Presentation is loading. Please wait.

Presentation is loading. Please wait.

Zaki S. Saldi, Jennifer X. Wen Warwick FIRE, School of Engineering

Similar presentations


Presentation on theme: "Zaki S. Saldi, Jennifer X. Wen Warwick FIRE, School of Engineering"— Presentation transcript:

1 MODELING THERMAL RESPONSE OF POLYMER COMPOSITE HYDROGEN CYLINDERS SUBJECTED TO FIRE
Zaki S. Saldi, Jennifer X. Wen Warwick FIRE, School of Engineering University of Warwick The International Conference on Hydrogen Safety Yokohama, October 2015

2 Background California Fuel Cell Partnership cafcp.org Zero carbon emission from transportation  hydrogen energy  HFCV. Low volumetric energy density H2 (5.6 MJ/L)  compressed storage, carbon fibre reinforced composite (CFRP) tank, MPa. Type-3 (aluminum liner), type-4 (HDPE plastic liner). Slow venting of H2 through TPRD required in the event of accident with fire (GB: road vehicle fires in ). Fire resistance: minutes (type 4) (Weyandt, 2006), 12 minutes (type 3) (Zalosh & Weyandt, 2005), too short.

3 Background Brennan & Molkov (2011):
Venting within 3 min requires TPRD diameter of 5 mm. At 35 MPa, the corresponding flow rate is about 390 g/s (overpressure kPa, garage destroyed in less than 2 s) In an open space, 70 MPa & 5 mm: “no harm” separation distance from the car is about 50 m. Need for enhanced fire resistance ( > 1 hour) This illustrates the serious and more severe hazards posed by hydrogen release as compared with hydrocarbon gases. Given a certain venting time based on the cylinder fire resistance (suppose the resistance is only 3 minutes), with typical TPRD diameter of 5 mm and inner pressure 35MPa, the flow rate is 390 gr/s, leading to catastrophic overpressure (10-20 kPa in 2 s). At 70MPa & 5 mm diameter, “no harm” separation distance is impractical, about 50 m. Fire resistance of preferably more than 1 hour would ensure slower venting and thus considerably much less severe hazards as well as timely evacuation by first responders.

4 Objectives To develop a thermo-mechanical model with decomposition & degradation mechanism for calculating the response of composite cylinder under thermal and pressure loadings. To use experimentally obtained physical-chemical properties in the model. To use the model for chosen fire testing scenarios & to aid in the design of cylinders with enhanced fire resistance.

5 Outline Background & objectives Numerical models Results Summary
CFD model for external fire FE model for cylinder structure Results H2 cylinder subjected to propane fire (burst testing) Summary

6 Numerical models CFD Heat Flux FE
Mouritz et al (2009) Computational Fluid Dynamics for fire. Finite Element simulation for cylinder thermo-mechanics (heat transfer, decomposition, degradation). One-way coupling through heat flux from fire (CFD) to cylinder (FE).

7 CFD model for external fire
Combustion: extended Eddy Dissipation Concept. Finite Volume, Large Eddy Simulation, FireFOAM (OpenFOAM). Wang, C.J. , Wen, J.X. and Chen, Z.B. 'Simulation of Large-Scale LNG Pool Fires Using FireFOAM' Combustion Science and Technology, 2014 Wang, C.J. , Wen, J.X. and Chen, Z.B. , Dembele, S. 'Predicting radiative characteristics of hydrogen and hydrogen/methane jet fires using FireFOAM' International Journal of Hydrogen Energy, 2014

8 CFD model for external fire
Extended EDC model: Original version: Magnussen (1989) Currently used: modification of total turbulent kinetic energy & dissipation rates to include subgrid scale quantities to accommodate LES (Chen, 2011). CD1 = 0.135, CD2 = 0.5, kSGS: one-equation linear eddy model (Menon, 1996) Fine structures assumed as stationary homogeneous perfectly stirred reactor. Reaction rates depend on: mass transfer rates with surrounding fluids, mass fraction of the fine structures, reacting fraction.

9 FE model for H2 cylinder Energy equation Source terms
- Gas convection: - Time rate of change due to decomposition: - Enthalpies of composite & gas: Decomposition model based on: Henderson, J., Wiecek, T., “A Mathematical Model to Predict the Thermal Response of Decomposing, Expanding Polymer Composites”, Journal of Composite Materials, Vol. 4, pp , 1987. Implemented in Elmer, open source FE code.

10 FE model for H2 cylinder A Model for the Thermal Response of Polymer Composite Materials with Experimental Verification, J.B. Henderson, J.A. Wiebelt and M.R. Tant, Journal of Composite Materials 1985, 19: 579

11 Results Problem under investigation:
H2 cylinder in propane fire  Fire exposure burst test, Zalosh & Weyandt (2005). Mesh sensitivity analysis Effects of radiation sub-model Effects of heat transfer coefficient at cylinder wall Decomposition of CFRP cylinder Initial estimate of cylinder fire resistance.

12 H2 cylinder, propane fire
Experiment: Zalosh R., and Weyandt N., Hydrogen Fuel Tank Fire Exposure Burst Test, SAE paper number , 2005. Type-4 composite cylinder Initial pressure 34.3 MPa. Propane flowrate scfh. HRR ~ 370 kW (95% burning eff). Rupture time 6 min 27 s, internal pressure at rupture = 357 bar.

13 H2 cylinder, propane fire
Hexahedral, unstructured mesh snappyHexMesh OpenFOAM tool Three meshes for mesh sensitivity study: ~0.7M, ~1.7M, ~3.4M

14 H2 cylinder, mesh sensitivity
In order to study the mesh sensitivity analysis, we monitor the heat power coming into the cylinder for three different meshes. We monitor this parameter, as in this CFD study we are mainly interested only in heat transfer (heat flux) from the fire to the cylinder, as this will be used for boundary condition in the subsequent FE model to predict thermal response of the cylinder. Although the mean values of the heat flux (at t = 6s, top figure) seem to differ only a little for the three meshes, the incoming heat power (surface integral of the instantaneous heat flux, bottom figure) is insensitive to changing the mesh from 1.7M to 3.4M, whereas the difference is clear between 0.7M and 1.7M. Accordingly, we use 1.7M in our further analysis.

15 Radiation sub-model Radiation is modeled by solving the radiative heat transfer equation using fvDOM. Two sub-models for absorption & emission properties of water vapour and carbon dioxide: Grey mean absorption coefficient. Weighted sum of grey gas. As the heat transfer in fire is dominated by radiation, we also study the influence of different radiation submodels on the heat transfer to cylinder. Here, we used two radiation submodels to calculate the emission and absorption coefficients of CO2 and H2O. The submodels are grey mean absorption model (GM) and weighted sum of grey gas (WSGG), respectively. The influence of radiation submodels on the fire is shown by the time evolution of maximum temperature and radiative heat fraction (top). It can be seen that the grey mean model leads to a higher radiative heat fraction and lower maximum temperature. The difference in maximum temperature between both submodels is around 250 K, corresponding to a difference in radiative heat fraction of around Unfortunately, the maximum temperature and/or radiative fraction in the burst experiment are unknown, so we can not validate the radiation submodels in terms of their influence on the fire thermal fields. However, we are only more interested in the heat transfer from the fire to the cylinder. From bottom figure, we can see the difference in the mean heat flux at cylinder wall predicted using the two radiation submodels. There is a difference of around 2 kW/m2 in the maximum mean heat flux, but in general the mean heat flux predicted using WSGGM can be 3-4 kW/m2 higher. Since the heat flux predicted using WSGG model is higher, we decided to employ this submodel for our further analysis.

16 Heat transfer coefficient
Heat flux at cylinder wall: h [W/(m2K)] Ref. Mean Qin, wall [kW] Mean heat flux max [kW/m2] 13 Ramroth, 2006 8.46 14.27 50 - 18.93 36.22 80 (*) Rafi & Nadjai, 2012 21.9 47.04 One of the current limitations of employing a one-way coupling between CFD and FE is that we rely on the above boundary condition on the cylinder wall. Using this bc, we treated the wall as if it acts as a heat sink operating at a certain rate, which is controlled by parameter h (heat transfer coefficient). The value for h is uncertain. From literature, for CFRP in hydrocarbon fire we found two suggested values (13 and 80, as listed in the table), so we investigated the influence of changing this parameter on the predicted heat flux. As expected, incoming heat power and heat flux increase with the heat transfer coefficient. The coefficient of 13 suggested by Ramroth represents only convection, whereas that suggested by Rafi and Nadjai (2012) (h = 80 W/(m2K)) represents combination of both convective and radiative heat transfer from fire, and gives the highest maximum mean heat flux among the three coefficients used. This value is used further in the analysis. (*) Overall coefficient that combines convective and radiative heat transfer, used in further analysis.

17 H2 cylinder, propane fire
The movie on the left shows the isocontour of T = 1200 K, on the right it show the mid-plane cut of temperature field. The flame seems to have developed at around 3-4 s, where the flame height reaches quasi-steady state. This simulation is for mesh 1.7M, h=80, and with WSGG radiation submodel.

18 H2 cylinder decomposition
Around 50% of CFRP decomposed after 600 s Using the predicted mean heat flux from the previous movie, we performed the FE analysis using Elmer on the cylinder thermal response. For the simulation of the thermal response of the cylinder, circumferential slice at the where maximum heat flux is located is selected. This is done in order to afford the 2-D axisymmetric simulation, instead of a full 3-D simulation of the thermal field, aimed at saving computational time. The thermal simulation was carried out until final time t = 600 s. The temperature fields at t = 120, 300, and 600 s are shown (left figure). The heating by the external propane fire also leads to the decomposition of the CFRP, as represented by the density rate of change (right figure). Once the decomposition temperature onset is reached ( K), the decomposition reaction starts to consume continuously the CFRP resin. The density ratio fields, i.e. the ratio between the instantaneous CFRP density and the initial density, are also shown in Figure 10. At t = 600 s, about half of the CFRP in the through-thickness direction has been completely decomposed. Unfortunately, there is no quantitative data from Zalosh burst experiment to compare with, apart from the cylinder fire resistance (burst time). Ideally speaking, it would have been nice if we also know the following: Heat flux and/or temperature distribution at cylinder wall Decomposition field

19 H2 cylinder fire resistance
Hu et al, IJHS, 2008 Fire resistance (initial estimate) based on internal pressure (function of temperature). (Deming WE, Shupe LE. Some physical properties of compressed gases, III. Hydrogen. Phys Rev 1932;40:848–59  covering -2150C < T < 5000C and p up to 1200 atm). Internal pressure at failure time in experiment (Zalosh, 2005): 357 bar. Predicted fire resistance: 399 s (6 mins 39 s), Zalosh experiment: 6 mins 27 s. For comparison sake, we also predicted fire resistance, as this is known from Zalosh experiment. In the experiment, cylinder burst at 6 min 27s, where the internal pressure is 357 bar. We assume that fire resistance is the time where burst happens. Accordingly, we monitor the increase of internal pressure. The internal pressure is calculated based on the temperature at the liner-hydrogen interface, using p-T relation taken from the above paper on data on compressed hydrogen. We obtain that the internal pressure of 357 bar is reached at 6 mins 39 s, close to the value reported in Zalosh experiment.

20 Summary LES simulation of fire using FireFOAM.
One way coupling between CFD (fire) & FE (cylinder) through mean heat flux predicted by CFD. Around 50% of tank CFRP decomposed after 600 s. Type-4 cylinder in propane fire (hydrogen tank fire exposure burst test, Zalosh & Weyandt, 2005): Fire resistance based on initial prediction using internal pressure (6 min 39s) closely matches experimental value (6 min 27s). More validation needed (e.g. heat flux). Ongoing works on thermo-mechanical FE model implementation in Elmer to predict stress distribution, structure failure index using progressive failure criteria.

21 ご清聴ありがとうございます Acknowledgements EPSRC Ref EP/K021109/1
Dr D. Makarov, Prof. V. Molkov (University of Ulster) & Dr T. Mays (University of Bath). Advisory board. ご清聴ありがとうございます


Download ppt "Zaki S. Saldi, Jennifer X. Wen Warwick FIRE, School of Engineering"

Similar presentations


Ads by Google