 2. SOME ISSUES ARISING FROM THE INTERPRETATION OF THE RESULTS The interpretation of the results of this experimental research focused on the seat of the fire, the plume, pressure variations, and heat transfers. (a) Figure 4 Combustion phase, during which the pool is completely ablaze, FLIP 6 test. Figure 5 Darkening of the atmosphere in the enclosure during the FLIP 6 test. (b) ( Gas pressure, a disruptive physical magnitude One of the notable results of the FLIP tests is the phenomenon of large-amplitude pressure oscillations, which appears in the case of large pools (2 m 2 and more). This is interpreted as a succession of extinctions and re-ignitions preceding the final extinction of the fire through lack of oxygen. High gas temperature, significant evaporation of the fuel and an oxygen concentration close to the fire extinction threshold, appear to form the necessary conditions for these pressure oscillations. Figure 6 shows the change in pressure during a test involving a fuel pool of 3.2 m 2 and the consequences on the intake and exhaust flows of the experimental enclosure. The pressure peak following ignition of the pool causes a reversal in the direction of flow at the intake and a significant increase in gas flow in the exhaust duct. During the pressure oscillation phase, the flows in the ventilation system are also strongly disrupted, reversing several times with amplitudes sometimes greater than those observed at the time of ignition. The maximum pressure oscillation amplitude has been quantified as a function of the fire s power (pools from 2 to 5 m 2 ). The maximum pressure oscillation amplitude has been quantified as a function of the fire s power (pools from 2 to 5 m 2 ). 74 INSTITUT DE RADIOPROTECTION ET DE SÛRETÉ NUCLÉAIRE

2 Study of the plume in tests carried out by IRSN is aimed at assessing to what extent existing literature correlations are applicable to fires in a confined, ventilated compartment. ( The smoke plume, driving force of convective flows in the compartment The flow of gas in the smoke plume increases with the power of the fire and may reach much higher values than the compartment s nominal ventilation flow rate, as figure 7 shows, which corresponds to the test involving a pool of 3.2 m 2. The gas flow rate values obtained indicate that the flow of the plume acts predominantly on the convective movements inside the containment. Study of the plume in tests carried out by IRSN is aimed at assessing to what extent existing literature correlations in previously published studies, which describe the plume of fires in a free atmosphere on a small or medium scale, are applicable to fires in a confined, ventilated compartment. In this case, the fires are characterized by surrounding flame and plume conditions that are variable over time, whereas they are constant in a free atmosphere 1,2,3,4. Interpretation of the results of the FLIP tests shows 5,6 that the McCaffrey model 4, developed for a pool fire in a free atmosphere, can be extended to a fire in a confined and ventilated compartment using as reference temperature the Figure 7 3.2 m 2 pool mass flow rates in the enclosure at the half-way point of the fire Figure 6 Gas pressure in the PLUTON containment during a 3.2 m 2 pool fire and consequences for ventilation flows at the intake and exhaust. Gas pressure of the compartment Gas flow rates at the intake Gas flow rates at the exhaust References 1 - G. Heskestad, Fire Plume, SFPE Handbook of Fire Protection Engineering, NFPA Publication, 1995. 2 - E. E. Zukoski, Properties of Fire Plumes, Combustion Fundamentals of Fire, Academic Press Ltd, 1995. 3 - M. A. Delichatsios, Air Entrainment into Buoyant Jet Flames and pool fires, SFPE Handbook of Fire Protection Engineering, NFPA Publication, 1995. 4 - B. McCaffrey, Purely Buoyant Diffusion Flames : Some Experimental Results, Report NBSIR 79-1910, National Bureau of Standards, 1979. 5 - Thermal Plume of Large Pool Fires in Confined and Forced Ventilation Enclosures, Proceeding of the Third International Seminar on Fire and Explosion Hazards (10-14 April 2000, Lancashire, UK). 6 - Fire Plume from Large Pool Fires in Forced-Ventilated Enclosures, Proceeding of the International Conference on Engineered Fire Protection Design (11-15 June 2001, San Francisco, USA). Safety of nuclear facilities SCIENTIFIC AND TECHNICAL REPORT 2002 75

Figure 8 temperature of the surrounding gases (variable over time) rather than a constant value. The decrease, according to height, of the maximum temperature of the plume gases is thus relatively accurately predicted by this generalized model for fires of moderate power, i.e. for pools whose surface area is less than 2 m 2 under the experimental conditions of the FLIP tests (figure 8). The analysis shows, however, the limitation of the model in the case of high-power fires (surface area of the pool greater than 2 m 2 ) in which the interaction of the plume with the ceiling significantly influences the behavior of the plume (figure 9). Other correlations of the literature have to be analyzed, and consideration must be given to adapting a generalized correlation of this type to the simplified approach of zonal computer Temperature in the thermal plume for FLIP tests 1, 2, 7 and 5. codes, or even to developing a specific correlation with the plume in a confined, ventilated environment. Temperature field in the compartment The instantaneous thermal field of the gases in the compartment, shown in figure 10, can be used to locate the flame envelope at an instant when the atmosphere of the compartment is completely obscured by soot. Figure 11 (page 77) shows a succession of extinctions and re-ignitions of the pool. QUALIFYING THE FLAMME_S - SIMEVENT CODE The so-called zonal computer code, FLAMME_S, is used to simulate a fire in an industrial facility to determine its thermodynamic (temperatures and pressure) and chemical (concentration of species) consequences. It can function in coupled mode with the SIMEVENT computer code, which simulates the air behavior of a ventilation system (transport of gases and soot, clogging of the very high-efficiency filters, etc.). The FLIP tests extend the FLAMME_S field of qualification to fires located close to a wall. The code has been applied to the FLIP test conditions, in non-coupled mode (FLAMME_S on its own with a simple representation of the ventilation system by pressure losses and fixed boundary conditions) and in coupled mode (full detailed modeling of the experimental ventilation system by SIMEVENT). The agreement between the cal- Figure 9 Temperature in the thermal plume for FLIP tests 6 and 8. Figure 10 Thermal field in the PLUTON containment during the FLIP 6 test (3.2 m 2 pool). 76 INSTITUT DE RADIOPROTECTION ET DE SÛRETÉ NUCLÉAIRE

2 An uncontrolled fire in a nuclear plant could lead to a deterioration in the level of safety and threaten the containment of radioactive materials. ( culated and the experimental results is clearly improved for very low-power fires when the ventilation system is modeled in detail (calculation in coupled mode). On the other hand, whether the calculation is performed in coupled mode or not, the pressure and temperature of the gases in the compartment are overestimated for large fires (power greater than 500 kw), which is explained by poor estimation of heat exchanges between the gases and the walls of the compartment (figure 12 and figure 13, page 78). This is because the thermal stratification of the gases, especially marked in the case of high-power fires, cannot be described by a two-zone simplified model (upper zone of hot gases and smoke, lower zone of cool gases), which assumes that each zone is at a homogeneous temperature. Thus, not making allowance for local values (variable convective transfer coefficient, temperature distribution below the ceiling) may lead to poor estimation of global values (average temperature of the gases in the compartment and their pressure). This finding highlights the limitations of zonal codes like FLAMME_S in simulating a fire and demonstrates the utility of multidimensional codes (field codes) which calculate local values. The research program on electrical cabinet fires CONTEXT OF THE RESEARCH A nuclear plant may contain a large number of electrical cabinets - (approximately two thousand in a pressurized water reactor). This kind of equipment forms the source of one quarter of fire starts due to equipment faults. An uncontrolled fire in a nuclear plant could lead to Safety of nuclear facilities Figure 11 View of the phenomenon of extinction and re-ignition during a 1.5 m 2 pool fire (t adim = instant/duration of fire). SCIENTIFIC AND TECHNICAL REPORT 2002 77

a deterioration in the level of safety and threaten the containment of radioactive materials. By way of illustration, the preliminary studies conducted as part of the probabilistic safety analysis relating to 900 MWe PWRs, associated with fire, in particular the study of a fire occurring in the control room, have shown that combustion damage to electrical cabinets or consoles could make a major contribution to the probability of damage to the reactor core. The information needed for studying the consequences of the fire scenarios caused relates to: the heat release rate of a fire in an electrical cabinet versus time; the damage of an electrical cabinet submitted to an external heat flux; the mode of damage of the electrical cabinets adjacent to the cabinet set on fire; the temperature evolution inside the cabinet as a function of external temperature; identification of the failure modes linked to the thermal damage. Data on fires in electrical cabinets is scarce in the open literature. So, it has proved necessary to improve the knowledge on this type of fire with a research program for developing then qualifying a simplified model implemented in the FLAMME_S calculation code. Figure 12 Gas pressure of the compartment calculated with the FLAMME_S - SIMEVENT code (coupled calculation and non-coupled calculation), compared with the experimental values on the left for a high-power fire and on the right for a low-power fire. Figure 13 Average temperatures of the compartment calculated with the FLAMME_S - SIMEVENT code (coupled calculation and non-coupled calculation), compared with the experimental temperatures on the left for a high power fire and on the right for a low power fire. 78 INSTITUT DE RADIOPROTECTION ET DE SÛRETÉ NUCLÉAIRE

2Safety of nuclear facilities In this research program, special attention is paid to determining fire power. ( SETTING UP THE RESEARCH PROGRAM This program started in 1998 with a bibliographical study that especially focused on the tests carried out by VTT (Finland) as they use real electrical cabinets. The simulations created with the FLAMME_S software for recalculating the VTT tests highlighted the difficulty of describing the fire in an electrical cabinet (the power released over time, the plume characteristics, etc.). This is because the phenomena are numerous and complex in a fire in an electrical cabinet; they include in particular: the extreme confinement of the flame; the scarcity of the oxygen needed for combustion; the pyrolysis of components under the effect of intense heat fluxes inside the electrical cabinet; gas flows (air inlet and gas outlet) via the electrical cabinet ventilation openings and in the electrical cabinet itself. The bibliographical study identified the following main parameters: the area of the electrical cabinet s ventilation openings, both at the intake, at the bottom, and the exhaust, at the top; the filling of the electrical cabinet; the pressure drop of the electrical cabinet; the location of the ignition point; the mass of fuel. The complexity and diversity of the types of electrical cabinets identified in a real facility led to adopting a research procedure with a large preliminary analytical part. The object of these analytical tests, christened CARMELA 1, is to quantify the effect of the parameters considered predominant on the evolution of the power of an electrical cabinet fire. This analytical approach basically involves experimentally quantifying how, for a given fire load density, the characteristics of the electrical cabinet (its filling, the areas of the ventilation openings, etc.) modify the heat released by the fire. The first three parameters, namely the area of the ventilation openings, the filling of the cabinet and the arrangement of components, modify the air flow in the cabinet and therefore the inflow of oxygen for combustion. The location of the ignition point influences flame propagation. In this research program, special attention is paid to determining fire power, a fundamental physical value for characterizing the fire and assessing its consequences for its environment in a real facility (other electrical cabinets, cable runs, neighboring premises, associated ventilation system, etc.). The plume of hot gases originating from the electrical cabinet is also a significant point of the study since, in a real facility, this flow may cause thermal stresses on equipment (like cable runs) and lead to their being damaged or even to their malfunctioning. Global tests (christened CARMELO) using real electrical cabinets will only be carried out later, in 2003. In parallel with this, the characterization of the component materials of a real electrical cabinet (electrical components, printed circuit boards, ducting, cable and wiring, etc.) is planned on a small scale, based on standardized tests (cone calorimeter, TEWARSON apparatus). THE CARMELA TESTS The experimental device The electrical cabinet is simulated by a steel box with dimensions comparable to those of a real electrical cabinet containing a fuel. The latter, a vertically positioned Plexiglas plate, represents a fire load rather than electrical and electronic components fitted in a real electrical cabinet. The cabinet is equipped with ventilation openings, located in the bottom of the door and on the top of the cabinet. The filling of the cabinet is simulated by inert steel blocks and metal plates located at various 1 - CARMELA : acronym for Combustion of an analytical electrical cabinet. SCIENTIFIC AND TECHNICAL REPORT 2002 79

Figure 14 View of the CARMELA device during a test. The first phase of thethe first phase of the CARMELA program includes 15 tests carried out in 2000 and 2001.CARMELA program includes 15 tests carried out in 2000 and 2001. ( heights inside the cabinet. These structures enable pressure drop in the electrical cabinet to be varied. The ignition point is located at the bottom or halfway up the combustible plate. The device is placed under a large exhaust hood whose on-line analyses before filtering are used to determine the power of the fire. Mass loss rate of fuel, temperature, pressure and gas concentration are measured in the cabinet, as well as temperature and velocity of the plume. The first phase of the CARMELA program includes fifteen tests carried out in 2000 and 2001. Figure 14 shows a view of the whole device during a test. Flame can be seen coming out of the upper ventilation orifice, on the top of the electrical cabinet model, as well as the smoke plume flowing into the exhaust hood. Figure 15 Typical progress of fire power as a function of time. Interpretation The interpretation of experimental results is in progress. By way of illustration, figure 15 shows the typical evolution of fire power versus time, in dimensionless values: the power increases during the flame spreading phase on the combustible plate, reaches a maximum, then falls abruptly when the plate collapses into the bottom of the cabinet on softening under the effect of temperature. The object of the empirical model now being prepared is to assess the evolution of fire power as a function of the electrical cabinet s characteristics. It is therefore necessary to describe the growth phase of fire power, characteristic of flame spreading in the electrical cabinet. A simple model normally used for complex material fires (chairs, cars, etc.) has been applied to the 80 INSTITUT DE RADIOPROTECTION ET DE SÛRETÉ NUCLÉAIRE

2 experimental results of the CARMELA tests. This model express fire power. Q (in kw) as a function of time squared :. Q = α. t 2 where α is the growth factor (in kw.s -2 ) and t, time (in secs.). Examination of the power obtained experimentally between the beginning of the fire (ignition of the gas burner) and the instant corresponding to the maximum value of the power (t = t 2 ) reveals two phases: a very slow fire propagation phase on the fuel plate (t t 1 ) followed by a faster fire propagation phase (t 1 t t 2 ). Instant t 1 of the transition between the two phases corresponds to an abrupt change in the velocity of flame spreading. The power evaluated by the model is best adjusted with the experimental result for each of these two phases acting on the growth factor α:. Q = α 1. t 2 for t t 1. Q = α 2. (t - t 1 ) 2 + α 1. t 1 2 for t 1 t t 2 The application of this model is illustrated in figure 16. The current analysis of the CARMELA tests is aimed at correlating the growth factors α with the characteristics of the electrical cabinet. As an example, the variation in the growth factor α 2 (second phase of fire propagation on the plate) as a function of the area of the ventilation openings in the door is shown in figure 17. This factor evolves linearly as a function of this area. Figure 18 illustrates the influence of the crosssection of the openings in the door on the peak fire power for a given cross-section of opening in the top of the electrical cabinet. The gas temperature fields inside the cabinet (figure 19, page 82) have especially shown that the temperature apparently becomes homogeneous and equal to 600 C as soon as the propagation phase begins (figure 16). A second series of CARMELA analytical tests will be carried out in 2003. Its content is defined based on the interpretation of this first series of tests and on the empirical cabinet fire model currently being developed. Eventually, this model will be qualified and if necessary adjusted based on CARMELO tests, which bring into play real electrical cabinets. Figure 16 Representation of a fire power build-up in an electrical cabinet. Figure 17 Evolution of the growth factor α 2 as a function of the cross-section of the ventilation openings in the door. Figure 18 Influence of the cross-section of the ventilation openings in the door of the electrical cabinet on the maximum fire power (as a dimensionless value) for two values of cross-section of opening in the top. Safety of nuclear facilities SCIENTIFIC AND TECHNICAL REPORT 2002 81

1 Figure 19 Gas temperature changes in the electrical cabinet. Ignition of the plate Flame propagation on the vertical plate Start of the propagation phase Conclusion The research work conducted at IRSN in the field of fire has already led to substantial improvements in understanding the development of a fire and its consequences for maintaining the containment of radioactive substances present in nuclear facilities. Many gains have been made both on the experimental level and on that of developing and qualifying the computer codes used for safety assessments. Nevertheless, experimental and theoretical studies are still needed to improve our knowledge of some phenomena and their modeling, as well as to expand the sphere of application of the computer codes. Thus, the program on electrical cabinet fires will continue in 2002 and 2003 in order to provide answers to safety concerns in this matter in 2003. The FLIP program will end in 2002, during which year IRSN will begin an experimental research program on fires in a multi-compartment configuration in a special full-scale experimental device christened DIVA (fire, ventilation and airborne contamination device). 82 INSTITUT DE RADIOPROTECTION ET DE SÛRETÉ NUCLÉAIRE