Fire scenarios modelling for the safe design of a passenger rail carriage

Transcription

1 Fire scenarios modelling for the safe design of a passenger rail carriage Andreini A., Da Soghe R., Facchini B., Giusti A. 1, L. Caruso ; G. Luconi 2, (2)- Troiano D. 3, 1, Engineering Department Sergio Stecco University of Florence 2, Trenitalia S.p.A,Technical Direction, Florence, Italy; 3, Italcertifer SCpA Technical Area and certification Florence Italy Effective fire safety design of rail infrastructure is underpinned by application of realistic design fires representing edible train fire incidents. One of many objectives for passenger train designers is to minimise both the probability of a large interior fire occurring and the maximum fire size and duration. However, current understanding of fire development in passenger train interiors is limited and existing methods for design fire estimation applied to trains are based on ude assumptions and, therefore, appropriate design fires for passenger trains are uncertain. In this paper a summary of the results of a research programme sponsored by the Italian national railways will be presented. The programme involves university as well as experimental laboratories and fire suppression systems manufacturing companies. The project is focused on the development of numerical tools for the reliable fire scenarios simulation and on the definitions of standards for the fire safety design of rail carriage. The programme experimental campaign has considered full scale and maquette tests and have provided valuable data in terms of exhaust gases concentration, temperature, heat flux measurements. The experimental data have been used to validate the advanced design tools used by Trenitalia (CFD codes and correlative formulations). One of the main goal of the research is to define a reliable procedure for the design of a fire extinguish system for railway passenger train by the use of the code FDS (Fire Dynamic Simulator) developed by the NIST (National Institute of Standard and Technologies). In order to reach the programme objectives, several aspects have been deeply analysed. More in details attention has been paid on the combustion and pyrolysis modelling and a best practice for the solid fuel s burning properties evaluation has been defined. CFD has proved to be an excellent tool for this kind of analysis as it provides a reliable evaluation of the interesting physical quantities that are analyzed to develop a fire safety design of rail carriage. The methodologies defined are a novel and potentially industrially useful contribution to rail infrastructure design. 1 INTRODUCTION A itical issue which is becoming ineasingly important in the design of passenger trains is the estimate of carriage fire resistance aimed at preserving passenger safety. As in other scientific fields, two basically different approaches can be used to study a fire scenario: experimental investigations and numerical simulations. It goes without saying that, in an industrial context, it is fundamental to develop proper tools which allow design engineers to choose the best solution in

2 reasonable time without waste of resources and, from this point of view, the numerical approach is undoubtedly the most appropriate. However, the available numerical tools are not able to give reliable predictions because of the limited knowledge about fire development characteristics and material pyrolysis process. Therefore, especially when simplified and less time consuming numerical models are used, it is indispensable to carry out experimental investigations to support numerical simulations in both defining material properties and validating numerical procedures. The pyrolysis has been identified as the most important physical process that regulates the behaviour of a fire scenario. So this research is mainly focused on defining an appropriate procedure aimed at determining the pyrolysis parameters starting from cone calorimeter tests, the experimental test chosen to measure solid material properties. A full-scale test has also been planned to validate the numerical procedure. 2 FDS4 PYROLYSIS MODELS As also desibed by White [1], in FDS4 there are two different approaches to model pyrolysis. The first one, which in the following will be called direct method, consists in assigning a specific HRR (Heat Release Rate) per unit area to each material that characterizes the fire scenario; therefore, since the production rate of gaseous fuel per unit area is controlled by the combustion process, the mass flow rate of gaseous fuel is independent of thermal feedback to the surface. The inception of the pyrolysis process is controlled by a parameter called ignition temperature: when the material achieves this temperature the gaseous fuel release begins. On the contrary, the second approach, which will be referred to as the indirect method, is based on the heat of vaporization and it accounts for thermal feedback to the solid surface; in this case the code allows the user to model the material either as a thermoplastic or a charring fuel. An ignition temperature has to be specified together with other specific parameters (see Section 3.2) which can be obtained from a cone calorimeter test by applying particular post-processing procedures as will be desibed below. 3 CONE CALORIMETER TEST AND POST-PROCESSING PROCEDURE When a solid material is modeled using the direct approach, the HRR curve (i.e. a curve which desibes the heat generated by the combustion process as a function of time) and the ignition temperature T ig are sufficient to set up a FDS4 simulation. However, when the indirect method is considered, besides T ig, the following material properties are also required: the heat of vaporization, the itical mass flux, the maximum burning rate and the thermal inertia kρc (thermally-thick material) or the product ρcδ (thermally-thin material). All these parameters can be obtained starting from cone calorimeter test data by means of a particular postprocessing procedure. In this work the post-processing procedure desibed by Boon [2] (s ee Section 3.2) has been used at first. Then a further iterative step has been introduced in order to obtain a better fitting of experimental data (see Section 4). 3.1 CONE CALORIMETER TEST Cone calorimeter is, at present, the most commonly used tool for bench-scale heat release rate measurement. Besides the heat release rate, it is also possible to measure other parameters such as effective heat of combustion, mass loss rate, ignitability, smoke, soot and toxic gases production [2]. During a cone calorimeter test (ISO 5660), a specimen is exposed to a constant level of exposure heat flux from a conical heater. Volatile gases, which are released from the specimen during the test, are ignited by an electric spark ignitor. The experimental setup usually includes a gas analyser which allows measurements of CO 2, CO, O 2 and other toxic gas concentrations. The mass loss of the specimen is also recorded during the experiment as well as the soot characteristics. The HRR is calculated using the oxygen consumption method whilst for the other parameters a post processing procedure is required. 3.2 POST-PROCESSING PROCEDURE As desibed by Boon [2], in order to derive the material properties from the cone calorimeter test data, ignition data measured at a minimum of three exposure heat fluxes are required. Furthermore tests at each exposure heat flux should be replicated in order to have a statistically representative set of data. The ignition temperature can be determined following the subsequent procedure [2,4] based on a thermal balance equation applied to the specimen. First of all the average time to ignition t ig for each exposure heat flux is calculated; then average times are correlated by plotting (1/ t ig ) n against exposure heat fluxes. The value of n can vary between and 1 and it has to be chosen in such a way as to obtain the highest correlation coefficient R 2. After that one has to determine the x-intercept from the best-fit line through the data. The x-intercept is taken as the itical heat flux for ignition q which is an estimate of the minimum heat

3 flux for ignition, i.e. the heat flux below which ignition under practical conditions cannot occur. Finally the ignition temperature T ig can be calculated by iteratively solving the following equation: q h ( T T ) ( T c ig 4 ig T 4 ) where h c is the convection heat transfer coefficient in kw/(m 2 K), ε is the surface emissivity at ignition, σ is the Stefan-Boltzmann s constant and T is the ambient temperature. When the solid material is modeled using the indirect method, the following parameters are also needed. First of all the material has to be classified either as thermally thin or thermally thick. The exponential coefficient n, previously determined, gives a iterion to operate such a choice. In particular, if n is closer to the material can be classified as thermally-thick and in this case the thermal inertia kρc has to be determined for the simulation. The procedure to determine the thermal inertia is outlined below [2,4]. First of all the total heat transfer coefficient h tot is computed: h tot q ( T T ) ig Once again, ignition times have to be correlated but this time using n = and including data point q on the x-axis. After that the slope of a straight line drawn through the data point q on the x-axis and data point for the highest heat flux is to be determined. The apparent thermal inertia kρc in kw 2 s/(m 4 K 2 ) can then be calculated by means of the following expression where slope stands for the slope of the straight line previously drawn: kc h 2 tot slope q On the other hand, when the exponential parameter n is closer to 1, the material is identified as thermallythin. For such a type of material, FDS4 requires the product ρcδ to be specified. The procedure to compute ρcδ requires the subsequent steps [5]: correlate the ignition times using n = 1, determine the slope of a linear line fit through the data, compute the product of ρcδ (in kj/m 2 K) by means of: c 1 slope( T ig T ) The heat of vaporization ΔH v is determined following the procedure of Quintiere [3]: peak heat release rates from cone calorimeter tests are plotted against the exposure heat flux levels. Then the slope of a linear q peak fit through the data has to be determined (this slope is equal to effective heat of combustion ΔH c,eff divided by heat of vaporization ΔH v ). Starting from the effective heat of combustion ΔH c,eff obtained from the cone calorimeter tests (an average value is used), the heat of vaporization can be computed using: H v H c, eff slope The itical mass flux m is a property that quantifies a itical condition for ignition. FDS4 requires both this parameter and the ignition temperature T ig to be specified (for the indirect modeling approach) in order to allow the code to compute pre-exponential factor A and activation energy E of the Arrhenius equation used to calculate gaseous fuel mass flow rate (the following equation refers to a thermoplastic material): m Ae E / RT In this way the fuel burns at the itical mass flux rate when its surface temperature reaches the ignition temperature. The itical mass flux has been computed using the mass loss rate peak value recorded in the

4 cone calorimeter test. This is a conservative approximation as it is expected the true itical mass flux to be lower than the value used. 4 CONE CALORIMETER DATA POST-PROCESSING RESULTS Six different materials, which are considered itical for the full-scale experiment, have been chosen to be tested using the cone calorimeter in such a way as to have an accurate characterization of their properties. Experiments have been carried out at the CSI S.p.A. laboratory in Bollate, Milan - Italy (Figure 1 shows some pictures of experimental tests whilst in Figure 2 the CSI S.p.A. experimental facility is depicted). First of all the material properties have been computed following the procedure desibed in Section 3. After that, in order to assess the reliability of the post-processing procedure, a numerical simulation of the cone calorimeter test has been performed. Figure 3 shows the numerical array used to reproduce the cone calorimeter tests whilst Table 1 summarizes the most important parameters derived from the cone calorimeter test. The parameters computed in the post-processing procedure are given as inputs (the material is modeled using the indirect approach) to the FDS4 cone calorimeter simulation; the main result of the numerical simulation is the material HRRPUA (Heat Release Rate Per Unit Area) curve which can be directly compared with the experimental one. Parameter Seat (cover + foam) Floor (rubber) Curtain Headrest (cover) Wall Ceiling Heat of reaction [kj/kg] Heat of vaporization [kj/kg] Density [kg/m 3 ] Ignition temperature [ C] Heat capacity [kj/(kgk)] Mass lost rate peak [g/s] Table 1: FDS inputs derived from cone calorimeter tests Figure 1: Cone calorimeter experimental tests

5 Figure 2: CSI S.p.A. experimental facility Figure 3: Numerical reproduction of the cone calorimeter Figure 4 illustrates such comparison. The graphs reported in this figure show that experimental and numerical data are very different from each other; it is clear that the cone calorimeter post-processing procedure previously desibed is not able to give an accurate estimate of the parameters that characterize the pyrolysis process and a further post-processing step is required. As also pointed out in recent publications available in literature [2] the low reliability of the previous procedure may be due to the presence of fire retardant additives into the analyzed materials. In order to improve the post-processing procedure, a numerical sensitivity analysis to the various parameters that characterize the indirect pyrolysis model was performed showing that the main parameters that affect the HRRPUA are the heat of vaporization and heat capacity of the solid fuel. In particular, the heat of vaporization mainly affects the maximum value of the HRRPUA curve whilst the heat capacity controls the ignition time.

6 Figure 4: Comparison between experimental and numerical HRRPUA In light of these observations, a numerical procedure able to set the best values of the heat of reaction and the heat capacity can be defined: the numerical simulation of the cone calorimeter test is repeated varying such parameters in order to obtain a HRRPUA curve much closer to the experimental data. Figure 5 shows how the setting up of the two parameters can lead to a correct fitting of experimental measurements. As regards the heat of vaporization, the best fitting was generally reached using a final value about 50% less than the initial value. Furthermore, it is important to note that the specimen response is also affected by the exposure heat flux since the pyrolysis process is strictly connected to the temperature reached by the specimen itself during the cone calorimeter tests. However the exposure heat flux is usually a result of a fire scenario numerical simulation and it varies during the simulation making the setting up of the parameters a very time consuming iterative process. For this reason the exposure heat flux dependence is simply considered by assigning a set of parameters that allow the fitting of the experimental HRRPUA curve obtained using the exposure heat flux which is expected to be much closer to the real one.

7 Figure 5: HRRPUA after heat of vaporization and heat capacity manipulation 5 COMPARISON BETWEEN DIRECT AND INDIRECT PYROLYSIS MODELING APPROACH In order to overcome the uncertainty on material properties, the code FDS4 also allows the user to model the pyrolysis process using a direct approach. The two modelling strategies, direct and indirect, are substantially different from each other (as said above, the main difference concerns the thermal feedback on the material which is not considered in the direct approach) and, as will be shown below, they lead to different results. The influence of the two modelling strategies on the numerical solution has been assessed using the fire scenario showed in Figure 6. It consists in two facing seats and a burner on one of them which allows the fire to be ignited. The burner is on during all the simulation (600 seconds). It is important to note that in the direct approach the production rate of gaseous fuel is assigned in such a way that the cone calorimeter simulation gives the same HRRPUA of the indirect strategy. The heat release rate obtained with the two different approaches is shown in Figure 7. In the direct approach (blue line) the left seat begins to furnish the imposed gaseous fuel mass flow rate after about 15 seconds when the ignition temperature is reached. After about 120 seconds from the beginning of the simulation the left seat results fully burned and the heat release rate becomes equal to the burner HRR (300 kw). In this case the fire does not propagate to the right seat since this one does not reach the ignition temperature.

8 Figure 6: Fire scenario Figure 7: Heat Release Rate of the two seats fire scenario On the contrary, when the indirect pyrolysis model is used (red line), the fire also propagates to the right seat after about 220 seconds as shown by the second HRR peak in Figure 7 and by the temperature field in Figure 8. This simple fire scenario shows how different models may lead to substantially different results. Since the aspects related to thermal feedback are considered fundamental in a typical passenger train fire scenario, the indirect approach has been chosen for the subsequent steps of this research activity. Figure 8: Temperature field at 220 s predicted by the indirect approach 6 FDS4 SIMULATION OF A PASSENGER TRAIN FIRE SCENARIO The pyrolysis parameters obtained in the post processing procedure have been used to set up a passenger train fire scenario simulation. The simulation attempts to reproduce the fire scenario of the full scale experimental test which is planned to definitively validate the numerical procedure. Figure 9 shows the simple geometry used to reproduce the carriage; in Figure 10 a detail of the carriage internal arrangement is reported.

9 Figure 9: Numerical model of the carriage Figure 10: Carriage internal arrangement details 7 CONCLUSIONS In this work a reliable procedure for the evaluation of the pyrolysis parameters required by FDS4 has been selected and tested. The results are encouraging as the cone calorimeter experiments have been successfully replicated by the CFD. As stressed above, once the pyrolysis parameters have been provided to FDS4 the code internally compute the pre-exponential factor A and the activation energy E of the Arrhenius equation used to calculate gaseous fuel mass flow rate. These two parameters can be easily acquired from a FDS4 calculation and so the existing material property libraries can be used in more update CFD solver such as FDS5. The simulation of a real passenger carriage fire scenario is now ongoing and the results will be presented at the congress.

10 8 NOMENCLATURE Latin symbols A Pre-exponential factor [m/s] c Specific heat [kj/(kg K)] E Activation energy [kj/kmol] h c Convection heat transfer coefficient [kw/(m 2 K)] h tot Total heat transfer coefficient [kw/(m 2 K)] HRR Heat Release Rate [kw] HRRPUA Heat Release Rate Per Unit Area [kw/m 2 ] k Thermal conductivity [kw/(m K)] m Mass loss rate [kg/(s m 2 )] m Critical mass flux [kg/(s m 2 )] q Critical heat flux [kw/m 2 ] q Peak heat release rate [kw/m 2 ] peak T Temperature [K] T ig Ignition temperature [K] T Reference temperature [K] Greek symbols ρ Density [kg/m 3 ] δ Thickness [m] σ Stefan-Boltzmann s constant (5.67e-11) [kw/m 2 K 4 ] ε Surface emissivity at ignition [-] ΔH c,eff Effective heat of combustion [kj/kg] ΔH v Heat of vaporization [kj/kg] 9 BIBLIOGRAPHY [1] White N., 2009, Fire Development in Passenger Trains. Master Thesis, Centre for Environment Safety and Risk Engineering, Victoria University, Australia [2] Boon H. C., 2005, Numerical Simulation of a Metro Train Fire. Master Thesis, Department of Civil Engineering, University of Canterbury, New Zealand [3] Quintiere, J. G., 1993, A Simulation Model for Fire Growth on Materials Subject to a Room-corner Test. Fire Safety Journal, 20, pp [4] Janssens, M. L., and Grenier, A. T. An Improved Method for Analyzing Ignition Data of Composites. Proceedings of 23rd International Conference on Fire Safety 1997, Millbrae, California, pp [5] Mikkola, E., and Wichman, I. S On the ignition of Combustibles Materials. Fire and Materials, 14, pp. 87 to 96.