VALIDATION OF CONTAINER ANALYSIS FIRE ENVIRONMENT (CAFE) CODE FOR MEMORIAL TUNNEL FIRE VENTILATION TEST PROGRAM

Transcription

1 Proceedings of the ASME 21 Pressure Vessels and & Piping Division / K-PVP Conference PVP21 July 18-22, 21, Bellevue, Washington, USA PVP21-2 PVP VALIDATION OF CONTAINER ANALYSIS FIRE ENVIRONMENT (CAFE) CODE FOR MEMORIAL TUNNEL FIRE VENTILATION TEST PROGRAM N.R. Chalasani Research Assistant University of Nevada, Reno Mechanical Engineering Department Mail Stop 312, Reno Nevada, U.S.A chalasan@unr.nevada.edu Miles Greiner Professor University of Nevada, Reno Mechanical Engineering Department Mail Stop 312, Reno Nevada, U.S.A greiner@unr.edu Ahti Suo-Anttila Computational Engineering Analysis Albuquerque, NM, U.S.A ABSTRACT The Container Analysis Fire Environment (CAFE) computer code was developed at Sandia National Laboratories to predict the response of spent nuclear fuel (SNF) transport packages in large fires. CAFE s fire model has been benchmarked using measurements from large, unconfined outdoor fires. In the current work CAFE simulations are benchmarked using data acquired in two fires from the Memorial Tunnel test series. The Memorial Tunnel, a decommissioned highway tunnel in West Virginia, is 85 m (2,8 ft) long, 4.38 m (14.3 ft) wide, and has a 3.2% slope. In both fires, the time-dependent air temperature and speed were measured at several locations throughout the tunnel during 5 MW fires. The first test used forced ventilation and the upper portal of the tunnel was sealed. Shortly after the fire started, air was forced into the tunnel at a location between the sealed portal and the fire, forcing the air flow toward the lower portal. The second test used natural ventilation, in that both portals were open and there was no forced flow. However, wind outside the tunnel appeared to cause a net flow inside, even before the fire started. While the Memorial Tunnel Fire test conditions and results were well documents, some details were not available to the current authors. This necessitated the used of some assumptions. CAFE simulations accurately reproduced many of the characteristics of the temporal and spatial variation of the measured air speed and temperature. The maximum simulated temperatures for the forced and naturally ventilated tests were, respectively, 26 o F (14 o C) and 21 o F (94 o C) below the corresponding measured values. This work will be used to assess the accuracy of CAFE in predicting the likely response of SNF packages in historic transportation tunnel fires. INTRODUCTION Spent nuclear fuel is transported away from reactor sites in thick walled packages [1]. Before these packages receive a certificate from US Nuclear Regulatory Commission, their manufacturer must demonstrate they will maintain their containment, shielding and criticality control functions after a series of severe events [2]. These events include a 9-m drop onto an unyielding flat surface, a 1-m drop onto a steel puncture bar, full engulfment in an 1472 o F (8 C)-fire for 3 minutes, and water submersion. Transportation risk studies must assess the likelihood and consequence of all possible accidents [3, 4]. The assessment of package performance under severe conditions is performed using both experiments and engineering analysis. The Container Analysis Fire Environment (CAFE) computer code has been developed at Sandia National Laboratories to predict the response of transport packages to severe accident conditions [5]. CAFE links a computational fluid dynamics (CFD) fire simulator with a finite element (FE) package model. CAFE employs a number of physics-based fuel 1 1 Copyright 21 by ASME

2 evaporation, turbulent transport, reaction chemistry, and radiation heat transfer models. Parameters for these models must be determined based on comparison of CAFE simulation results with measurements acquired in large fire experiments [6, 7]. Large-scale outdoor fire tests have been performed to acquire data to benchmark CAFE [8, 9]. In these experiments, truck- and rail-package-sized pipe calorimeters were suspended over JP8 jet fuel fires. The calorimeter interior surface temperatures and external wind conditions were measured during and after a series of roughly 3-minute fires. CAFE simulations were performed using computational domains that modeled the experimental facility, and the measured wind conditions as boundary conditions [1-12]. These simulations were performed with different parameters for the physics-based models. The calorimeter temperature results from these simulations were compared to the experimental data to determine the most appropriate model parameter values, and to benchmark the simulation methods. The current work is part of an effort to predict the response of spent nuclear fuel transportation packages to historic tunnel fires. For example, severe transportation fires occurred in Caldecott highway tunnel (near Oakland, California) and Howard Street rail tunnel (in Baltimore, Maryland) in 1982 and 21 respectively [13, 14] The behavior of fires in tunnels may be somewhat different from that of outdoor fires because the supply of air in tunnel fires is somewhat limited. Before fire simulation codes can be used with confidence to simulate the potential response of spent nuclear fuel transport packages in tunnel fire events, these codes must be benchmarked against data acquired in relevant tunnel fires. A series of 98 tunnel fire experiments were performed in the Memorial Tunnel, a decommissioned highway tunnel in West Virginia, from 1993 to 1995 [15]. Various fire sizes and ventilation schemes were used. The tests were performed to determine the effect of several different forced and natural convection ventilation schemes within the tunnel. These tests provide data that can be used to benchmark fire simulation codes. However, one shortcoming is that they do not measure heat transfer to large objects in the experiments. Moreover, it is somewhat difficult for researchers who did not conduct the fire tests to know all details of the fires. This makes it difficult to use the data for benchmarking purposes. The Fire Dynamics Simulator (FDS) computer code, developed by the National Institute of Standards and Technology (NIST) [16], was benchmarked using forced [17] and natural [18] ventilation tests conducted in the Memorial Tunnel. FDS has been used to predict the temperatures in the Caldecott and Howard Street tunnel fires [18, 19], so they could be used to predict the possible response of transport casks if they had been located near those fires [19, 2] The goal of the current work is to benchmark CAFE against data acquired in for forced (test # 321A) and naturally (test # 52) ventilated fires using data acquired in the Memorial Tunnel fire test sequence. Future work will use CAFE to predict the possible response of a cask if it had been located near these fires. Those results will be compared to predictions made using FDS. MEMORIAL TUNNEL FIRE VENTILATION TEST PROGRAM The Memorial Tunnel is a two-lane road tunnel built in 1953 as part of the West Virginia Turnpike. Figure 1a shows the tunnel cross section, and consists of a lower rectangular roadway region, and an upper semicircular ventilation duct. The area of the roadway and duct regions are approximately 36 m 2 (39 ft 2 ) and 24 m 2 (26 ft 2 ), respectively. Extensive modifications were made to the tunnel ceiling and the duct above the tunnel roadway to facilitate various ventilation schemes. For forced ventilation tests the ceiling dividing the roadway region from the upper ventilation duct was in place. For natural ventilation tests, that ceiling was removed. Figure 1b shows an axial slice through the tunnel, and indicates the location of important components. The tunnel is 85 m (2,8 ft) long and has a 3.2 % upgrade from the south (lower) to the north (upper) portals. Four steel pans were installed approximately one-quarter of the tunnel length (238 m) from the south portal. The fire pans were set about.75 m above the tunnel floor and were filled with.15 m of water on which the measured supply of fuel oil (low sulfur Number 2 fuel oil) floated. A 4.5 m 2 (48 ft 2 ) exposed fuel surface area produced a nominal heat release rate of approximately 1 MW. Different pan sizes were used to produce nominal 1, 2, 3 and 5 MW fires. The 2, 3 and 5 MW pans were used together to generate a 1 MW fire. A closed loop system was use to monitor the fuel level in the pan and pump fuel into it at a rate that caused the fuel level to be fairly stable. The pump flow rate versus time was recorded. Temperature and velocity sensors were installed on 14 instrument trees (identified as loops) at various axial locations throughout the tunnel (Figure 1b). Table 1 shows the list of locations of components within the tunnel. The thermocouple loops are instrumented seven thermocouples for forced ventilation tests and eight thermocouples for natural ventilation. These thermocouples are identified as A through H, and their respective elevations above the tunnel floor are listed in Table 2. Velocity sensors were installed on only 1 loops (Table 1) at the same elevations as thermocouples. A total of 98 fire tests were performed in the Memorial Tunnel test sequence. The current work described in this paper focuses on two 5 MW tests that were used by other investigators to benchmark the FDS code. Test 321A used forced ventilation. In that experiment the north porthole was sealed and air was forced into the tunnel through a 28 m 2 (3 ft 2 )- port in the ceiling at z = -135 m (north of the fire). Cochard [15] used this data to benchmark FDS. Test 52 used 2 2 Copyright 21 by ASME

3 Uphill Ventilation Port Downhill North Portal Fan Room Loop 214 Loop 213 Loop 211 Loop 29 Loop 28 Loop 27 Loop 37 Loop 36 Loop 35 y Fire Loop 34 Loop 33 z Loop 32 Loop 31 Loop 22 Fan Room 8 m South Portal Figure 1 Memorial tunnel configuration cross-section Axial schematic showing fire location, ventilation port, thermocouple loops, south (Downhill) and north (Uphill) portals. Z [m] Loop Instrumentation U = Uphill of fire, D = Downhill of fire 615 North Portal (Upper) U T & V U T U T U T & V U T & V U -135 Ventilation port U T & V U T & V U T U T & V U -5 Fuel Pan N/A - 5 Fuel Pan N/A T & V D 3 33 T D T & V D T & V D T & V D 238 South Portal (Lower) D T = Thermocouples, V = Pitot Tubes Table 1 Instrumented loops with temperature and velocity sensors located at various cross sections throughout the tunnel between portalto-portal. Location of south (lower), north (upper) and ventilation port. Forced Ventilation Natural Ventilation y [m] y [m] Table 2 Locations of temperature and velocity sensors on the instrumented loops at different elevations in the tunnel for forced (321A) and natural (52) ventilation fire tests. 3 3 Copyright 21 by ASME

4 natural ventilation. The ceiling was removed, and the north (uphill) portal was 75%-sealed. McGrattan and Hamins [16] used this data to benchmark FDS. CAFE SIMULATIONS Figure 2a and 2b shows the computational grids used in the CAFE simulations of forced ventilation (321A) and natural ventilation (52) tests respectively. The experimental north and south portals are at z = -615 and 238 m, respectively. However, since the north (uphill) portal was sealed, and in order to reduce the computational domain size, the ends of the domain are at z = ±238 m (i.e. the length of the tunnel in the computational grid is 476 m). Hence, only seven instrument loops (which are modeled as 3-mm-thick vertical steel plates) at z = -62, -29, -11, 12, 3, 66 and 18 m are included in the grid as shown in Fig. 2c. Figure 2c also shows, ventilation port, vehicle simulants (modeled as rectangular steel plates) and fuel pan (modeled as insulated box) inside the tunnel. The axial location of the fuel pan center is z =. Its surface was placed at z = 1 m. The ceiling is at y = 4.33 m, and the side walls are at x = ± m. Figure 2a shows that the grid includes the tunnel walls, ceiling and floor. The side walls are modeled as 1 mm thick concrete. The number of grids in the x, y, and z directions are 33, 23 and 97, respectively, for a total of 73,623 cells. For the natural ventilation test (52) the ceiling was removed. Hence, the height of the tunnel is 7.9 m and the whole length (85 m) of the tunnel is modeled with a semicircular dome as the ceiling shown in Fig. 2b. Since, the instrument loops 27, 28, 29, 211, 213, 214 and 22 are more than 1 m away from the fuel pan, only seven instrument loops are modeled in the natural ventilation simulations as well. The number of grids in the x, y and z directions are 23, 29 and 112, respectively for a total of 74,74 cells. CAFE COMBUSTION MODEL DESCRIPTION The current combustion model used in the tunnel fire scenarios is an improved version of the previous combustion model [3]. It has been modified to be more general and applicable to a wider range of combustion, e.g. pool fires and flares. The combustion model also includes an eddy breakup effect to account for turbulent mixing, but that treatment is identical to the previously reported model and will not be discussed further here. The chemical kinetics phase of the combustion model is similar to the earlier model in that combustion is treated in three steps: an initial partial hydrogen burning phase (within the hydrocarbon fuel); an intermediate species burning and soot production phase; and a soot burning phase. The kinetics coefficients of the hydrogen burning phase are set to values which insure that the fire continues burning. This is a numerical requirement due to using large computational cells and has little impact upon the accuracy of the overall combustion model. (c) Vehicle Simulants Ventilation Port Fuel Loop Loop Loop Figure 2 Computational grid showing tunnel walls, instrumentation loops, fuel pan and vehicle stimulants. Axial view Forced ventilation Natural ventilation (c) 4 Side view. 4 Copyright 21 by ASME

5 In the first step, hydrogen burning is treated according to the following combustion kinetics molar equation: dx dt F = A * e * X F * X O2 Ta T In this expression, X F is the number of moles of fuel in a computational cell, and X is the number of moles of oxygen O 2 in the cell. The pre-exponential coefficient is A = 1 13, and T a = 1,5 K is an activation temperature. All of the kinetics coefficients reported here are in SI units. The kinetics follows the stoichiometry equation in which the coefficients are mass based (kg), and Intr is an intermediate species. 1kg Fuel +.53kg O 2.596kg H 2 O +.933kg*Intr MJ The second step is combustion of the intermediate species to form combustion products (H 2 O and CO 2 ) and soot. The kinetics equation for this reaction is identical to the one shown above, except that the coefficients are different, and the number of moles of fuel (X F ) is replaced by the number of moles of the intermediate species (X Intr ). The pre-exponential coefficient is A = 5*1 11 and the activation temperature is T a = 15,5 K. This second step follows the stoichiometry in the equation shown below. 1kg Intr kg O kg PC +.23kg Soot MJ In this expression, PC represents products of combustion (a mixture of H 2 O and CO 2 ). Again the stoichiometry coefficients are mass based. The third and final reaction is the combustion of soot. The kinetics equation for soot combustion is of a global reaction rate form based on mass: dm dt Soot.66 = A * T * M * M * e.33 Soot O2 Ta T In this expression, M Soot and M O2 are the masses (kg) of Soot and O 2 in a computational cell. The pre-exponential coefficient is A = 2*1 8, and the activation temperature is T a = 25,5 K. The global kinetics form also includes two exponents,.33 on soot mass (which accounts for the surface area weighting effect), and.66 on absolute temperature T. The third kinetics step follows the stoichiometry in the equation shown below. 1kg Soot kg O2 3.66kg PC + 32 MJ FORCED VENTILATION TESTS Boundary Conditions Figure 3a shows the ventilation air volume flow rate versus time after the fire was started, for forced ventilation Test 321A. This air entered the tunnel from a port in the ceiling at the location z = -135m. There was no ventilation for the first two minutes of the fire. The flow rate rose to between 75 and 1 m 3 /s (158, 9 and 211, 9 ft 3 /min respectively) for the next 28 minutes, and then stopped. Since the tunnel has a crosssectional area of 36 m 2, this corresponds to an average ventilation speeds between 2.1 and 2.7 m/s (413 and 531 fpm). Boundary conditions on the tunnel ends were set to the local hydrostatic pressure. This allowed flow to enter and leave the tunnel as needed by local pressure conditions. Figure 3b shows the heat release rate versus time for this test. The line marked "" shows the fuel pump rate times the fuel heat of reaction during the test. The flow rate to maintain the fuel level constant rose quickly for the first two minutes, and oscillated about a fairly constant value for the next 15 minutes. It then dipped for around 4 minutes, and then rose for 5 minutes. The fuel was essentially shut off approximately 26 minutes after it was turned down. Even though, this test is categorized as a 5 MW fire size test, the heat release rate is always less than the 5 MW except between time t = 22 and 24 min. Two simulations of this experiment were performed. In the first, the measured time dependent fuel flow rate was divided by the area of the fuel pool, and that time dependent fuel flux was injected into the domain. In the second, a constant fuel flux, equal to the total fuel volume divided by the pool area and total fire time (3 minutes) was applied. This corresponds to a heat release rate of 4 MW. That constant flow rate corresponds to the horizontal line in Figure 3b marked -Average. CAFE Simulation Results The line in Fig. 3b marked "CAFE-Variable" is the heat release rate calculated from simulation with the timedependent fuel injection rate. It follows the fuel injection rate but predicts a slightly lower heat release rate. The line marked "CAFE-Average" shows the calculated heat release rate from CAFE simulation for the constant fuel flux calculation. It begins above the rate predicted based on the average flow rate, but decreases to be near that value at the end of the simulation. Figure 4 shows maximum temperature at each thermocouple loop versus loop location relative to the fire, z for the forced ventilation Test 321A. Results are given at t = 2 minutes (when ventilation is first turned on), and t = 15 minutes (well after the ventilation is activated). Symbols are used for the measured data, thick lines are used for the simulations using the constant fuel flux, and thin lines are used for the timedependent fuel flux. 5 5 Copyright 21 by ASME

6 When the ventilation is first turned on, the highest temperature is north (uphill) of the fire (at z = -11 m), but it shifts to the south (downhill) of the fire (z = 12 m) after the ventilation is activated (in the direction of the net flow). The agreement with the data for the highest temperatures is better for the constant fuel injection rate simulations than for the variable fuel rate calculations. For this reason, the remaining results are given for the constant fuel flux simulations. Air Volumetric Flow rate [m 3 /s] Heat Release Rate [MW] min 5 min 3 min Time after ignition, t [min] - Average CAFE - Variable 1 min CAFE - Average Time after ignition, t [min] Figure 3 Forced ventilation test (321A) Time dependent volume air flow rate through ventilation port (28 m 2 ) at z = -135 m north of fuel pan. Heat Release Rate versus time. Line marked experiment shows the time dependent flow rate of fuel to the pan times the fuel heat of reaction, whereas the line marked experiment-average shows the average of heat of reaction from the experiment for 3 min. The lines marked cafe and cafe-average show the calculated time dependent and average heat release rate from CAFE simulations using experiment and experiment-average heat of reaction as boundary condition Maximum t = 2 min Loop location relative to fire, Z [m] CAFE - Variable CAFE-Average t = 15 min Figure 4 Maximum temperature at each thermocouple loop versus axial location z along the length of the tunnel for forced ventilation Test 321A at t = 2 and 15 min. Symbols (), thick lines (CAFE- Average) and thin lines (CAFE). Figure 5a and 5b show gas velocity towards the south portal versus time for three pitot tubes (y =.3, 3 and 4 m) located at loops 35 (z = -11 m) and 34 (z = 12m) south of ventilation port (z = -135 m). The measured data are shown on the left, and the CAFE simulations results are on the right. Since there is no ventilation for the first two minutes of fire, the gas at these loops is moving towards the north (uphill) portal due to natural ventilation (positive velocity). It is interesting to note that, the magnitude of gas velocity due to natural ventilation during the first two minutes is around 2 m/s (393 fpm) at 35, which is approximately equal to the average ventilation velocity after the ventilation is turned on as shown in Figure 3b. In contrast the velocity of gas for the first two minutes at loop 34 is around.5 m/s. When the ventilation is turned on after at t = 2 min, the gas velocities reverses its directions (negative velocity magnitude) almost immediately (t = 2 min) for loop 35 which is closer to the ventilation port. This oscillates around a constant velocity thereafter. But for loop 34 which is the farthest from the ventilation port, the reverse in gas velocity direction occurs later compared to loop 35. At around six minutes the maximum velocity of the gas is approximately 6 m/s (118 fpm) and oscillates around the same magnitude until 26 minutes. But, after 26 minutes the velocity of gas decreases in magnitude, even though the ventilation is turned on. This may be caused by the fire as the fuel injection is turned off at around the same time. Simulated gas velocities exhibit the same trends. There is a fairly good agreement between the simulation and measurements. 6 6 Copyright 21 by ASME

7 Velocity towards south portal [fpm] 6 Ventilation - On Fuel injection - Off 4 y =4 m y =.3 m Velocity towards south portal [fpm] Simulation 6 Ventilation - On 4 2 y =4 m y =.3 m Velocity towards south portal [fpm] Ventilation - On Fuel injection - Off y =.3 m y =4 m Velocity towards south portal [fpm] Ventilation - On y =4 m y =.3 m Figure 5 Gas velocity versus time for forced ventilation Test 321A. Loop 35 (z = -11 m) Loop 34 (z = 12 m). Simulation Air Supply Fuel injection - OFF Air Supply y = 4 m y =.3 m y = 4 m y =.3 m y = 4 m Air Supply Fuel Injection - OFF y =.3 m Air Supply y =.3 m y = 4 m Figure 6 Loop temperature versus time for forced ventilation Test 321A Loop 35 (z = -11 m) Loop 34 (z = 12 m) 7 7 Copyright 21 by ASME

8 Figure 6a shows temperature versus time for the three thermocouples (y =.3, 3 and 4 m) located at z = -11 m (uphill of the fuel pool at loop 35). The measured data are shown in the plot on the left, and the CAFE simulation results is on the right. Since there is no ventilation for the first two minutes of the fire, hot fire plume initially moves toward this location. The ventilation port is located north (uphill) of fire (at z = -135 m). After the ventilation is activated, the plume is blown away from loop 35. The measured and simulated temperatures on loop 35 rise at the beginning of the fire, reach a peak around t = 2.4 min (just after the ventilation is turned on), and then decrease. There is a fairly good agreement between measured and simulated temperatures. Figure 6b show similar graphs for loop 34, which is south (downhill) of fire at z = 12 m. Since this loops is downhill of fire, the fire plume does not reach it until after the ventilation is turned on. It is interesting that even though the loop is 12 m from the fire and the net ventilation is only on the order of 2 m/s, the measured temperatures of this loops begin to respond to the fire very soon after t =. The simulated temperatures also exhibit the same trends. The measured temperatures decrease rapidly after the fuel pump is deactivated, whereas the simulated temperatures do not. This is because the constant fuel flow rate is applied as a boundary condition for the first 3 minutes of the fire (future simulations will stop injecting fuel at t = 25 min). y [Height of the Tunnel, m] y [Height of the Tunnel, m] Figure 7 Measured (symbols) and simulated (lines) temperatures versus elevation of the tunnel for forced ventilation Test 321A Uphill oops 35, 36 and 37 (z = -11, -29 and -62 m) at t =2.4 min Downhill loops 31, 32, 33 and 34 (z = 18, 66, 3 and 12m) at t =15 min. Figure 7 shows the measured (symbols) and simulated (line) temperature versus elevation for all the loops for forced ventilation Test 321A. Figure 7a shows temperature versus elevation for the uphill loops 37 (z = -62 m), 36 (z = -29 m) and 35(z = -11 m) at t = 2.4 min (after ventilation is first turned on, where loop 35 has peak temperature), while Fig. 7b shows the temperatures of the downhill loops 34 (z = 12 m), 33 (z = 3 m), 32 (z = 66 m) and 31 (z = 18 m) at t = 15 min (well into the fire). In all cases the measured and simulated temperatures near the ceiling are hotter than those near the floor. The maximum deviation between the measured and simulated temperatures is 21 o F (94 o C) in the uphill loops whereas 14 o F (6 o C) in downhill loops at t = 2 and 15 min respectively. There is good qualitative and quantitative agreement between the simulations and measurements. Heat Release Rate [MW] Air Volume Flow Rate [m 3 /s] CAFE - Average - Average Figure 8 Natural ventilation Test 52 Heat Release Rate versus time. Line marked experiment shows the time dependent flow rate of fuel to the pan times the fuel heat of reaction, whereas the line marked experimentaverage shows the average of heat of reaction from the experiment for 15 min. The line marked cafe-average shows the calculated average heat release rate from CAFE simulation using experiment-average as boundary condition time dependent air flow rate measured at loop 22 (z = 218 m), 2 m away from the south portal. 8 8 Copyright 21 by ASME

9 NATURAL VENTILATION TESTS Boundary Conditions Figure 8a shows heat release rate versus time for natural ventilation test (test # 52). Line marked "" shows the fuel pumping rate times its heat of reaction during the test. The heat release rate rose quickly for the first one minute to 55 MW and in two minutes dropped quickly to 1 MW. For the next 12 minutes of the test, the heat release rate oscillates similarly. Hence, an average heat release rate (39.5 MW) is obtained by averaging the heat release rate for the 15 minute time period shown as "-Average". It was not possible to match the pre-fire tunnel ventilation distribution by applying a wind boundary condition outside the tunnel. Therefore the measured flow conditions at the tunnel inlet were used. Figure 8b shows volume air flow rate measured at loop 22 (z = 218 m), 2 m away from the south portal. Even though this is a natural ventilation test, high magnitude of air flow rate is measured at this location before the fire and is indicative of significant forced flow conditions due to windy conditions outside the tunnel. CAFE Simulation Results In Figure 8a the line marked "CAFE-Average" is the heat release rate calculated from CAFE simulation. The heat release rate is fairly high (6 MW) for the first 1.5 minutes, dipped to 45 MW and oscillated around 45 MW for the rest of the test. Figure 9a shows temperature versus time for four thermocouples (y =.3, 2.4, 5.7 and 7.3 m) located at z = -11 m (uphill of the fuel pool at loop 35). The measured data is shown on the left, and the CAFE simulation result is on the right. Due to buoyancy induced gas motion and the external wind, the hot fire plume moves towards this location. The measured and simulated temperatures at y = 5.7 and 7.3 m rise quickly to reach 14 o F (76 o C) and 172 o F (938 o C) respectively at t = 2.3 min and then decrease initially for 2 min and oscillate thereafter for the rest of the fire experiment. Whereas the temperatures measured at y =.3 and 2.4 m rise more slowly throughout the experiment. Similar trends are predicted in the simulation, except that simulations under predict the measured data. Figure 9b shows similar graphs for loop 34, which is downhill of fire, the fire: plume moves away from this location y = 7.3 m Simulation 12 y = 5.7 m 9 6 y = 2.4 m 3 y =.3 m y = 7.3 m 9 6 y = 2.4 m y = 5.7 m 3 y =.3 m y = 7.3 m y = 5.7 m y = 7.3 m 3 y = 5.7 m y =.3 & 2.4 m 3 y =.3 & 2.4 m Figure 9 Loop temperature versus time for natural ventilation (test # 52) test. Loop 35 (z = -11 m) Loop 34 (z = 12 m) 9 9 Copyright 21 by ASME

10 throughout the experiment. Temperatures measured at y =.3 and 2.4 m are nearly equal and constant. Measured temperatures at y = 5.7 and 7.3 m rise quickly to their peak temperatures 65 o F (343 o C) and 53 o F (277 o C) respectively at t = 1.3 min and are fairly constant thereafter. This is due to the relatively small gas motion near the tunnel floor. Simulated temperatures follow the measured trends but over-predict the measured data t = 3 min t = 1 min CAFE length of the Tunnel,z [m, Relative to fire location] Figure 1 Maximum temperature at each thermocouple loop versus axial location, z along the length of the tunnel for natural ventilation Test 52 at t = 3 and 11 min. Symbols () and lines (CAFE). Figure 1 shows maximum temperature at each thermocouple loop versus axial location (z) along the length of the tunnel, Results are given at t = 3 and 1 minutes. Symbols are used for measured data and lines for simulations. The maximum measured temperatures at t = 3 and 1 minutes are located north of fire at loop 35 (z = -11 m). This is due to the gas motion inside the tunnel. The maximum temperatures from simulation at t = 3 and 1 minutes are also at loop 35 (z = -11 m). However the simulations under-predict measured temperature by 186 o F (86ºC) at 3 minutes. There is a fairly good agreement between the simulations and measurements. Figure 11 shows the measured (symbols) and simulated (line) temperature versus elevation for all the loops. Results are given at t = 1 min. Symbols are used for the measured data and lines are used for the simulations. Figure 11a shows temperature versus elevation for the uphill loops 37 (z = -62 m), 36 (z = -29 m) and 35(z = -11 m) while Fig. 11b shows the temperatures of the downhill loops 34 (z = 12 m), 33 (z = 3 m), 32 (z = 66 m) and 31 (z = 18 m). In all cases the measured and simulated temperatures near the ceiling are hotter than those near the floor except for Loop 31 (z = 18m). As expected for the natural ventilation, the uphill loops (Loop 35, 36 and 37) exhibit higher temperatures than the downhill loops. The simulations accurately predict that the temperatures near ceiling for the uphill loops, but under predict near the floor and at the center. y, Height of the Tunnel [m] y, Height of the Tunnel [m] For the downhill loops (Loops 31, 32, 33 and 34), the measured and simulated temperatures near the floor are approximately the same. This is due to the absence of gas motion in the downhill portion of the tunnel. There is a fairly good agreement between the simulation and measurements. ACKNOWLEDGEMENTS 34 This paper describes work performed by the University of Nevada, Reno and for the NRC under Contract No The activities reported here were performed on behalf of the NRC Office of Nuclear Material Safety and Safeguards, Division of Spent Fuel Storage and Transportation. This paper is an independent product of UNR and does not necessarily Figure 11 Measured (symbols) and simulated (lines) temperatures versus elevation of the tunnel for natural ventilation test (test 52) at t = 1 min Loops 35, 36 and 37 (z = -11, -29 and -62 m) Loops 31, 32, 33 and 34 (z = 18, 66, 3 and 12m). 1 1 Copyright 21 by ASME

11 reflect the view or regulatory position of the NRC. The NRC staff views expressed herein are preliminary and do not constitute a final judgment or determination of the matters involving transportation of spent nuclear fuel. REFERENCES [1] Saling, J. H., and Fentiman, A. W., 21, Radioactive Waste Management, New York: Taylor and Francis,. [2] Packaging and Transportation of Radioactive Material, Title 1, Part 71, Code of Federal Regulations, U.S. Nuclear Regulatory Commission, Washington D.C. 3] Fischer, L. E., Chou, C. K., Gerhard, M. A., Kimura, C. Y., Martin, R. W., Mensing, R. W., Mount, M. E., Witte, M. C., 1987, " Shipping Container Response to Severe Highway and Railway Accident Conditions," NUREG/CR-4829, Volume 1, Lawrence Livermore National laboratory. [4] Sprung, J. L., Ammerman, D. J., Breivik, N. L., Dukart, R. J., Kanipe, F. L., Koski, J. A., Mills, G. S., Neuhauser, K. S., Radloff, H. D., Weiner, R. F., Yoshimura, H. R., March, 23, "Reexamination of Spent Fuel Shipment Risk Estimates, NUREG/CR-6672, Vol.1, Sandia National Laboratories, Albuquerque [5] Suo-Anttila, A. J., Container Analysis Fire Environment User Manual [6] Greiner, M., and Suo-Anttila, A.J., 24, Validation of the Isis-3D Computer Code for Simulating Large Pool Fires Under a Variety of Wind Conditions, ASME J. Pressure Technology, 126, pp [7] Greiner, M., and Suo-Anttila, A.J., 26, Radiation Heat Transfer and Reaction Chemistry Models for Risk Assessment Compatible Fire Simulations, Journal of Fire Protection Engineering, Vol. 16, pp [8] Kramer, M. A., Greiner, M., Koski, J. A., Lopez, C., and Suo-Anttila, A., 23 Measurements of Heat Transfer to a Massive Cylindrical Object Engulfed in a Circular Pool Fire, Journal of Heat Transfer, Vol. 125, pp [9] Greiner, M., del Valle, M., Lopez, C., Figueroa, V., and Abu-Irshaid, E., 29, Thermal Measurements of a Rail-Cask- Size Pipe-Calorimeter in JP8 Fuel Fires, ASME 29 Summer Heat Transfer Conference, HTC , July 19-23, 29, San Francisco, California U.S.A. [1] Are, N., Greiner, M., Suo-Anttila, A., 25, Benchmark of a Fast running Computational Tool for Analysis of Massive Radioactive Material Packages in Fire Environments, ASME Journal of Pressure Vessel Technology, Vol. 127 pp [11] Greiner, M., Chalasani, N. R., and Suo-Anttila, A., 28, Thermal Protection Provided by Impact Limiters to Containment Seal Within a Truck Package, ASME Journal of Pressure Vessel Technology, Vol. 13. [12] del Valle, M. A., Kramer, M. A., Lopez, C., Suo-Anttila, A., and Greiner, M., 27, Temperature Response of a Rail- Cask-Size Pipe Calorimeter in Large Scale Pool Fires, proceedings of the 15 th International Symposium on the Packaging and Transportation of Radioactive Materials (PATRAM) [13] NTSB/HAR-83/ Multiple Vehicle Collisions and Fire: Caldecott Tunnel, near Oakland, California, April 7, National Transportation Safety Board, Bureau of Accident Investigation, Washington D.C. [14] NTSB Fire Group Factual Report. NTSB Public Docket, Report Issue Date: October 22. NTSB Case Reference Number: DCA 1 MR 4; Accident Site Location Reference: Baltimore, MD (Howard Street Tunnel); Accident Date Reference: July 18, 21. [15] Massachusetts Highway Department, 1996, Memorial Tunnel Fire Ventilation Test Program, Interactive CD-ROM & Comprehensive Test Report, [16] McGrattan, K. B, Forney, G. P., Floyd, J. E., and Hostikka, S., November 21, Fire Dynamics Simulator (Version 2) User s Guide. Technical Report NISTIR 6784, National Institute of Standard and Technology, Gaithersburg, Maryland, USA. [17] Cochard, S., Validation of The Freeware Fire Dynamics Simulator Version 2. for Simulating Tunnel Fires. [18] McGrattan, K. B., and Hamins, A., February 23, Numerical Simulation of the Howard Street tunnel Fire, Baltimore, Maryland, July 21, NUREG/CR- 6793, National Institute of Standards and Technology, Washington D.C. [19] Adkins, H. E., Koeppel, B. J. Jr., Cuta, J. M., Guzman, A. D., Bajwa, C. S., December 26, Spent Fuel Transportation Package Response to the Caldecott Tunnel Fire Scenario, NUREG/CR-6894, Rev.1, Pacific Northwest National Laboratory, Washington D.C. [2] Adkins, H. E., Cuta, J. M., Koeppel, B. J. Jr., Guzman, A. D., Bajwa, C. S., December 26, Spent Fuel Transportation Package Response to the Baltimore Tunnel Fire Scenario, NUREG/CR-6886, Rev.2, Pacific Northwest National Laboratory, Washington D.C Copyright 21 by ASME