An Experimental Analysis of the Warehouse Pallet Impact in Terms of Fire Loading

Transcription

1 An Experimental Analysis of the Warehouse Pallet Impact in Terms of Fire Loading Truchot B.*, Leroy G., Fouillen F., Rahman S. INERIS, Parc ALATA, Verneuil en Halatte (France) *Corresponding author ABSTRACT The fire behaviour of goods pallets is a crucial element for characterising the warehouse fire hazard. Characteristics of the fire depend not only of the nature of the goods but of the nature and quantity of the packaging, commonly composed by combustibles as cardboard or plastic too. Consequently fire tests were managed for real pallets, including packaging, for different nature of goods. One of the main issue in such test consists in being representative of the warehouse storage configuration because of interaction between pallets during fire propagation. A test protocol is proposed here to take account of fire in the neighborhood, at the ground level, by introducing a previously calibrated wood pallet fire. Results are given for three different food pallets. The first is composed by fresh salad with the main interest to include a large amount of plastic punnets. The second is composed by frozen meat to show the influence of the frozen water included in the foods on the fire development. The last one is a frozen butter pallet to highlight the specific behaviour of a liquifiable product. Finally, while several calorimetric methods are available, none can consider differently the sample matter than the wood aggression, to determine their relative specific contribution to the global heat release rate. KEYWORDS: Fire tests, Pallets characterization, warehouse. INTRODUCTION Because of the large variety of commodities that can be stored in warehouses, designing fire protection measures or predicting consequences in case of fire is nowadays a great challenge. Whatever the model that can be used for the evaluation of those consequences, one of the key issues is still the heat release rate (HRR) of individual elements. In the warehouse configuration, it consists in the commodity pallets that are stored including goods but also packaging. The objective of this paper is to present the methodology developed for determining this HRR and then describing the results for the different types of commodities. Such data can then be used in different types of models, as 3D computational fluids dynamics codes as FDS [1] or global approaches as Flumilog [2], to evaluate the global fire characteristics with taking into account the warehouse specificities. One of the main difficulties is, in such an evaluation, to consider the pallet surrounding with fire interaction. While the objective is not here to measure the global fire propagation, but to evaluate the HRR of a given sample, a general measurement of the whole energy is not enough. A specific protocol is then proposed to introduce a calibrated fire in the surrounding to generate an external thermal flux representative of the one than can be received in case of warehouse fire. This external aggression is generated by different types of wood pallets arrangements on three of the four faces of the sample pallet. The global HRR can then be measured using classical calorimetric methods and, assuming that the wood pallet arrangement fire is reproducible, the specific HRR of the sample to be evaluated. Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8), pp Edited by Chao J., Liu N. A., Molkov V., Sunderland P., Tamanini F. and Torero J. Published by USTC Press ISBN: DOI: /c.sklfs.8thISFEH

2 Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8) Then, after having described the burning test protocol and the experimental facilities that were used, results are detailled for three goods pallets with their specificities. The first is a salad pallet, having in mind that burning a salad is quite difficult, the main fire load is then composed by the packaging made of plastic punnets and cardboards. The second pallet consists in cardboard of frozen meat. Considering meat is composed by a large amount of fat, the fire load is then given by the food, that also contains a large quantity of frozen water. The last test proposed here consists in burning a frozen butter pallet. What is interesting in such a test is the ability of the butter to generate a pool with a specific risk of ignition. EXPERIMENTAL PROTOCOL As detailed in the introduction, the aim of the fire tests achieved at the pallet scale was to evaluate their specific behaviour in case of fire. To let such tests possible with several types of goods, a limited number of pallets has to be used, this induced specifications in terms of fire protocol. This protocol is deeply presented in a French document [1] and is summarised hereafter. Before going any further in the protocol description, it is important to highlight that the final objective is to determine the HRR and total released energy for the pallet. It is then important that such tests has to be achieved in large scale calorimeter. The protocol described in [1] and shortly in the present paper is focussed on the INERIS facilities. INERIS fire gallery Experimental device that was used for the experimental campaign is the INERIS fire gallery. This fire gallery was described in some previous papers [4] but relevant details are given hereafter. This gallery is 50 m long with a 3 m width and 1.8 m height section that corresponds to a two lanes tunnel at third scale. This fire gallery is equipped with a fan that can be control to manage the air flux into the tunnel. Photography of INERIS fire facilities is presented on Fig. 1. Figure 1. INERIS fire facilities. One of the main interests of this installation consists in the smoke treatment system installed downstream. This system, design as for garbage furnace emissions treatment, enables to capture not only the toxic products as carbon oxides or acid gases but chronical toxic compounds, as dioxin or PAH (Polycyclic Aromatic Hydrocarbons) too [4]. 268

3 Part II Fire Fire tests sequence Considering that the objective of each tests series consists in determining the fire behaviour in storage configuration, it is important to evaluate the impact of the neighbourhood in terms of fire development. To reach such an objective, the fire in the surrounding, which influences the fire development of the sample, is of primary importance and has to be representative of the HRR of the sample itself. Then, a full test series is composed by three individual experiments, Fig. 2: free burning of a sample, at pallet scale, characterisation of a calibrated wood fire to represent the surrounding, fire test of the sample with calibrated fires in the surrounding. The first test consists in the HRR evaluation of the sample itself with no other external aggression than the ignition. For that case, ignition is made thanks to two 100 kw propane burners directed to the bottom of the sample, commonly the wood pallet that support goods. The application duration for burners is 5 minutes to let then the fire developed itself. It is important to note that this ignition source represents a quite important source compared to those that can occur in warehouse. The second test consists in burning a given number of wood pallets, the number and arrangement are designed based on the sample fire behaviour. These pallets will represent the external aggression, i.e. the surrounding fire, for the last experiment. Then its characteristics have to be determined with a quite good accuracy. To reach such an objective, pallets are arranged around an instrumented inert cube located in the middle of the arrangement. The third and last test is identical to the second one with the sample in the middle of wood pallets instead of instrumented cube. Measurements and results managment Figure 2. Three tests arrangement scheme. As described previously, the objective of each series of tests is to determine the HRR of a given pallet. Consequently the main measurement apparatus consists in a gas analysis to let calorimetry possible. To ensure a correct evaluation of gas concentration downstream the fire, CO, CO 2 and O 2 probes are located in two different positions in the smoke extraction duct. A comparison between the two measurement points is achieved to demonstrate the correct mixture homogeneity. Additional gas measurements are achieved thanks to an FTIR (Fourier Transform Infra-Red) spectrometer. The objective is not only to provide a complementary CO 2 concentration evaluation but also to give a representation of smoke composition, including acid gases mainly. These concentration probes are obviously coupled with flow rate measurement. Well known equation for calorimetry is then used based on both oxygen consumption (OC) and Carbon Dioxide Generation (CDG) methods. For OC, the instantaneous HRR, P(t), is then evaluated thanks to [7]: 269

4 Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8) 0 XO X 2 O W 2 O2 0 0 Pt () = αo m ( 1 ) 2 a XHO X 2 CO. (1) 2 1 XO W 2 a 0 In this relation, X G represents the molar fraction of G species in smoke, X, the molar fraction of G species G in the ambient, W G is the molar weight (kg/mol) of species G and is the incident air mass flow rate (kg/s). Then, the same HRR is also computed based on CO and CO 2 concentration measurements [8]: ( ) = αco CO + αco CO Pt m m, (2) 2 2 where ṁ CO 2 is the CO 2 mass flow rate and is ṁ CO the CO mass flow rate. In those two relation, α x is the corresponding calorimetric coefficient. The standard values were used for OC and CDG coefficient, i.e MJ/kg for α O 2, 13.3 MJ/kg for α and 11.1 MJ/kg for α CO. Of course, while the OC coefficient is few impacted by the nature of the combustible, CDG coefficients are. Consequently, in case of major differences between the HRR computed with the two methods, a detailed analysis of these coefficients value were achieved. In order to have a local evaluation of the fire behaviour, video camera and (0/100 kw/m²) cooled flux meters are distributed around the sample at different heights. CO 2 WOOD PALLETS ARRANGMENTS CALIBRATION Before testing a large number of goods pallets, several tests were made in order to characterize the behavior of a wood pallet series. Tests were achieved for 1 to 4 pallets on three of the four sides of an instrumented noncombustible cube composed with Siporex. For the 4 pallets test, two different pallets arrangements were done. It is important to note that, prior to those tests, the wood pallets were dried during several days in a regulated stove. Thanks to the gas analyses previously described, the HRR was determined for those different configurations. Characteristics and main results in terms of HRR and energy release are detailed in Table 1 for all arrangements. These configurations are only preliminary tests to provide data for the pallet fire tests series. Case number Wood mass for each face / Total mass (kg) Table 1. Wood pallets arrangement characteristics. Pallets arrangement 270 Peak HRR (kw) Total energy (MJ) Heat of combustion (MJ/kg) CAR-1 12 / 36 A single pallet on each face CAR-2 29 / 87 CAR-3 35 / 105 CAR-4 46 / 138 CAR-5 49 / pallets on each face distributed on 1 level 3 pallets on each face distributed on 1 level 4 pallets on each face distributed on 1 level 4 pallets on each face distributed on 2 level The first conclusion of this test series is the low combustion efficiency for a small number of pallets. In the most characteristic case, for 1 pallet, only 65% of the initial wood mass was lost at the end of

5 Part II Fire the fire test, the auto feeding is not sufficient enough to enable a full development of the fire. The second conclusion consists in the pallets arrangement for optimizing the HRR. The last case, CAR-5, shows that, for an identical number of pallets and quite the same mass, the HRR can be strongly different. This means that, using wood pallets for calibrated fire aggression requires to pay a specific attention to pallets arrangement. The HRR curves for those different tests are given on Fig. 3. Figure 3. HRR curves for the different wood pallets fire tests. On top of instrumentation described previously, flux meters were introduced inside the cube. The interest of those probes consists in following the fire development on each faces of the cube. Fig. 4 shows the evolution of HRR for the three faces based on the thermal flux measurement in the CAR-3 fire test. The main hypothesis for building those curves consists in assuming a constant radiative fraction during the whole fire. Such an hypothesis could led to overestimate the heat flux in the early stage of the fire. Figure 4. Example of thermal flux evolution for the three faces, CAR-3 case. 271

6 Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8) This analyze shows that, in the fire gallery configuration, the fire development is not exactly the same for each face, but the HRR maximum value and the duration of the peak is similar. Regarding the objective of those series of tests to evaluate the individual pallet HRR in warehouse fire configuration, this only means that the aggression on each face will be time shifted, which is quite similar to a real configuration. In the experimental case, this time shifting is probably induced by the air flow difference between the three faces that led to differences in terms of fire propagation even for right and left pallets arrangement. FOOD PALLET Fire tests described in the present paper concern food pallets. A large variety of such pallets were tested [9], only 3 are detailed here. It concerns salad pallet, frozen meat pallet and a frozen butter pallet that has a specific behaviour in case of fire. More details are given for salad pallet, for the other, results are limited to one graph containing HRR curves for the three tests. Salad pallet Because of the objectives of those tests, the pallet composition is one of the key issue. The details of this first pallet are given in Table 2. Table 2. Salad pallet composition for free burning test and wood pallet aggression test. Pallet compound (kg) Free burning test With wood pallet aggression Wood (pallet) Cardboard Punnet (polycarbonate) plastic film Salad Total The HRR and energy curves are presented on Fig. 5. Figure 5. HRR and energy curve for salad pallet without wood aggression. 272

7 Part II Fire The first important point highlighted by this curve is the dependence of the HRR curves on the ignition source. While, even in a real warehouse fire ignition, the first minutes should be ignition source dependent, it is important to keep that in mind for pallet fire calibration. This curve shows that, a maximum of 600 kw is reached in the first 5 minutes that corresponds to the propane burner period. This curve clearly shows the influence of the burner regarding the HRR and the importance to stop such an powerful fire source. After burners stop, the HRR quickly decreases to about 50 kw. 44 kg of mass were lost during this test, this means that about half of the pallet did not burn. Because, in real fire, high ignition source cannot be strictly avoided, the maximum HRR has to be considered for the wood charge to be placed in the neighborhood. This corresponds to the fire of two wood pallets on each faces. The relation with such an aggression is for both the peak HRR and total potential energy, this is important, mainly not for this test where potential energy and HRR is in great accordance, but for the other. The HRR curve for the two pallets fire, for test with wood aggression, and the HRR that corresponds to the sample alone in a surrounding fire is reproduced on Fig. 6. Figure 6. HRR for the wood aggression, sample with the aggression and resulting HRR for the sample. Two main conclusions has to be discussed for this test. The first is that almost all the mass of the pallet burnt during this test with wood aggression. This clearly indicates that, burning a single pallet is not representative of the behavior of the sample in real fire condition. Adding a representative but calibrated charge in the surrounding enables to get closer to real conditions. Some differences can still be quoted as for instance the fact that the pallets are located on the ground which limits chimney effect and probably influence the HRR, compared to a real configuration. The second conclusion is regarding the maximum HRR reached by the salad pallet during this test that is quite twice the one reached during the free burning test. It is important to note that the main assumption here consists in assuming that the HRR curve is identical for wood pallets, for the instrumented test and for the sample wood aggression test because of impossibility to distinguish the contribution of each combustible with classical calorimetry methods. An improvement for that should consists in using a tracer for the wood contribution. Frozen meat pallet As for the salad case, the pallet composition is the key for such test representability. Composition of the meat pallet is given in Table

8 Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8) Table 3. Frozen meat pallet composition for free burning test and wood pallet aggression test. Pallet compound (kg) Free burning test With wood pallet aggression Wood (pallet) Cardboard plastic film Meat Total As indicated previously, for this case and followings, HRR curves are given on one picture for the sample alone, the wood aggression and the sample with aggression in the surrounding on Fig. 7. Reasons detailed hereafter, aggression was made, for this test, using 4 pallets on each faces but on two levels to reduce the peak HRR. This corresponds, for this test, to the potential energy that can be released during the fire test. Figure 7. HRR curve for frozen meat fire tests with details (right). The main interest of this test consists on the impact of the large fire aggression on the HRR of the sample. It shows that, whether two HRR peaks are shown in case of aggression in the surrounding, the total HRR is globally lower than the HRR observed during the sample free burning. On top of that, while the wood fire was still present around the sample, the sample combustion was not so important after about 10 minutes. This should be explained by the fact that, for this test, meat was frozen. When heating the sample with an important thermal flux due to the important fire in the surrounding, an important release of water occurs and consequently reduces the HRR due to the sample itself. Once again, it is important to keep in mind that the main limit of such an evaluation consists in assuming the same fire behavior for the wood pallets for the two tests. THE SPECIFIC CASE OF FROZEN BUTTER The characteristics of the frozen butter pallet are given in Table

9 Part II Fire Table 4. Frozen butter pallet composition for free burning test and wood pallet aggression test. Pallet compound (kg) Free burning test With wood pallet aggression Wood (pallet) Cardboard plastic film Butter Total The HRR curves for tests corresponding to such a frozen butter pallet are given on Fig. 8. Figure 8. HRR curve for frozen butter fire tests. Note: Value of the second part of the curve (90 min) were influence by a ventilation change. During the sample free burning test, the peak HRR was about 950 kw for a total energy released of 1560 MJ, only 119 kg were lost during this test. While the beginning of the HRR curve is classic during the 30 first minutes with a quick increase of the HRR to its maximum before it decreases when stopping the burner, one can observe on this graph that HRR increases 40 minutes after ignition. This increase was caused by the pallet collapse. A butter pool was formed on the ground but was not ignited. Because of this low energy release compared to the available amount, only 3 wood pallets were placed on each faces for the sample fire with aggression. During the fire test with aggression, the HRR due to the pallet was, during the first phase of the fire, five times the maximum value observed during the free sample burning test. As for the sample free burning test, the curve presents two peaks, the first during the initial burning before pallet falling, the second due to the ignition of the butter pool that induced a pool fire. What is important to note here is that, because of the important HRR during this pool fire, the air flow rate was multiplied by about 2 but impossible to measure. Consequently, the real HRR maximum value, for the second peak, was probably higher than 10 MW. This value is however not highly important because the real HRR that should be observed in a real configuration would mainly be function of the pool surface. The HRR was here limited thanks to a 9 m 2 pool limit around the pallet. 275

10 Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8) This test confirms that a pallet free burning is clearly not representative of its behaviour in case of fire in a storage and a second test, with an aggression has to be achieved. It also shows that, for specific matter that can be liquefied before burning, the nature of the fire can be strongly modified when a sufficient amount of energy is given. ACKNOWLEDGMENT The authors thanks the French ministry of environment for funding this study, the USNEF and AFILOG for providing the sample for testing. CONCLUSIONS The objective of this paper was to propose and illustrate a fire protocol in order to determine the specific HRR of goods pallets that can be stored in warehouses by taking into account that fire can propagates to the surrounding. First of all, because of the fire behavior depends not only of the good itself but of the packaging too, the whole pallet has to be tested. Next, because the objective is not to evaluate the global fire propagation but the specific HRR induced by the pallet, a specific method has to be set up. It consists here in generating a surrounding fire thanks to specific arrangements of wood pallets. Thanks to this method, and using standard OC and CDG calorimetry, it is possible to determine the global HRR and, assuming that the wood fire behaves similarly between tests, to focus on the HRR due to the sample itself. Of course, this consists in the main assumption required here because the wood fire can also be influenced, not only in terms of HRR maximum value but in terms of kinetic too. Consequently, uncertainties on the HRR value can be quite important and has to be kept in mind when using those values for safety design. To go further in that specific pallet HRR characterization, new method is required to measure the HRR. The first should be to determine the specific calorimetric coefficient of each product for both OC and CDG and then, determine the burning rate of each material by making those two method corresponding. Of course, while such an approach is theoretically possible, its application to real configuration is not realistic because the large number of products distributed on the pallet and considering that the burning rate of each is varying along time. The simplest solution should consist in using a tracer for the wood contribution. REFERENCES 1. FDS user guide, NTST Special Publication Truchot, B. Fire Test Protocol for the Characterisation of Pallet Fire, INERIS Report n DRA A, Boehm, M., Fournier, L., and Truchot, B. Smoke Stratification Stability: Presentation of Experiments, Tunnel Safety and Ventilation, Graz, pp , Truchot, B. in Fire Safety Sciences News No. 38, Thauvoye, C., Russo, P., Blanchet, J. M., Duplantier, S., Kruppa, J., Muller, A., Patej, S., Taveau, J., and Zhao, B. Method for Calculating Heat Fluxes from a Warehouse Fire, International Performance-Based Codes and Fire Safety Design Methods, Janssens, M., and Parker, W. J. Oxygen Consumption Calorimetry, In: Babrauskas, V. Grayson, S. J. (Eds.), Heat Release in Fire, London, Tewarson, A. Generation of Heat and Fire Products, Technical Report of Factory Mutual, Truchot, B. Rapport d opération-base Données Produits Flumilog,