- Email: [email protected]
Hydrogen fuel-cell forklift vehicle releases in enclosed spaces
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/he
Hydrogen fuel-cell forklift vehicle releases in enclosed spaces W.G. Houf a, G.H. Evans a, I.W. Ekoto a,*, E.G. Merilo b, M.A. Groethe b a b
Sandia National Laboratories, 7011 East Ave, MS 9052, Livermore, CA 94551-0969, USA1 SRI International, Menlo Park, CA 94025, USA
article info
abstract
Article history:
Sandia National Laboratories has worked with stakeholders and original equipment
Received 12 December 2011
manufacturers (OEMs) to develop scientific data that can be used to create risk-informed
Received in revised form
hydrogen codes and standards for the safe operation of indoor hydrogen fuel-cell fork-
16 May 2012
lifts. An important issue is the possibility of an accident inside a warehouse or other
Accepted 25 May 2012
enclosed space, where a release of hydrogen from the high-pressure gaseous storage tank
Available online 23 June 2012
could occur. For such scenarios, computational fluid dynamics (CFD) simulations have been used to model the release and dispersion of gaseous hydrogen from the vehicle and to
Keywords:
study the behavior of the ignitable hydrogen cloud inside the warehouse or enclosure. The
Hydrogen fuel-cell forklift
overpressure arising as a result of ignition and subsequent deflagration of the hydrogen
Simulations
cloud within the warehouse has been studied for different ignition delay times and ignition
Experimental validation
locations. Both ventilated and unventilated warehouses have been considered in the
Hydrogen codes and standards
analysis. Experiments have been performed in a scaled warehouse test facility and compared with simulations to validate the results of the computational analysis. Copyright ª 2012, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
1.
Introduction
Hydrogen powered fuel-cells are replacing lead-acid batteries as the power source for commercial warehouse forklifts. The onboard storage of compressed hydrogen gas along with indoor refueling in the closed warehouse environment introduces new operational requirements that must be considered in the development of safety guidelines. In order to provide a scientific basis for these safety guidelines, validated engineering models of unintended hydrogen releases are needed for scenario and risk analysis. Sandia has been working with original equipment manufacturers (OEMs) to develop an understanding of unintended
releases from hydrogen powered forklift vehicles in enclosed spaces. Our initial efforts have focused on the gaseous hydrogen indoor dispensing code in NFPA 52 (extracted and also appearing in NFPA 2) [1,2] as it applies to hydrogen forklift vehicles in warehouses. Based on Table 9.4.3.2.1 in NFPA 52 a maximum of 0.8 kg of hydrogen may be dispensed and used in a forklift vehicle indoors without warehouse ventilation if the warehouse volume is at least 1000 m3 and has a ceiling height of at least 7.62 m. Larger fueling rates are allowed in unventilated warehouses if the warehouse volume is increased in accordance with Table 9.4.3.2.1 specifications. Forklift vehicle fueling and operation outside the parameter range specified by Table 9.4.3.2.1 is allowed if the warehouse is
* Corresponding author. Tel.: þ1 925 294 6586; fax: þ1 925 294 2276. E-mail address: [email protected] (I.W. Ekoto). 1 Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear Security Administration under Contract DE-AC04-94-AL85000. 0360-3199/$ e see front matter Copyright ª 2012, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijhydene.2012.05.115
8180
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
ventilated at a rate of a least 0.3 m3/min/m2 (1 ft3/min/ft2) of room area, but no less than 0.03 m3/min/0.34 m3 (1 ft3/min/ 12 ft3) of room volume. As a first approach a set of indoor hydrogen release experiments was designed based on OEM forklift specifications, NFPA 52 guidance, and a leak size based on an OEM generated failure modes and effects analysis (FMEA). A release scenario was defined where the entire contents (0.8 kg) of a 35 MPa hydrogen tank onboard a fuel-cell forklift was released and ignited at different times and locations inside a full-scale (1000 m3) warehouse. Hydrogen was released through a 6.35 mm diameter opening, meant to simulate either a medium sized leak from the FMEA or a release from a thermally activated pressure relief device (TPRD), into an onboard enclosure that contained the forklift fuel-cell storage tank. Hydrogen then exited through a grill on the forklift side and entered into the warehouse. Warehouse sizing, ventilation, and amount of onboard hydrogen were based on the parameters outlined in section 9.3.3 of NFPA 52 as discussed above. From the full-scale release scenario, subscale experiments were designed and performed in a scaled warehouse test facility (Fig. 1) at the SRI International Corral Hollow Experiment Site (CHES) to provide model validation data. Computational fluid dynamics (CFD) simulations of hydrogen releases from the fuel-cell powered forklift inside the warehouse were performed using the Sandia developed FUEGO [3] CFD model. Concentrations taken from FUEGO simulations were then used in a FLACS (2010) [4] model of the warehouse to simulate the overpressure generated by ignition of the hydrogen cloud at different ignition delay times. Subscale model experiment and simulation results were compared to validate the modeling approach. The validated model was then used to perform additional full-scale release simulations to provide information for risk-informed hydrogen codes and standards development. Details from the experiments and modeling are described in the following sections.
2.
Scaled warehouse experiment
Model validation experiments were performed in a hardened, subscale warehouse facility located at the SRI International Corral Hollow Experiment Site (CHES) located near Livermore, CA. Fig. 1a shows a photograph of the SRI test warehouse prior to construction of a rigid front wall that was added for the
deflagration experiments. Appropriate scaling factors [5] were determined to create a set of scaled-warehouse tests that resembled as closely as possible the full-scale warehouse release scenario. This scaling approach has been successfully used for previous tests at SRI that involved releases from scaled hydrogen fuel-cell vehicles in road tunnels [6]. The test facility volume and height were respectively (1/2.8)3 and 1/2.8 relative to the 1000 m3 full-scale warehouse with a 7.62 m ceiling height. The scaled warehouse dimensions were 3.64 m wide by 4.59 m long by 2.72 m high, with a total volume of 45.4 m3. A layout of the experimental setup is given in Fig. 1b. The mass of hydrogen released in the scaled test was related to the mass of hydrogen released in the full-scale scenario by the volume ratio, (1/2.8)3. The time for the scaled mass release was related to the time for full-scale mass released by the Froude number and a dimensionless timescale [5]. The transient nature of the hydrogen tank blowdown was modeled with the Sandia developed compressible network flow analysis code, NETFLOW [7,8]. The NETFLOW model accounted for choked flow at the 3.56 mm release orifice, expansion into the enclosure, and exhaust of the hydrogen through the steel perforated plate (51% flow area) on the forklift model side. The initial tank pressure for the experiments was 13.45 MPa and the tank volume was chosen to hold the correct amount of scaled hydrogen mass. The 3.56 mm choked flow orifice with a 0.75 discharge coefficient was selected so that experimental tank blowdown characteristics matched computed subscale values. Fig. 2 shows a comparison of the measured mass release rate from the experiment with the calculated and full- and subscale mass release rates. The forklift model measured 0.343 m long by 0.343 m wide by 0.435 m high and an image is shown in Fig. 3a. These dimensions are consistent with a scaled version of a class 3 “Reach Type” forklift. Hydrogen was released into a 131 mm wide by 131 mm high by 88 mm deep enclosed space on the forklift side facing the back wall of the garage. The release was initiated by actuation of a fast response, pneumatically operated valve (w75 ms activation time) located between the hydrogen storage tank and nozzle orifice. During a release, hydrogen from the nozzle passed through a porous frit and into an enclosure packed with steel beads that occupied w61% of the enclosure volume, and were held in place by a perforated steel plate with a 51% open area (see Fig. 3b). The bed of steel beeds and performated plate diffused the momentum of the choked hydrogen release and ensured a uniform velocity at the grate exit.
Fig. 1 e (a) Photograph and (b) layout of the scaled-warehouse test facility used for release experiments. The model forklift shown with release opening directed toward back wall.
100
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
10-1 10-2 10-3
Flow Rate [kg/sec]
Full Scale, Calculation SRI Scale, Calculation SRI Scale, Measurement
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
Time [sec]
Fig. 2 e Calculated and measured scaled mass release rates along with a comparison to the calculated full-scale mass release rates (based on 0.8 kg of hydrogen forklift release in a 1000 m3 full-scale warehouse).
Three different types of tests were performed in the scaled warehouse experiments: (1) hydrogen dispersion tests to measure the concentration variation with time in the unignited hydrogen cloud released from the model forklift; (2) ignition and flame front imaging tests to determine the shape of the flame front and its propagation speed; and (3) a set of deflagration tests where the overpressure produced from igniting the hydrogen cloud was determined for a given ignition delay time and ignition location. Here the ignition delay time is defined as the time between the beginning of the release and activation of the spark ignition source. Note that to have a consistent comparison of experimental results, tank mass flow rate and warehouse concentration measurements were referenced to the valve opening time; while propagation flame speed, flame imaging, and overpressure measurements were referenced to the spark time. These reference times have likewise been maintained when reporting simulation results in the following sections. The front wall configuration of the warehouse was altered depending on the type of test performed, and the various
8181
configurations are illustrated in Fig. 4. For the flame propagation imaging and dispersion tests, the warehouse front end was either left open (Fig. 1a) or covered with a high-density polyethylene (HDPE) plastic sheet (Fig. 4a), which allowed infrared and standard video visualization of the flame propagation. For the overpressure tests the warehouse open end was sealed with a steel wall that contained a center access doorway covered by a plywood blowout panel (Fig. 4bed). The ignition source was a spark igniter located either 3 cm above the top rear edge of the model forklift (on the side with the release opening grill) or 10 cm below the warehouse ceiling on the same vertical line (the midline of the grill on the side of the forklift). For the dispersion tests, oxygen depletion measurements from fast-response Teledyne UFO 130-2 oxygen sensors were used to measure hydrogen concentrations. Two dispersion tests were conducted within a well-sealed warehouse, while a third was conducted with active mechanical ventilation. The 6.3 m3/min ventilation flow rate was selected based on application of the scaling factors to the NFPA 52 full-scale warehouse ventilation guidelines. Active ventilation was accomplished by placing two muffin fans in the 3.05 m long and 34.0 cm diameter exhaust vent tube located on the back wall of the warehouse (see Fig. 1b). When the fans were running, air was pulled into the warehouse through an open vent (120.0 cm wide by 6.35 cm high) located near the bottom of the plywood pressure blowout panel on the steel front wall of the warehouse (see Fig. 4d). The flow rate through the exhaust tube was verified beforehand with an anemometer along seven radial locations 2.75 m downstream in the tube. Fig. 5 shows an infrared flame propagation image sequence from the scaled warehouse experiment when ignition occurred near the forklift top rear edge (3 s ignition delay time). The flame can be observed traveling upward along the released plume before propagating into the hydrogen/air mixture that had collected along the ceiling. When the release was ignited 10 cm below the warehouse ceiling (image sequence not shown) the flame propagated along the ceiling and was directed downward as it approached the warehouse side-walls. Note, however, that no downward propagation of the flame along the hydrogen release plume was observed. For the confined deflagration overpressure tests, the open end of the warehouse was sealed with a steel wall, and
Fig. 3 e (a) Photograph showing the exterior and release opening for the model forklift used in the experiments. (b) Schematic showing the interior of the model forklift with hydrogen bottle, refilling line, porous frit, steel bead enclosure and perforated steel release opening.
8182
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
Fig. 4 e Photographs of scaled-warehouse test facility showing the different front wall configurations used for the various experiments.
Fig. 5 e Infrared image sequence from the scaled warehouse experiments that illustrate flame front propagation approximately 0, 33, 400, and 800 ms after ignition. The ignition location was 3 cm above the top rear edge of the model forklift and occurred 3 s after the beginning of the release. The hydrogen was released through the back side of the model forklift, directed toward the back wall of the warehouse.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
a plywood blowout panel was used to cover the center access doorway (Fig. 4bed). Seven confined ignition tests were performed, which included several different blowout panel configurations and tests with either active or natural ventilation using the vent tube on the back wall and the vent near the bottom of a plywood pressure blowout panel (Fig. 4d). Pressure transducers were used to measure the generated overpressure, while fast-response thermocouples mounted along the facility ceiling were used to measure horizontal flame front propagation speed. The maximum measured overpressure in the test series was 24.6 kPa, which occurred when the warehouse was wellsealed and the ignition source was located 3 cm above the model forklift. When ignition occurred 10 cm below the ceiling, the maximum measured overpressure was moderately lower at 19.0 kPa. It was observed that small increases in warehouse ventilation area led to significant reductions in deflagration overpressure. Hence, the leakage area of the scaled warehouse was measured for each test configuration by replacing the plywood blowout panel with a computerized fan system, Model 3 Minneapolis Blower Door, manufactured by The Energy Conservatory. For the well-sealed warehouse configuration the leakage area was measured to be approximately 36.7 cm2. The measured exhaust tube leakage area with the muffin fans mounted in place was 197 cm2 and the total measured leakage area for the ventilated warehouse was 972 cm2. These leakage areas were incorporated into warehouse deflagration models for the purposes of comparing the simulated ignition overpressures with the experimental data. It is important to note that the rise in overpressure may have resulted in facility deformation, which in turn would have led to an associated increase in the effective leakage area. Only a brief summary of the experiments has been given here and further details can found in Ekoto et al. [9].
3. Scaled warehouse simulations and model validation The computational fluid mechanics code, FUEGO [3], was used to perform simulations of hydrogen fuel-cell forklift releases inside the scaled warehouse. FUEGO is a Sandia National Laboratories’ developed code designed to simulate turbulent reacting flow and heat transfer [3] on massively parallel computers, with a primary focus on heat transfer to objects in pool fires. The code was recently adapted for compressible flow and hydrogen combustion. The discretization scheme used in FUEGO is based on the control volume finite element method [10], where the partial differential equations of mass, momentum, and energy are integrated over unstructured control volumes. A standard two equation (ke3 ) turbulence model was applied to close the Favre averaged equations [11]. Transport equations were solved for the mass fractions of each chemical species, except for the dominant species (nitrogen in the present simulations) which was computed by constraining the sum of the species mass fractions to equal one. For compressible flow, the ideal gas equation of state was used to relate the density and pressure of the gas mixture. For the calculations reported here, the first order upwind scheme was used for the convective terms. The code solved the
8183
conservation equations in a time-dependent manner. A validation study that assesses the model’s ability to predict velocity and concentration decay along the centerline of unignited hydrogen free jets and the centerline temperature for laboratory-scale hydrogen jet flames is reported in Houf et al. [12]. FUEGO dispersion simulation boundary conditions for the floor and walls of the warehouse enclosure were no-slip and constant temperature (294 K) as were the boundary conditions on the forklift sides. The vent in the back wall of the enclosure was assigned an open boundary condition with a total pressure of 1 atm, a temperature of 294 K, oxygen and nitrogen mole fractions of 0.21 and 0.79 respectively, and negligible levels of turbulence (k ¼ 0.11 cm2/s2, 3 ¼ 1.51 e4 cm2/s3). For simulations involving forced ventilation, the ventilation mass flow rate was used to specify an inflow boundary condition on the inlet vent area in the front wall of the warehouse. Transient jet inflow boundary conditions for the hydrogen exiting the grill on the side of the forklift (velocity, temperature, concentration) were computed with NETFLOW [7,8] and input as transient inlet boundary conditions on the side of the forklift in the warehouse model. A grid resolution study was performed to insure that the concentration results were independent of mesh spacing and time step. In addition, a total mass balance on the amount of hydrogen in the enclosure and exiting the vents was performed and compared with the total mass of hydrogen released from the forklift. A FLACS [4] model was used to perform transient simulations of hydrogen gas cloud ignition along with the associated overpressure generated by the rapid deflagration wave propagation and expansion of hot combustion product gases inside the enclosure. FLACS solved the compressible NaviereStokes equations on a three-dimensional Cartesian grid using a finite volume method. Conservation equations for mass, momentum, enthalpy, turbulence, and species were solved in conjunction with the ideal gas law equation of state using the SIMPLE [13] pressure correction algorithm extended for compressible flow. The numerical model used a second-order scheme for diffusive fluxes and a second order “kappa” scheme (hybrid scheme with weighting between second order upwind and second order central differences, with delimiters for some equations) to resolve the convective fluxes. Although the model contained a second order scheme for time, it was not used for the deflagration calculations due to the requirement of short time steps for stability. A standard ke3 model was used for turbulence with modifications for turbulence generation behind sub-grid objects. In the simulation, FLACS treated the deflagration wave as a collection of flamelets, where the burning velocity for each was computed based on the following three submodels: (a) a laminar burning velocity model that describes the laminar burning velocity as a function of gas mixture, concentration, temperature, pressure, oxygen concentration in air and the amount of inert diluents, (b) a model describing quasi-laminar combustion in the first phases of flame propagation after ignition, and (c) a model that describes the turbulent burning velocity as a function of turbulence parameters such as the intensity and length scale. A roughly cubic grid was used for all FLACS deflagration calculations with a mesh spacing of w10 cm. A number of grid resolution studies were performed to ensure the calculations
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
were independent of mesh size. The basic equations used in the FLACS model are described in Hjertager [14,15] and validation of the model against experimental data for hydrogen deflagration can be found in Middha and Hansen [16]. High-pressure hydrogen gas was exhausted from the forklift side grating through a 131 mm by 131 mm opening with a uniform exit velocity and was directed toward the back wall of the warehouse. In the forced ventilation simulations the FUEGO scaled warehouse model included the open area where the exhaust tube was located in the warehouse back wall, and the lower vent located at the bottom of the plywood blowout panel. Fig. 6 shows a still frame, 3 s into the release, from the simulation of the transient hydrogen cloud in the scaled warehouse for a case without active ventilation. From the image, the hydrogen plume can be observed rising from the forklift side vent and collecting along the ceiling. Initial conditions for the FLACS simulations were based on threedimensional concentrations maps extracted from the FUEGO dispersion simulations at the time of ignition. The FLACS model included the ventilation openings in the back and front walls and additional models were built that allowed simulations to be performed with either an open warehouse front wall or the plywood blowout panel. The small warehouse leakage areas measured with the blower apparatus as part of the experiments were likewise incorporated into the model by adding a small slot of equivalent leakage area near the bottom center of the front wall. Pretest FUEGO/FLACS simulations of the scaled warehouse were performed prior to the experiments to determine the appropriate placement of concentration sensors and to estimate the expected overpressure from ignition of the hydrogen releases. Fig. 7 summarizes the predicted peak overpressures for four separate warehouse configurations, with ignition located near the forklift model, as a function of ignition delay time. For all conditions with the exception of the inclusion of passive ventilation, the peak overpressure occurred after a 3 s ignition delay, with a corresponding maximum deflagration overpressure predicted to be roughly 30 kPa if the warehouse
0.35 Closed Front Wall (Length - 4.59 m)
0.30 Overpressure (barg)
8184
Closed Front Wall (Length = 6.10 m)
0.25 0.20
Closed Front Wall Vent in Back Wall (30 cm x 30 cm)
0.15 0.10 Open Front Wall
0.05 0.00
0
1
2
3
4
5
6
7
Ignition Delay Time (sec) Fig. 7 e Pretest FLACS simulation results of the maximum deflagration overpressure as a function of ignition delay time for different scaled warehouse test configurations. Note that these pretest results neglect wall heat transfer and effective leakage area.
was completely sealed and the effects of wall heat transfer were neglected. Accordingly, the 3 s ignition delay time was selected for the experiments, and it was decided that a pressure relief blowout panel should be included in the experimental setup to prevent unintended damage to the test facility. A comparison of predicted mole fractions relative to experimental measurements, both with and without forced ventilation, is given in Fig. 8 at four sensor locations. Close agreement is exhibited between the simulation and experimental results. Importantly, both the experimental and simulation datasets indicated the effect of forced ventilation on the hydrogen concentration profiles is small for this leak size.
1 prediction S01 no ventilation prediction S01 ventilation data S01 1026 data S01 1027 prediction S05 no ventilation prediction S05 ventilation data S05 1026 data S05 1027 prediction S04 no ventilation prediction S04 ventilation data S04 1026 data S04 1027 prediction S08 no ventilation prediction S08 ventilation data S08 1026 data S08 1027
H2 mole fraction
0.8 0.6 0.4 0.2 0 0 Fig. 6 e Simulation snapshot of the flammable volume of hydrogen (4%e75% mole fraction) in the scaled warehouse experiment 3 s into the release for the condition without ventilation, along with the location of the concentration sensors.
1
2
3
4
5 6 time (s)
7
8
9
10
Fig. 8 e Comparison of measured and predicted concentrations with and without forced ventilation in the scaled warehouse at four different sensor locations (test 1026 performed without active ventilation and test 1027 performed with active ventilation).
8185
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
a
b
30
5 Simulation (Natural Ventilation) Simulation (Forced Ventilation)
Simulation - Sealed (No Heat Transfer)
4
Simulation Sealed (Heat Transfer) 2
20
Simulation - 36.7 cm Leak Area (No Heat Transfer)
15 Data
10
2
Simulation - 36.7 cm Leak Area (Heat Transfer)
Data (Natural Ventilation) Data (Force Ventilation)
3
2
1
5 0
Overpressure (kPa)
Overpressure (kPa)
25
0
2
4 6 Time (sec)
8
10
0
0
0.5
1
1.5 2 2.5 Time (sec)
3
3.5
4
Fig. 9 e Comparison of simulation and experimental data of deflagration overpressure for the well-sealed subscale warehouse with ignition 3 cm above the forklift for the case with (a) no ventilation where the simulations progressively include the effective leakage area, wall heat transfer, or both effects; and (b) either forced ventilation (w6.3 m3/min) or natural ventilation (ventilation fans off) through the vents. Note that the reported time is relative to the spark ignition.
The experimental deflagration overpressure and decay rate within the scaled warehouse was dependent of the rigidity of the enclosure and how well it was sealed. Several different strength blowout panels were examined to investigate the impact on overpressure characteristics once they were breached. Initial blowout panels were built from a single plywood sheet (1.27 cm thick) bolted to the steel wall. A wooden frame (5.08 cm by 10.16 cm thick) was used to increase attachment strength to the facility (see Fig. 4b). For the well-sealed warehouse results shown in Fig. 9a where no breach was desired, the number of plywood sheets was increased to 3. Metal Unistrut was then mounted across the front surface of the plywood as shown in Fig. 4c to further increase the panel’s strength and to minimize panel bending. This reinforced panel did not breach and remained relatively well-sealed for the test duration. Fig. 9a shows a comparison of predicted deflagration overpressure from the simulations with data from a pressure
a
sensor located in the center of the warehouse sidewall near the forklift and positioned 1.24 m off the floor. Simulation results are shown with and without wall heat transfer, both for the completely sealed warehouse and for the warehouse with the measured effective leakage area (36.7 cm2) included. The near forklift ignition point was used for these simulations. Once wall heat transfer and the measured effective leakage rate are included into the simulations, the results exhibit excellent agreement with the experimental data in both the magnitude and shape of the pressure profiles. Although the pressure decay from the peak value was somewhat slower than was observed in the measurements, facility deformation that resulted from large stresses due to the deflagration pressure wave likely increased the effective leakage area, and hence increased the pressure relief. Note also that while the presented results are relative to the spark time, there was an initial w1 kPa pressure rise prior to ignition that was observed for both the experimental and simulation results, and was
b
25
Simulation (Well Sealed)
Overpressure (kPa)
20
Data (Well Sealed)
15
Simulation (Blowout Panel Breach)
10 5
Data (Blowout Panel Breach)
0 -5
0
1
2
3
4 5 Time (sec)
6
7
8
Fig. 10 e (a) Single frame of a high-speed video showing breach of plywood blowout panel on front of the scaled warehouse after ignition of hydrogen release from model forklift. (b) Comparison of simulation and experimental data for a well-sealed warehouse with and without plywood blowout panel breach (3 s ignition delay time). Ignition location is 3 cm above the forklift and time is referenced to the spark ignition.
8186
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
25 Simulation (Ign. above forklift)
Overpressure (kPa)
20
15 Simulation (Ign. 10 cm from ceiling)
10
Fig. 12 e Simulation results that show the time evolution of flammable hydrogen volumes (4%e75% mole fraction) that result from the release of 0.8 kg of hydrogen into the fullscale warehouse for cases with and without forced ventilation.
Data (Ign. 10 cm from ceiling)
5 Data (Ign. above forklift)
0
0
1
2
3
4
5
6
Time (sec) Fig. 11 e Comparison of simulation results and experimental data for ignition 3 cm above forklift (3.0 s ignition delay time) and ignition 10 cm from warehouse ceiling (3.5 s ignition delay time). Note that the simulations include the 36.7 cm2 effective ventilation area and wall heat transfer.
attributed to the rapid expansion of compressed hydrogen into the enclosed volume. Fig. 9b compares simulation and experimental deflagration overpressure data for the scenarios with either forced (ventilation fans on) or natural ventilation (ventilation fans off) through the warehouse vents. From the experiments, overpressure for the case with forced ventilation was approximately 3 kPa while the case with natural (passive) ventilation was slightly higher at 4.2 kPa. Note that the difference in overpressure was probably within the experimental variability of the tests. Nonetheless, predicted overpressure profiles were similar in magnitude and shape, with the roughly1 kPa difference in the conditions with and without mechanical ventilation preserved. The only major difference of note was that the simulation pressure peak was shifted w0.4 s earlier in time. These results demonstrate that the impact of forced ventilation on the peak deflagration overpressure remained small. It is also worth noting that the initial overpressure increase observed for the well-sealed conditions due to the hydrogen release was not observed in either the experimental or simulation results for the conditions with
ventilation. Further details and discussion about this phenomenon can be found in Ekoto et al. [9]. A simulation of the breach of the pressure relief panel was modeled by incorporating a panel of the same size and location as the plywood blowout panel into the FLACS model. In the model the pressure relief panel was set to open at the same overpressure where the experimental data indicated a breach occurred (w13 kPa). Fig. 10a shows an image frame from a high-speed movie taken of the single sheet plywood blowout panel during breach. Fig. 10b compares simulation and experimental overpressure data for the well-sealed warehouse with and without breach of the plywood pressure blowout panel. Results for the simulation and experiment again exhibited good agreement, with Helmholtz pressure wave oscillations observed in the pressure data for both conditions, although the ringing intensity was higher in the experiments. These results indicate that pressure relief panels can provide an immediate and effective means to reduce deflagration pressure rise, and the simulations are able to capture the characteristic ringing behavior that results once these panels breach. The final comparison was for the well-sealed warehouse condition with the ignition point located 10 cm below the warehouse ceiling. For this condition, the ignition delay was increased from 3.0 to 3.5 s to ensure a sufficiently flammable hydrogen/air mixture at the release point was present. Results are compared in Fig. 11, and the near forklift overpressure data is included for comparison. The experimental overpressure data exhibited a roughly 5.6 kPa reduction in the peak value relative to the near forklift ignition condition. The associated reduction for the simulations, however, was more modest at around 1.5 kPa. The disagreement may be due to the
Table 1 e Comparison of simulation predicted peak deflagration overpressure from transient hydrogen releases in scaled and full-scale warehouses for a 3.0 s ignition delay time and ignition 3 cm above the top of the forklift.
Scaled-warehouse (S.F. ¼ 1/2.8) Full-scale warehouse
100% Sealed (no heat transfer)
100% Sealed (heat transfer to walls)
Natural ventilation (vents open)
Active ventilation (vents open with forced flow)
29.8 kPa 27.2 kPa
26.7 kPa 24.5 kPa
4.9 kPa 14 kPa
4.2 kPa 15.5 kPa
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
8187
Fig. 13 e Simulation results of the flammable hydrogen cloud (4%e75% mole fraction) 3 s into the 0.8 kg hydrogen release for the full-scale warehouses with a 7.62 m high ceiling and volumes of 1000 m3, 1500 m3, and 2000 m3 (without ventilation).
fact that a single sheet plywood blowout panel was used for the ceiling ignition test, while a more rigid 3-sheet plywood blowout panel was used for the tests where ignition occurred 3 cm above the forklift. The single sheet plywood panel may have undergone greater deformation and allowed a greater amount of leakage during the pressure rise period. Another possible explanation is that the FLACS simulation for the ceiling ignition condition indicates that the flame propagated down along the release plume toward the model forklift. This result contradicts the observation from the infrared flame visualization that downward flame propagation along the release plume did not occur. The FLACS flame propagation model appeared to not distinguish between upward and downward propagating hydrogen flammability limits, but instead assigned a uniform value regardless of propagation direction. Although the generally accepted volumetric concentration for the upward-propagating lower flammability limit of hydrogen in air is 4%, experimental data in the literature indicate that the limit may be as high as 7.2% for horizontally-propagating flames and 9.5% for downwardpropagating flames [17,18].
4.
Full-scale warehouse simulations
Full-scale forklift-in-warehouse release simulations were performed with the FUEGO/FLACS models to evaluate
whether the full-scale warehouse release and deflagration would behave in the same manner as observed in the scaled warehouse experiments. The geometry of the scaled SRI warehouse, including the forklift and its location, was converted to a full-scale warehouse model by multiplying all of the lengths in the model by a factor of 2.8. This full-scale warehouse had a ceiling height of 7.62 m and was 12.86 m long by 10.20 m wide with a volume of 1000 m3. A total of 0.8 kg of hydrogen was released from the forklift tank at 35 MPa through a 6.35 mm opening into a 0.118 m3 enclosure with a vent open area on the forklift side equal to 674 cm2. The NETFLOW compressible network flow analysis code was again used to model the transient nature of the tank blowdown assuming a 50% blockage by the grill (see Fig. 2); the results were applied for the FUEGO boundary condition at the exhaust opening on the forklift side. Full-scale front and back wall vents were incorporated into the model, while active and passive (natural) ventilation, along with completely sealed scenarios were also examined. For the active ventilation simulations the full-scale ventilation rate of 0.03 m3/min/0.34 m3 of room volume, as specified in NFPA 52 [1], was used. Table 1 shows comparisons of simulated peak ignition deflagration overpressure in the scaled and full-scale warehouse releases for a 3 s ignition delay time. For the well-sealed warehouse conditions the peak ignition deflagration overpressure with wall heat transfer for the full-scale warehouse
Fig. 14 e Simulation results of the flammable hydrogen cloud (4%e75% mole fraction) 6 s into the 0.8 kg hydrogen release for the full-scale warehouses with a 7.62 m high ceiling and volumes of 1000 m3, 1500 m3, and 2000 m3 (without ventilation).
8188
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
was 24.5 kPa as compared to 26.7 kPa in the (1/2.8) scaled warehouse geometry. The results for the full-scale warehouse show that forced flow ventilation had little effect on deflagration overpressure as compared to the case where the vents were left open without forced flow. A similar effect of forced ventilation on the deflagration overpressure for the scaled warehouse experiments was observed (see Fig. 9b). Profiles of the predicted facility flammable volumes, defined as the total volume of flammable hydrogen/air mixtures, are depicted in Fig. 12 for the cases with passive ventilation and mechanical ventilation at the rate specified by NFPA 52 (6.3 m3/min) [1]. These results indicate that 3 s into the release there was no discernible difference in flammable volumes, and only a modest 10% difference was predicted by 16 s into the simulation. A series of full-scale warehouse simulations were also performed for the 0.8 kg forklift release, where the warehouse volume was increased from 1000 m3 to 1500 m3 and 2000 m3, while keeping the warehouse ceiling height constant at 7.62 m. The location of the forklift remained midway between the front and back walls of the warehouse and at a fixed distance of 3.04 m from the nearer sidewall in all cases. Figs. 13 and 14 show the flammable hydrogen clouds (4%e75% mole fraction) for a 0.8 kg forklift release in warehouses of volumes 1000 m3, 1500 m3, and 2000 m3 with a 7.62 m ceiling with and without ventilation at 3 s and 6 s after the start of the release, respectively. Fig. 15 shows the predicted peak overpressure for ignition of these clouds at ignition delay times of 3 s and 6 s. The results indicate that the peak deflagration overpressure decreased almost linearly with increased warehouse volume.
5.
Summary and conclusions
Simulation results of an unintended release from a hydrogen fuel-cell forklift vehicle in a full-scale warehouse have been performed for the case where hydrogen gas is vented from the side of the forklift vehicle as a result of a medium sized leak or thermal activation of a pressure relief device. The same modeling approach used in the full-scale warehouse release was validated in a scaled model by comparing simulation results with measured data from a series of scaled-warehouse release experiments performed at the SRI Corral Hollow Experiment Site. Results of the simulations were found to be in good agreement with the experimental data. The results of the modeling and experiments demonstrated that passive natural ventilation or pressure relief panels can be an effective means to mitigate deflagration overpressure arising from ignition of hydrogen plumes created by medium- and large-scale transient releases. However, the impact of forced ventilation had little effect on hydrogen concentration or deflagration overpressure in the early stages of the currently sized transient release scenario. Full-scale simulation results further indicated that increased warehouse volume beyond current requirements resulted in a linear reduction in peak overpressure.
Acknowledgments This work was supported by the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, Fuel Cell Technologies Program under the Safety, Codes, and Standards subprogram element managed by Antonio Ruiz.
references
Fig. 15 e Simulation results that show the peak ignition overpressure in a well-sealed full-scale warehouse for a 0.8 kg hydrogen release where ignition occurs 3 cm above the top of the forklift and the ceiling height is 7.62 m. Results include heat transfer to the warehouse walls and show the effect of changing the ignition delay time and warehouse volume.
[1] NFPA 52 Vehicular Gaseous Fuel Systems Code 2010 Edition, NFPA, 1 Batterymarch Park, Quincy, MA 02169e07471. [2] NFPA 2 Hydrogen Technologies Code 2011 Edition, NFPA, 1 Batterymarch Park, Quincy, MA 02169e07471. [3] Moen CD, Evans GH, Domino SP and Burns SP. A multimechanics approach to computational heat transfer. In: proceedings 2002 ASME Int. Mech. Eng. congress and exhibition. New Orleans, IMECE2002e33098: November 17e22, 2002. [4] FLACS version 9.1 user’s manual. Bergen, Norway: GEXCON; 2010. [5] Hall DJ, Walker S. Scaling rules for reduced-scale field releases of hydrogen fluoride. J Hazard Mater 1997;54:89e111. [6] Houf WG, Evans GH, Merilo EG, Groethe MA and James SC. Simulation of hydrogen releases from fuel-cell vehicles in tunnels. In: 18th world hydrogen energy conference. Essen, Germany: May 16e21, 2010. [7] Winters WS. A new approach to modeling fluid/gas flows in networks. Sandia National Laboratories; July 2001. Report SAND2001-8422. [8] Winters WS. Implementation and validation of the NETFLOW porous media model. Sandia National Laboratories; October 2009. Report SAND2009-6838. [9] Ekoto IW, Merilo EG, Houf WG, Evans GH and Groethe MA. Experimental investigation of hydrogen release and ignition from fuel cell powered forklifts in enclosed spaces. Int J hydrogen energy, (2012), http://dx.doi.org/10.1016/j.ijhydene. 2012.03.161.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
[10] Minkowycz WJ, Sparrow EM, Schneider GE, Pletcher RH. Elliptic systems: finite element method 1, handbook of numerical heat transfer. New York: J. Wiley and Sons, Inc; 1988. p. 379e420. [11] Papageorgakis GC, Assanis DN. Comparison of linear and nonlinear RNG-based k-epsilon models for incompressible turbulent flows. Numer Heat Transfer, Part B 1999;35:1e22. [12] Houf WG, Evans G, Schefer RW. Analysis of jet flames and unignited jets from unintended releases of hydrogen. Int J Hydrogen Energy 2009;34:5961e9. [13] Patankar SV. Numerical heat transfer and fluid flow. Hemisphere Pub; 1980.
8189
[14] Hjertager BH. Computer-simulation of turbulent reactive gas-dynamics. Model Ident Control 1984;5:211e36. [15] Hjertager BH. Three-dimensional modeling of flow, heat transfer, and combustion, handbook of heat and mass transfer. Houston: Gulf Publishing Company; 1986. p. 304e50. [16] Middha P, Hansen OR. Using computational fluid dynamics as a tool for hydrogen safety studies. J Loss Prevent Proc 2009; 22:295e302. [17] Coward HF, Jones GW. Limits of flammability of gases and vapors. Bur Mines Bull 1952;503. [18] Zebetakis MG. Flammability characteristics of combustible gases and vapors. U.S. Bur Mines Bull 1965;627.
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/he
Hydrogen fuel-cell forklift vehicle releases in enclosed spaces W.G. Houf a, G.H. Evans a, I.W. Ekoto a,*, E.G. Merilo b, M.A. Groethe b a b
Sandia National Laboratories, 7011 East Ave, MS 9052, Livermore, CA 94551-0969, USA1 SRI International, Menlo Park, CA 94025, USA
article info
abstract
Article history:
Sandia National Laboratories has worked with stakeholders and original equipment
Received 12 December 2011
manufacturers (OEMs) to develop scientific data that can be used to create risk-informed
Received in revised form
hydrogen codes and standards for the safe operation of indoor hydrogen fuel-cell fork-
16 May 2012
lifts. An important issue is the possibility of an accident inside a warehouse or other
Accepted 25 May 2012
enclosed space, where a release of hydrogen from the high-pressure gaseous storage tank
Available online 23 June 2012
could occur. For such scenarios, computational fluid dynamics (CFD) simulations have been used to model the release and dispersion of gaseous hydrogen from the vehicle and to
Keywords:
study the behavior of the ignitable hydrogen cloud inside the warehouse or enclosure. The
Hydrogen fuel-cell forklift
overpressure arising as a result of ignition and subsequent deflagration of the hydrogen
Simulations
cloud within the warehouse has been studied for different ignition delay times and ignition
Experimental validation
locations. Both ventilated and unventilated warehouses have been considered in the
Hydrogen codes and standards
analysis. Experiments have been performed in a scaled warehouse test facility and compared with simulations to validate the results of the computational analysis. Copyright ª 2012, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
1.
Introduction
Hydrogen powered fuel-cells are replacing lead-acid batteries as the power source for commercial warehouse forklifts. The onboard storage of compressed hydrogen gas along with indoor refueling in the closed warehouse environment introduces new operational requirements that must be considered in the development of safety guidelines. In order to provide a scientific basis for these safety guidelines, validated engineering models of unintended hydrogen releases are needed for scenario and risk analysis. Sandia has been working with original equipment manufacturers (OEMs) to develop an understanding of unintended
releases from hydrogen powered forklift vehicles in enclosed spaces. Our initial efforts have focused on the gaseous hydrogen indoor dispensing code in NFPA 52 (extracted and also appearing in NFPA 2) [1,2] as it applies to hydrogen forklift vehicles in warehouses. Based on Table 9.4.3.2.1 in NFPA 52 a maximum of 0.8 kg of hydrogen may be dispensed and used in a forklift vehicle indoors without warehouse ventilation if the warehouse volume is at least 1000 m3 and has a ceiling height of at least 7.62 m. Larger fueling rates are allowed in unventilated warehouses if the warehouse volume is increased in accordance with Table 9.4.3.2.1 specifications. Forklift vehicle fueling and operation outside the parameter range specified by Table 9.4.3.2.1 is allowed if the warehouse is
* Corresponding author. Tel.: þ1 925 294 6586; fax: þ1 925 294 2276. E-mail address: [email protected] (I.W. Ekoto). 1 Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear Security Administration under Contract DE-AC04-94-AL85000. 0360-3199/$ e see front matter Copyright ª 2012, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijhydene.2012.05.115
8180
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
ventilated at a rate of a least 0.3 m3/min/m2 (1 ft3/min/ft2) of room area, but no less than 0.03 m3/min/0.34 m3 (1 ft3/min/ 12 ft3) of room volume. As a first approach a set of indoor hydrogen release experiments was designed based on OEM forklift specifications, NFPA 52 guidance, and a leak size based on an OEM generated failure modes and effects analysis (FMEA). A release scenario was defined where the entire contents (0.8 kg) of a 35 MPa hydrogen tank onboard a fuel-cell forklift was released and ignited at different times and locations inside a full-scale (1000 m3) warehouse. Hydrogen was released through a 6.35 mm diameter opening, meant to simulate either a medium sized leak from the FMEA or a release from a thermally activated pressure relief device (TPRD), into an onboard enclosure that contained the forklift fuel-cell storage tank. Hydrogen then exited through a grill on the forklift side and entered into the warehouse. Warehouse sizing, ventilation, and amount of onboard hydrogen were based on the parameters outlined in section 9.3.3 of NFPA 52 as discussed above. From the full-scale release scenario, subscale experiments were designed and performed in a scaled warehouse test facility (Fig. 1) at the SRI International Corral Hollow Experiment Site (CHES) to provide model validation data. Computational fluid dynamics (CFD) simulations of hydrogen releases from the fuel-cell powered forklift inside the warehouse were performed using the Sandia developed FUEGO [3] CFD model. Concentrations taken from FUEGO simulations were then used in a FLACS (2010) [4] model of the warehouse to simulate the overpressure generated by ignition of the hydrogen cloud at different ignition delay times. Subscale model experiment and simulation results were compared to validate the modeling approach. The validated model was then used to perform additional full-scale release simulations to provide information for risk-informed hydrogen codes and standards development. Details from the experiments and modeling are described in the following sections.
2.
Scaled warehouse experiment
Model validation experiments were performed in a hardened, subscale warehouse facility located at the SRI International Corral Hollow Experiment Site (CHES) located near Livermore, CA. Fig. 1a shows a photograph of the SRI test warehouse prior to construction of a rigid front wall that was added for the
deflagration experiments. Appropriate scaling factors [5] were determined to create a set of scaled-warehouse tests that resembled as closely as possible the full-scale warehouse release scenario. This scaling approach has been successfully used for previous tests at SRI that involved releases from scaled hydrogen fuel-cell vehicles in road tunnels [6]. The test facility volume and height were respectively (1/2.8)3 and 1/2.8 relative to the 1000 m3 full-scale warehouse with a 7.62 m ceiling height. The scaled warehouse dimensions were 3.64 m wide by 4.59 m long by 2.72 m high, with a total volume of 45.4 m3. A layout of the experimental setup is given in Fig. 1b. The mass of hydrogen released in the scaled test was related to the mass of hydrogen released in the full-scale scenario by the volume ratio, (1/2.8)3. The time for the scaled mass release was related to the time for full-scale mass released by the Froude number and a dimensionless timescale [5]. The transient nature of the hydrogen tank blowdown was modeled with the Sandia developed compressible network flow analysis code, NETFLOW [7,8]. The NETFLOW model accounted for choked flow at the 3.56 mm release orifice, expansion into the enclosure, and exhaust of the hydrogen through the steel perforated plate (51% flow area) on the forklift model side. The initial tank pressure for the experiments was 13.45 MPa and the tank volume was chosen to hold the correct amount of scaled hydrogen mass. The 3.56 mm choked flow orifice with a 0.75 discharge coefficient was selected so that experimental tank blowdown characteristics matched computed subscale values. Fig. 2 shows a comparison of the measured mass release rate from the experiment with the calculated and full- and subscale mass release rates. The forklift model measured 0.343 m long by 0.343 m wide by 0.435 m high and an image is shown in Fig. 3a. These dimensions are consistent with a scaled version of a class 3 “Reach Type” forklift. Hydrogen was released into a 131 mm wide by 131 mm high by 88 mm deep enclosed space on the forklift side facing the back wall of the garage. The release was initiated by actuation of a fast response, pneumatically operated valve (w75 ms activation time) located between the hydrogen storage tank and nozzle orifice. During a release, hydrogen from the nozzle passed through a porous frit and into an enclosure packed with steel beads that occupied w61% of the enclosure volume, and were held in place by a perforated steel plate with a 51% open area (see Fig. 3b). The bed of steel beeds and performated plate diffused the momentum of the choked hydrogen release and ensured a uniform velocity at the grate exit.
Fig. 1 e (a) Photograph and (b) layout of the scaled-warehouse test facility used for release experiments. The model forklift shown with release opening directed toward back wall.
100
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
10-1 10-2 10-3
Flow Rate [kg/sec]
Full Scale, Calculation SRI Scale, Calculation SRI Scale, Measurement
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
Time [sec]
Fig. 2 e Calculated and measured scaled mass release rates along with a comparison to the calculated full-scale mass release rates (based on 0.8 kg of hydrogen forklift release in a 1000 m3 full-scale warehouse).
Three different types of tests were performed in the scaled warehouse experiments: (1) hydrogen dispersion tests to measure the concentration variation with time in the unignited hydrogen cloud released from the model forklift; (2) ignition and flame front imaging tests to determine the shape of the flame front and its propagation speed; and (3) a set of deflagration tests where the overpressure produced from igniting the hydrogen cloud was determined for a given ignition delay time and ignition location. Here the ignition delay time is defined as the time between the beginning of the release and activation of the spark ignition source. Note that to have a consistent comparison of experimental results, tank mass flow rate and warehouse concentration measurements were referenced to the valve opening time; while propagation flame speed, flame imaging, and overpressure measurements were referenced to the spark time. These reference times have likewise been maintained when reporting simulation results in the following sections. The front wall configuration of the warehouse was altered depending on the type of test performed, and the various
8181
configurations are illustrated in Fig. 4. For the flame propagation imaging and dispersion tests, the warehouse front end was either left open (Fig. 1a) or covered with a high-density polyethylene (HDPE) plastic sheet (Fig. 4a), which allowed infrared and standard video visualization of the flame propagation. For the overpressure tests the warehouse open end was sealed with a steel wall that contained a center access doorway covered by a plywood blowout panel (Fig. 4bed). The ignition source was a spark igniter located either 3 cm above the top rear edge of the model forklift (on the side with the release opening grill) or 10 cm below the warehouse ceiling on the same vertical line (the midline of the grill on the side of the forklift). For the dispersion tests, oxygen depletion measurements from fast-response Teledyne UFO 130-2 oxygen sensors were used to measure hydrogen concentrations. Two dispersion tests were conducted within a well-sealed warehouse, while a third was conducted with active mechanical ventilation. The 6.3 m3/min ventilation flow rate was selected based on application of the scaling factors to the NFPA 52 full-scale warehouse ventilation guidelines. Active ventilation was accomplished by placing two muffin fans in the 3.05 m long and 34.0 cm diameter exhaust vent tube located on the back wall of the warehouse (see Fig. 1b). When the fans were running, air was pulled into the warehouse through an open vent (120.0 cm wide by 6.35 cm high) located near the bottom of the plywood pressure blowout panel on the steel front wall of the warehouse (see Fig. 4d). The flow rate through the exhaust tube was verified beforehand with an anemometer along seven radial locations 2.75 m downstream in the tube. Fig. 5 shows an infrared flame propagation image sequence from the scaled warehouse experiment when ignition occurred near the forklift top rear edge (3 s ignition delay time). The flame can be observed traveling upward along the released plume before propagating into the hydrogen/air mixture that had collected along the ceiling. When the release was ignited 10 cm below the warehouse ceiling (image sequence not shown) the flame propagated along the ceiling and was directed downward as it approached the warehouse side-walls. Note, however, that no downward propagation of the flame along the hydrogen release plume was observed. For the confined deflagration overpressure tests, the open end of the warehouse was sealed with a steel wall, and
Fig. 3 e (a) Photograph showing the exterior and release opening for the model forklift used in the experiments. (b) Schematic showing the interior of the model forklift with hydrogen bottle, refilling line, porous frit, steel bead enclosure and perforated steel release opening.
8182
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
Fig. 4 e Photographs of scaled-warehouse test facility showing the different front wall configurations used for the various experiments.
Fig. 5 e Infrared image sequence from the scaled warehouse experiments that illustrate flame front propagation approximately 0, 33, 400, and 800 ms after ignition. The ignition location was 3 cm above the top rear edge of the model forklift and occurred 3 s after the beginning of the release. The hydrogen was released through the back side of the model forklift, directed toward the back wall of the warehouse.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
a plywood blowout panel was used to cover the center access doorway (Fig. 4bed). Seven confined ignition tests were performed, which included several different blowout panel configurations and tests with either active or natural ventilation using the vent tube on the back wall and the vent near the bottom of a plywood pressure blowout panel (Fig. 4d). Pressure transducers were used to measure the generated overpressure, while fast-response thermocouples mounted along the facility ceiling were used to measure horizontal flame front propagation speed. The maximum measured overpressure in the test series was 24.6 kPa, which occurred when the warehouse was wellsealed and the ignition source was located 3 cm above the model forklift. When ignition occurred 10 cm below the ceiling, the maximum measured overpressure was moderately lower at 19.0 kPa. It was observed that small increases in warehouse ventilation area led to significant reductions in deflagration overpressure. Hence, the leakage area of the scaled warehouse was measured for each test configuration by replacing the plywood blowout panel with a computerized fan system, Model 3 Minneapolis Blower Door, manufactured by The Energy Conservatory. For the well-sealed warehouse configuration the leakage area was measured to be approximately 36.7 cm2. The measured exhaust tube leakage area with the muffin fans mounted in place was 197 cm2 and the total measured leakage area for the ventilated warehouse was 972 cm2. These leakage areas were incorporated into warehouse deflagration models for the purposes of comparing the simulated ignition overpressures with the experimental data. It is important to note that the rise in overpressure may have resulted in facility deformation, which in turn would have led to an associated increase in the effective leakage area. Only a brief summary of the experiments has been given here and further details can found in Ekoto et al. [9].
3. Scaled warehouse simulations and model validation The computational fluid mechanics code, FUEGO [3], was used to perform simulations of hydrogen fuel-cell forklift releases inside the scaled warehouse. FUEGO is a Sandia National Laboratories’ developed code designed to simulate turbulent reacting flow and heat transfer [3] on massively parallel computers, with a primary focus on heat transfer to objects in pool fires. The code was recently adapted for compressible flow and hydrogen combustion. The discretization scheme used in FUEGO is based on the control volume finite element method [10], where the partial differential equations of mass, momentum, and energy are integrated over unstructured control volumes. A standard two equation (ke3 ) turbulence model was applied to close the Favre averaged equations [11]. Transport equations were solved for the mass fractions of each chemical species, except for the dominant species (nitrogen in the present simulations) which was computed by constraining the sum of the species mass fractions to equal one. For compressible flow, the ideal gas equation of state was used to relate the density and pressure of the gas mixture. For the calculations reported here, the first order upwind scheme was used for the convective terms. The code solved the
8183
conservation equations in a time-dependent manner. A validation study that assesses the model’s ability to predict velocity and concentration decay along the centerline of unignited hydrogen free jets and the centerline temperature for laboratory-scale hydrogen jet flames is reported in Houf et al. [12]. FUEGO dispersion simulation boundary conditions for the floor and walls of the warehouse enclosure were no-slip and constant temperature (294 K) as were the boundary conditions on the forklift sides. The vent in the back wall of the enclosure was assigned an open boundary condition with a total pressure of 1 atm, a temperature of 294 K, oxygen and nitrogen mole fractions of 0.21 and 0.79 respectively, and negligible levels of turbulence (k ¼ 0.11 cm2/s2, 3 ¼ 1.51 e4 cm2/s3). For simulations involving forced ventilation, the ventilation mass flow rate was used to specify an inflow boundary condition on the inlet vent area in the front wall of the warehouse. Transient jet inflow boundary conditions for the hydrogen exiting the grill on the side of the forklift (velocity, temperature, concentration) were computed with NETFLOW [7,8] and input as transient inlet boundary conditions on the side of the forklift in the warehouse model. A grid resolution study was performed to insure that the concentration results were independent of mesh spacing and time step. In addition, a total mass balance on the amount of hydrogen in the enclosure and exiting the vents was performed and compared with the total mass of hydrogen released from the forklift. A FLACS [4] model was used to perform transient simulations of hydrogen gas cloud ignition along with the associated overpressure generated by the rapid deflagration wave propagation and expansion of hot combustion product gases inside the enclosure. FLACS solved the compressible NaviereStokes equations on a three-dimensional Cartesian grid using a finite volume method. Conservation equations for mass, momentum, enthalpy, turbulence, and species were solved in conjunction with the ideal gas law equation of state using the SIMPLE [13] pressure correction algorithm extended for compressible flow. The numerical model used a second-order scheme for diffusive fluxes and a second order “kappa” scheme (hybrid scheme with weighting between second order upwind and second order central differences, with delimiters for some equations) to resolve the convective fluxes. Although the model contained a second order scheme for time, it was not used for the deflagration calculations due to the requirement of short time steps for stability. A standard ke3 model was used for turbulence with modifications for turbulence generation behind sub-grid objects. In the simulation, FLACS treated the deflagration wave as a collection of flamelets, where the burning velocity for each was computed based on the following three submodels: (a) a laminar burning velocity model that describes the laminar burning velocity as a function of gas mixture, concentration, temperature, pressure, oxygen concentration in air and the amount of inert diluents, (b) a model describing quasi-laminar combustion in the first phases of flame propagation after ignition, and (c) a model that describes the turbulent burning velocity as a function of turbulence parameters such as the intensity and length scale. A roughly cubic grid was used for all FLACS deflagration calculations with a mesh spacing of w10 cm. A number of grid resolution studies were performed to ensure the calculations
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
were independent of mesh size. The basic equations used in the FLACS model are described in Hjertager [14,15] and validation of the model against experimental data for hydrogen deflagration can be found in Middha and Hansen [16]. High-pressure hydrogen gas was exhausted from the forklift side grating through a 131 mm by 131 mm opening with a uniform exit velocity and was directed toward the back wall of the warehouse. In the forced ventilation simulations the FUEGO scaled warehouse model included the open area where the exhaust tube was located in the warehouse back wall, and the lower vent located at the bottom of the plywood blowout panel. Fig. 6 shows a still frame, 3 s into the release, from the simulation of the transient hydrogen cloud in the scaled warehouse for a case without active ventilation. From the image, the hydrogen plume can be observed rising from the forklift side vent and collecting along the ceiling. Initial conditions for the FLACS simulations were based on threedimensional concentrations maps extracted from the FUEGO dispersion simulations at the time of ignition. The FLACS model included the ventilation openings in the back and front walls and additional models were built that allowed simulations to be performed with either an open warehouse front wall or the plywood blowout panel. The small warehouse leakage areas measured with the blower apparatus as part of the experiments were likewise incorporated into the model by adding a small slot of equivalent leakage area near the bottom center of the front wall. Pretest FUEGO/FLACS simulations of the scaled warehouse were performed prior to the experiments to determine the appropriate placement of concentration sensors and to estimate the expected overpressure from ignition of the hydrogen releases. Fig. 7 summarizes the predicted peak overpressures for four separate warehouse configurations, with ignition located near the forklift model, as a function of ignition delay time. For all conditions with the exception of the inclusion of passive ventilation, the peak overpressure occurred after a 3 s ignition delay, with a corresponding maximum deflagration overpressure predicted to be roughly 30 kPa if the warehouse
0.35 Closed Front Wall (Length - 4.59 m)
0.30 Overpressure (barg)
8184
Closed Front Wall (Length = 6.10 m)
0.25 0.20
Closed Front Wall Vent in Back Wall (30 cm x 30 cm)
0.15 0.10 Open Front Wall
0.05 0.00
0
1
2
3
4
5
6
7
Ignition Delay Time (sec) Fig. 7 e Pretest FLACS simulation results of the maximum deflagration overpressure as a function of ignition delay time for different scaled warehouse test configurations. Note that these pretest results neglect wall heat transfer and effective leakage area.
was completely sealed and the effects of wall heat transfer were neglected. Accordingly, the 3 s ignition delay time was selected for the experiments, and it was decided that a pressure relief blowout panel should be included in the experimental setup to prevent unintended damage to the test facility. A comparison of predicted mole fractions relative to experimental measurements, both with and without forced ventilation, is given in Fig. 8 at four sensor locations. Close agreement is exhibited between the simulation and experimental results. Importantly, both the experimental and simulation datasets indicated the effect of forced ventilation on the hydrogen concentration profiles is small for this leak size.
1 prediction S01 no ventilation prediction S01 ventilation data S01 1026 data S01 1027 prediction S05 no ventilation prediction S05 ventilation data S05 1026 data S05 1027 prediction S04 no ventilation prediction S04 ventilation data S04 1026 data S04 1027 prediction S08 no ventilation prediction S08 ventilation data S08 1026 data S08 1027
H2 mole fraction
0.8 0.6 0.4 0.2 0 0 Fig. 6 e Simulation snapshot of the flammable volume of hydrogen (4%e75% mole fraction) in the scaled warehouse experiment 3 s into the release for the condition without ventilation, along with the location of the concentration sensors.
1
2
3
4
5 6 time (s)
7
8
9
10
Fig. 8 e Comparison of measured and predicted concentrations with and without forced ventilation in the scaled warehouse at four different sensor locations (test 1026 performed without active ventilation and test 1027 performed with active ventilation).
8185
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
a
b
30
5 Simulation (Natural Ventilation) Simulation (Forced Ventilation)
Simulation - Sealed (No Heat Transfer)
4
Simulation Sealed (Heat Transfer) 2
20
Simulation - 36.7 cm Leak Area (No Heat Transfer)
15 Data
10
2
Simulation - 36.7 cm Leak Area (Heat Transfer)
Data (Natural Ventilation) Data (Force Ventilation)
3
2
1
5 0
Overpressure (kPa)
Overpressure (kPa)
25
0
2
4 6 Time (sec)
8
10
0
0
0.5
1
1.5 2 2.5 Time (sec)
3
3.5
4
Fig. 9 e Comparison of simulation and experimental data of deflagration overpressure for the well-sealed subscale warehouse with ignition 3 cm above the forklift for the case with (a) no ventilation where the simulations progressively include the effective leakage area, wall heat transfer, or both effects; and (b) either forced ventilation (w6.3 m3/min) or natural ventilation (ventilation fans off) through the vents. Note that the reported time is relative to the spark ignition.
The experimental deflagration overpressure and decay rate within the scaled warehouse was dependent of the rigidity of the enclosure and how well it was sealed. Several different strength blowout panels were examined to investigate the impact on overpressure characteristics once they were breached. Initial blowout panels were built from a single plywood sheet (1.27 cm thick) bolted to the steel wall. A wooden frame (5.08 cm by 10.16 cm thick) was used to increase attachment strength to the facility (see Fig. 4b). For the well-sealed warehouse results shown in Fig. 9a where no breach was desired, the number of plywood sheets was increased to 3. Metal Unistrut was then mounted across the front surface of the plywood as shown in Fig. 4c to further increase the panel’s strength and to minimize panel bending. This reinforced panel did not breach and remained relatively well-sealed for the test duration. Fig. 9a shows a comparison of predicted deflagration overpressure from the simulations with data from a pressure
a
sensor located in the center of the warehouse sidewall near the forklift and positioned 1.24 m off the floor. Simulation results are shown with and without wall heat transfer, both for the completely sealed warehouse and for the warehouse with the measured effective leakage area (36.7 cm2) included. The near forklift ignition point was used for these simulations. Once wall heat transfer and the measured effective leakage rate are included into the simulations, the results exhibit excellent agreement with the experimental data in both the magnitude and shape of the pressure profiles. Although the pressure decay from the peak value was somewhat slower than was observed in the measurements, facility deformation that resulted from large stresses due to the deflagration pressure wave likely increased the effective leakage area, and hence increased the pressure relief. Note also that while the presented results are relative to the spark time, there was an initial w1 kPa pressure rise prior to ignition that was observed for both the experimental and simulation results, and was
b
25
Simulation (Well Sealed)
Overpressure (kPa)
20
Data (Well Sealed)
15
Simulation (Blowout Panel Breach)
10 5
Data (Blowout Panel Breach)
0 -5
0
1
2
3
4 5 Time (sec)
6
7
8
Fig. 10 e (a) Single frame of a high-speed video showing breach of plywood blowout panel on front of the scaled warehouse after ignition of hydrogen release from model forklift. (b) Comparison of simulation and experimental data for a well-sealed warehouse with and without plywood blowout panel breach (3 s ignition delay time). Ignition location is 3 cm above the forklift and time is referenced to the spark ignition.
8186
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
25 Simulation (Ign. above forklift)
Overpressure (kPa)
20
15 Simulation (Ign. 10 cm from ceiling)
10
Fig. 12 e Simulation results that show the time evolution of flammable hydrogen volumes (4%e75% mole fraction) that result from the release of 0.8 kg of hydrogen into the fullscale warehouse for cases with and without forced ventilation.
Data (Ign. 10 cm from ceiling)
5 Data (Ign. above forklift)
0
0
1
2
3
4
5
6
Time (sec) Fig. 11 e Comparison of simulation results and experimental data for ignition 3 cm above forklift (3.0 s ignition delay time) and ignition 10 cm from warehouse ceiling (3.5 s ignition delay time). Note that the simulations include the 36.7 cm2 effective ventilation area and wall heat transfer.
attributed to the rapid expansion of compressed hydrogen into the enclosed volume. Fig. 9b compares simulation and experimental deflagration overpressure data for the scenarios with either forced (ventilation fans on) or natural ventilation (ventilation fans off) through the warehouse vents. From the experiments, overpressure for the case with forced ventilation was approximately 3 kPa while the case with natural (passive) ventilation was slightly higher at 4.2 kPa. Note that the difference in overpressure was probably within the experimental variability of the tests. Nonetheless, predicted overpressure profiles were similar in magnitude and shape, with the roughly1 kPa difference in the conditions with and without mechanical ventilation preserved. The only major difference of note was that the simulation pressure peak was shifted w0.4 s earlier in time. These results demonstrate that the impact of forced ventilation on the peak deflagration overpressure remained small. It is also worth noting that the initial overpressure increase observed for the well-sealed conditions due to the hydrogen release was not observed in either the experimental or simulation results for the conditions with
ventilation. Further details and discussion about this phenomenon can be found in Ekoto et al. [9]. A simulation of the breach of the pressure relief panel was modeled by incorporating a panel of the same size and location as the plywood blowout panel into the FLACS model. In the model the pressure relief panel was set to open at the same overpressure where the experimental data indicated a breach occurred (w13 kPa). Fig. 10a shows an image frame from a high-speed movie taken of the single sheet plywood blowout panel during breach. Fig. 10b compares simulation and experimental overpressure data for the well-sealed warehouse with and without breach of the plywood pressure blowout panel. Results for the simulation and experiment again exhibited good agreement, with Helmholtz pressure wave oscillations observed in the pressure data for both conditions, although the ringing intensity was higher in the experiments. These results indicate that pressure relief panels can provide an immediate and effective means to reduce deflagration pressure rise, and the simulations are able to capture the characteristic ringing behavior that results once these panels breach. The final comparison was for the well-sealed warehouse condition with the ignition point located 10 cm below the warehouse ceiling. For this condition, the ignition delay was increased from 3.0 to 3.5 s to ensure a sufficiently flammable hydrogen/air mixture at the release point was present. Results are compared in Fig. 11, and the near forklift overpressure data is included for comparison. The experimental overpressure data exhibited a roughly 5.6 kPa reduction in the peak value relative to the near forklift ignition condition. The associated reduction for the simulations, however, was more modest at around 1.5 kPa. The disagreement may be due to the
Table 1 e Comparison of simulation predicted peak deflagration overpressure from transient hydrogen releases in scaled and full-scale warehouses for a 3.0 s ignition delay time and ignition 3 cm above the top of the forklift.
Scaled-warehouse (S.F. ¼ 1/2.8) Full-scale warehouse
100% Sealed (no heat transfer)
100% Sealed (heat transfer to walls)
Natural ventilation (vents open)
Active ventilation (vents open with forced flow)
29.8 kPa 27.2 kPa
26.7 kPa 24.5 kPa
4.9 kPa 14 kPa
4.2 kPa 15.5 kPa
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
8187
Fig. 13 e Simulation results of the flammable hydrogen cloud (4%e75% mole fraction) 3 s into the 0.8 kg hydrogen release for the full-scale warehouses with a 7.62 m high ceiling and volumes of 1000 m3, 1500 m3, and 2000 m3 (without ventilation).
fact that a single sheet plywood blowout panel was used for the ceiling ignition test, while a more rigid 3-sheet plywood blowout panel was used for the tests where ignition occurred 3 cm above the forklift. The single sheet plywood panel may have undergone greater deformation and allowed a greater amount of leakage during the pressure rise period. Another possible explanation is that the FLACS simulation for the ceiling ignition condition indicates that the flame propagated down along the release plume toward the model forklift. This result contradicts the observation from the infrared flame visualization that downward flame propagation along the release plume did not occur. The FLACS flame propagation model appeared to not distinguish between upward and downward propagating hydrogen flammability limits, but instead assigned a uniform value regardless of propagation direction. Although the generally accepted volumetric concentration for the upward-propagating lower flammability limit of hydrogen in air is 4%, experimental data in the literature indicate that the limit may be as high as 7.2% for horizontally-propagating flames and 9.5% for downwardpropagating flames [17,18].
4.
Full-scale warehouse simulations
Full-scale forklift-in-warehouse release simulations were performed with the FUEGO/FLACS models to evaluate
whether the full-scale warehouse release and deflagration would behave in the same manner as observed in the scaled warehouse experiments. The geometry of the scaled SRI warehouse, including the forklift and its location, was converted to a full-scale warehouse model by multiplying all of the lengths in the model by a factor of 2.8. This full-scale warehouse had a ceiling height of 7.62 m and was 12.86 m long by 10.20 m wide with a volume of 1000 m3. A total of 0.8 kg of hydrogen was released from the forklift tank at 35 MPa through a 6.35 mm opening into a 0.118 m3 enclosure with a vent open area on the forklift side equal to 674 cm2. The NETFLOW compressible network flow analysis code was again used to model the transient nature of the tank blowdown assuming a 50% blockage by the grill (see Fig. 2); the results were applied for the FUEGO boundary condition at the exhaust opening on the forklift side. Full-scale front and back wall vents were incorporated into the model, while active and passive (natural) ventilation, along with completely sealed scenarios were also examined. For the active ventilation simulations the full-scale ventilation rate of 0.03 m3/min/0.34 m3 of room volume, as specified in NFPA 52 [1], was used. Table 1 shows comparisons of simulated peak ignition deflagration overpressure in the scaled and full-scale warehouse releases for a 3 s ignition delay time. For the well-sealed warehouse conditions the peak ignition deflagration overpressure with wall heat transfer for the full-scale warehouse
Fig. 14 e Simulation results of the flammable hydrogen cloud (4%e75% mole fraction) 6 s into the 0.8 kg hydrogen release for the full-scale warehouses with a 7.62 m high ceiling and volumes of 1000 m3, 1500 m3, and 2000 m3 (without ventilation).
8188
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
was 24.5 kPa as compared to 26.7 kPa in the (1/2.8) scaled warehouse geometry. The results for the full-scale warehouse show that forced flow ventilation had little effect on deflagration overpressure as compared to the case where the vents were left open without forced flow. A similar effect of forced ventilation on the deflagration overpressure for the scaled warehouse experiments was observed (see Fig. 9b). Profiles of the predicted facility flammable volumes, defined as the total volume of flammable hydrogen/air mixtures, are depicted in Fig. 12 for the cases with passive ventilation and mechanical ventilation at the rate specified by NFPA 52 (6.3 m3/min) [1]. These results indicate that 3 s into the release there was no discernible difference in flammable volumes, and only a modest 10% difference was predicted by 16 s into the simulation. A series of full-scale warehouse simulations were also performed for the 0.8 kg forklift release, where the warehouse volume was increased from 1000 m3 to 1500 m3 and 2000 m3, while keeping the warehouse ceiling height constant at 7.62 m. The location of the forklift remained midway between the front and back walls of the warehouse and at a fixed distance of 3.04 m from the nearer sidewall in all cases. Figs. 13 and 14 show the flammable hydrogen clouds (4%e75% mole fraction) for a 0.8 kg forklift release in warehouses of volumes 1000 m3, 1500 m3, and 2000 m3 with a 7.62 m ceiling with and without ventilation at 3 s and 6 s after the start of the release, respectively. Fig. 15 shows the predicted peak overpressure for ignition of these clouds at ignition delay times of 3 s and 6 s. The results indicate that the peak deflagration overpressure decreased almost linearly with increased warehouse volume.
5.
Summary and conclusions
Simulation results of an unintended release from a hydrogen fuel-cell forklift vehicle in a full-scale warehouse have been performed for the case where hydrogen gas is vented from the side of the forklift vehicle as a result of a medium sized leak or thermal activation of a pressure relief device. The same modeling approach used in the full-scale warehouse release was validated in a scaled model by comparing simulation results with measured data from a series of scaled-warehouse release experiments performed at the SRI Corral Hollow Experiment Site. Results of the simulations were found to be in good agreement with the experimental data. The results of the modeling and experiments demonstrated that passive natural ventilation or pressure relief panels can be an effective means to mitigate deflagration overpressure arising from ignition of hydrogen plumes created by medium- and large-scale transient releases. However, the impact of forced ventilation had little effect on hydrogen concentration or deflagration overpressure in the early stages of the currently sized transient release scenario. Full-scale simulation results further indicated that increased warehouse volume beyond current requirements resulted in a linear reduction in peak overpressure.
Acknowledgments This work was supported by the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, Fuel Cell Technologies Program under the Safety, Codes, and Standards subprogram element managed by Antonio Ruiz.
references
Fig. 15 e Simulation results that show the peak ignition overpressure in a well-sealed full-scale warehouse for a 0.8 kg hydrogen release where ignition occurs 3 cm above the top of the forklift and the ceiling height is 7.62 m. Results include heat transfer to the warehouse walls and show the effect of changing the ignition delay time and warehouse volume.
[1] NFPA 52 Vehicular Gaseous Fuel Systems Code 2010 Edition, NFPA, 1 Batterymarch Park, Quincy, MA 02169e07471. [2] NFPA 2 Hydrogen Technologies Code 2011 Edition, NFPA, 1 Batterymarch Park, Quincy, MA 02169e07471. [3] Moen CD, Evans GH, Domino SP and Burns SP. A multimechanics approach to computational heat transfer. In: proceedings 2002 ASME Int. Mech. Eng. congress and exhibition. New Orleans, IMECE2002e33098: November 17e22, 2002. [4] FLACS version 9.1 user’s manual. Bergen, Norway: GEXCON; 2010. [5] Hall DJ, Walker S. Scaling rules for reduced-scale field releases of hydrogen fluoride. J Hazard Mater 1997;54:89e111. [6] Houf WG, Evans GH, Merilo EG, Groethe MA and James SC. Simulation of hydrogen releases from fuel-cell vehicles in tunnels. In: 18th world hydrogen energy conference. Essen, Germany: May 16e21, 2010. [7] Winters WS. A new approach to modeling fluid/gas flows in networks. Sandia National Laboratories; July 2001. Report SAND2001-8422. [8] Winters WS. Implementation and validation of the NETFLOW porous media model. Sandia National Laboratories; October 2009. Report SAND2009-6838. [9] Ekoto IW, Merilo EG, Houf WG, Evans GH and Groethe MA. Experimental investigation of hydrogen release and ignition from fuel cell powered forklifts in enclosed spaces. Int J hydrogen energy, (2012), http://dx.doi.org/10.1016/j.ijhydene. 2012.03.161.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 8 1 7 9 e8 1 8 9
[10] Minkowycz WJ, Sparrow EM, Schneider GE, Pletcher RH. Elliptic systems: finite element method 1, handbook of numerical heat transfer. New York: J. Wiley and Sons, Inc; 1988. p. 379e420. [11] Papageorgakis GC, Assanis DN. Comparison of linear and nonlinear RNG-based k-epsilon models for incompressible turbulent flows. Numer Heat Transfer, Part B 1999;35:1e22. [12] Houf WG, Evans G, Schefer RW. Analysis of jet flames and unignited jets from unintended releases of hydrogen. Int J Hydrogen Energy 2009;34:5961e9. [13] Patankar SV. Numerical heat transfer and fluid flow. Hemisphere Pub; 1980.
8189
[14] Hjertager BH. Computer-simulation of turbulent reactive gas-dynamics. Model Ident Control 1984;5:211e36. [15] Hjertager BH. Three-dimensional modeling of flow, heat transfer, and combustion, handbook of heat and mass transfer. Houston: Gulf Publishing Company; 1986. p. 304e50. [16] Middha P, Hansen OR. Using computational fluid dynamics as a tool for hydrogen safety studies. J Loss Prevent Proc 2009; 22:295e302. [17] Coward HF, Jones GW. Limits of flammability of gases and vapors. Bur Mines Bull 1952;503. [18] Zebetakis MG. Flammability characteristics of combustible gases and vapors. U.S. Bur Mines Bull 1965;627.