Final Report Summary - FIRECOMP (Modelling the thermo-mechanical behaviour of high pressure vessel in composite materials when exposed to fire conditions)
Executive Summary:
The use of hydrogen presents an opportunity for a carbon-neutral fuel source. There are a number of challenges that need to be overcome in order to exploit the benefits of hydrogen at large scale, in particular the storage of hydrogen must be secured. Successful large scale storage of hydrogen has been managed at high pressure with the use of composite vessels. Under normal working conditions a burst in service of a composite vessel is very unlikely, however when exposed to a fire the integrity of the composite material can be compromised and presents safety challenges. Thus research into protection strategies for the vessels in fire conditions is needed to combat the potential for a burst and limit hazardous the potential effects of hydrogen release.
The main objective of the FireComp project has been to better characterize the conditions that need to be achieved to avoid burst of composite vessels for CGH2 storage. To this aim, experimental work has been done in order to improve the understanding of heat transfer mechanisms and the loss of strength of composite high-pressure vessels in fire conditions coupled with modelling of the thermo-mechanical behaviour of the vessels. Both stationary and mobile applications have been considered including; automotive application, transportable cylinders, bundles and tube trailers.
Key activities undertaken during the FireComp project:
• Development of a risk methodology and proposed an approach for comparison with metallic cylinders.
• Modelling studies and large-scale experimental campaigns
• Comparison between experiments and modelling resulting in a good correlation with respect to time to burst.
• Development of a methodology to establish a safety pressure curve
• Development of modelling criterions for leak or burst scenarios
Scientific dissemination of the research has been achieved through presentations at conferences such as IChemE Hazards 26, ICHS and WHEC and public webinars.
Dissemination towards Regulations, Codes and Standards (RCS) has been achieved through the relationships established with experts from ISO TC58/SC3/WG24 committee and the new ISO Technical Report TR 13086-2 on topics regarding the qualification of high-pressure composite storage and sizing of its protections. FireComp has presented an excellent possibility to contribute to updating RCS on fire safety.
Project Context and Objectives:
Today’s challenge
Hydrogen is expected to be highly valuable energy carrier for the 21st century. To exploit its benefits at large scale, further research and technological developments are required. In particular, the performance of the storage of hydrogen under high pressure must be assessed. Keeping the integrity of composite pressure vessels in case of fire is necessary to protect people, infrastructure and the environment. More efficient protection strategies are then required to increase societal acceptance and consolidate the hydrogen energy business.
Better characterization of the conditions for safety.
The main objective of FireComp project was to better characterize the conditions that need to be achieved to avoid burst of composite cylinders in case of fire.
• Experimental work has been carried out to improve the understanding of heat transfer mechanisms and the loss of strength of composite high-pressure vessels in fire conditions.
• Then the modelling of the thermo-mechanical behaviour of these vessels has been set up. The model has been validated by full scale fire tests.
Different applications have been considered: automotive application, stationary application, transportable cylinders, bundles and tube trailers.
Perspectives of FIRECOMP project
Outcomes of the project are recommendations for the Regulation Codes and Standards regarding the qualification of high-pressure composite storage and sizing of its protections.
The FireComp project brings together partners with diverse expertise: a gaseous compressed hydrogen technology integrator as a coordinator (Air Liquide); a manufacturer of composite pressure vessels (Hexagon Composites); leading actors in international Regulation, Codes and Standards (HSL, Hexagon Composites, Air Liquide); experts in industrial risks (INERIS); experts in thermal radiation and mechanical behaviour of the composite (CNRS Pprime & CNRS LEMTA and SIEMENS); experts in thermal degradation, combustion and numerical simulation of composites (Edinburgh University, SIEMENS and CNRS Pprime) and an expert in European R&D collaborative project management (AYMING).
More information can be found on the internet website www.firecomp.info
Project Results:
FireComp Risk analysis (methodology)
The aim in this field was to describe the hydrogen systems currently being deployed by the partners of the consortium (Hexagon and Air Liquide) and to describe their fire safety strategies to avoid burst. This information is critical to define representative fire scenario depending on applications.
As a starting point, the possible fire scenarios were listed and classified according to first quick calculations of the heat levels involved. Various hydrogen systems with their possible environments were considered (see Figure 3)
Figure 3 Hydrogen systems considered
The risk analysis underlined three possible types of fire:
- Large engulfing fires
- “Partial” fires exposing an important part of the cylinder (especially for large cylinders), for example a car burning next to a hydrogen trailer
- “Local” or “punctual” jet fires, for example hydrogen jet from the pressure relief device of another burning cylinder
Punctual jet fires are known to lead to rapid burst of the cylinders, while leaving blind the TPRDs, due to the very high momentum of the flame. It was decided in the project to focus on large fires (partially or completely engulfing).
The quantitative study showed that for all applications, an acceptable level of safety can be reached, provided the pressure relief system has a low probability of failure on demand and 100% fire detection capacity. A new methodology was proposed for the design of safety strategies according to the conclusions of the QRA:
- The system integrating a composite cylinder should protect it efficiently against both local and engulfing fires
- Whatever other protections are used, the pressure relief system shall activate for all types of fires which are severe enough to lead to a burst
- The performance of the cylinder alone (without any protection) should be assessed in order to provide information to the integrator.
- The representative test should expose a large part of the cylinder, including a dome area, to a heat flux of 80 kW.m-2
Then, the integrator should design and test his safety devices, using
a) The end user’s own risk analysis, which depends on how and where the cylinder is used
b) the information received on the cylinder behaviour in fire (e.g. time to burst or safe pressure relief curve)
c) other requirements for the pressure relief (e.g. flame length, orientation, time, number of openings...)
The description of the hydrogen systems developed by Hexagon and Air Liquide allowed defining the representative fire scenarios (by simulation): the bonfire should reproduce the thermal heat fluxes observed in accidents involving fire. Similar levels of aggression should be reached in bonfire tests. Data from the fire safety barriers has been also gathered (essential for the risk analysis). The definition of methodology and the preliminary quantitative risk analysis of hydrogen systems have been conducted.
The objectives of the second phase of the study were to:
• Provide input data for the experimental tests campaign: level of thermal aggression to be reached in bonfire tests on composite vessels
• Define the methodology and gather the input data to conduct a Quantitative Risk Analysis (QRA) for the hydrogen systems developed by Hexagon and Air Liquide
• Quantify the bow ties of the risk analysis for each application
• Compare the results with traditional technologies (metallic cylinders)
The detail of the methodology that comes out of the project is gathered, with all supporting scientific arguments, in the public deliverables D6.4 and D6.5. It is based on obtaining (either experimentally or numerically) a safe pressure zone such as displayed onFigure 4, i.e. an area of pressure vs. time in fire within which there is no burst of the cylinder; and using it to design the safety devices for the structure considered (car, trailer, forklift or any other).
Figure 4 Example of safe pressure zone which can be used to design the fire safety strategy
FireComp Experimental Part #1: Sample Testing, Thermal & mechanical models
Thermal models
WP3 addressed the thermal properties of the composite material exposed to fire. In order to achieve the ultimate modelling objectives of the FireComp project (i.e. the simulation of the time to burst of the real tank subject to an inner pressure and a realistic arbitrary fire exposure), it is necessary to be able to sufficiently characterise the different aspects of the material thermal response. A significant part of this characterisation pertains to the pre-ignition conditions, when it is possible to applying some simplifying assumptions. The subsequent solid-phase degradation and combustion behaviours are inevitably much more complex and their generalised representation is challenging. Nevertheless both aspects are amenable to study via fire experimental techniques in conjunction with an array of modelling approaches.
The main vehicle of the experimental work was the cone calorimeter, a test apparatus with a heated coil which can generate different radiative heat fluxes in order to study the reaction-to-fire behaviours of the materials, shown below along with a degraded specimen. A controlled atmosphere version of the device was also used to explore the impacts of vitiated (oxygen starved) environments.
[Figure 5 Left: cone calorimeter apparatus. Right: degraded specimen after fire exposure]
The obtained values or correlation expressions for some of the essential thermal parameters, including those related to ignition itself, are summarised in Table 1:
[Table 1 Essential thermal parameters]
These include all of the relevant parameters needed to estimate the progression of the thermal response via heat transfer analysis (note that the thermal inertia has a single constant value, as it is an effective parameter, it cannot be temperature dependent). Whilst there are formal temperature limits on applicability of the thermal conductivity and specific heat, the impression of linear trends up to the bounds set by capability of the instrument suggests support for some extrapolation beyond – such representations are likely to be much more accurate than assuming constant values.
It is important to note the uncertainties in the measured parameters, in particular the ignition temperature which is obtained indirectly from the critical heat flux [2]. The latter value requires careful interpretation as it will vary according to the specimen thickness, orientation and rear-face boundary condition. The results reported above were obtained with the assistance of detailed numerical heat transfer analysis which described departures from more simplified treatments in the classical theory. A remaining source of uncertainty relates to the convective heat transfer coefficient. In equivalent cone calorimeter tests using a vertical orientation the measured critical heat flux rose to 15.0 ± 0.5 kW.m-2 representing the impact of the slightly enhanced convective heat transfer in the vertical flow. It seems likely that the heat transfer coefficient will have a very modest impact on results under many practical conditions.
Besides these macroscopic or engineering scale phenomena it is important to characterise the degradation behaviour at the material scale. Thermogravimetric analysis was used to establish the kinetics of the composite degradation under a) an inert atmosphere (100% nitrogen, so no reaction with air) and under b) air (21% oxygen). The reaction steps deduced by Genetic Algorithm optimisation are shown below, with parameter values for a modified Arrhenius expression A (Pre-exponential factor [s-1]); E (Apparent activation energy [kJ·mol-1]); n (Reaction order); v (Stoichiometric coefficient of product of reaction i) presented in the table:
[table]
Again there will be uncertainties in the representations but they provide a basis for prediction and testing.
Another aspect of the characterisation was examining the evolution of the emitted radiative spectrum, both pre and post ignition, and in isolation or in conjunction with mechanical testing. Identification of flame properties was achieved via spectrum reconstruction and it was shown that, as expected, there was no influence of mechanical loading or fibre orientation on the heat transfer processes.
To explore the post-ignition behaviours the flammability and combustibility properties of the composite material were evaluated for boundary conditions representing a range of real fire exposures, spanning heat fluxes of 15 to 65 kW.m-2 and allowing for the effects of vitiated fire environments by imposing different oxygen concentrations, i.e. nitrogen (~2%), normal air (21%) and an intermediate level of 11%, all by volume. It was demonstrated that at lower fluxes the influence of oxygen level can be dominant while the influence may be small or negligible at high heat fluxes. The same trends in mass loss behaviour were observed at each oxygen concentration, with a single peak at 20 kW.m-2 but a second peak appearing at higher fluxes, representing the pyrolysis of the char. The second peak is delayed at lower oxygen concentrations.
The evolution of gas mole fractions was consistent with the observations pertaining to the mass loss rates, with very low yields of the major species at the reduced oxygen levels. By contrast, the lightweight hydrocarbons (e.g. CH4, C2H2 and C2H4) showed more strongly under these conditions but were diminished at the intermediate level and very low under air, where they have been effectively eliminated by oxidation. The same trend was observed for NH3. NOx production follows the pattern of the major species concentration.
Flame spread was studied in a Lateral Ignition and Flamespread test (LIFT), shown right. The ignition time was in broad agreement with that achieved in the cone calorimeter at the same heat flux (50 kW.m2). There was a minor influence of the fibre orientation on the time to ignition, and the rate of flame propagation was also moderately influenced, being quickest for the EC90 orientation, intermediate for ECiso, and slowest for the EC45. There was also an influence on flame duration.
A final research element under WP3, not specifically mentioned in the original task breakdown, was an analysis of the charring rates achieved in the cone calorimeter. Two different specimen thicknesses were used and the rear face of the sample was either insulated or backed with an Aluminium block which served as a heat sink. For the latter case the results showed reasonably constant charring rates after a brief initial peak, and a reducing sensitivity to the imposed flux. Thus at longer time periods there was a convergence to upper values in the range 0.3-0.6 mm/min, presumably because the insulating behaviour of the char limits regression rates at higher fluxes. While higher rates are observed for insulating rear surfaces, with much stronger dependence on the incident flux, these conditions are expected to be of less relevance to many practical scenarios. Naturally, these results cannot be taken as definitive or assumed to apply in general, but it is noted that other aspects of the fire problem are subject to significantly larger uncertainties. In particular the main driver of degradation, the heat flux to the exposed surface(s), depends strongly on the nature of the fire and varies in space and time. Despite uncertainties in the charring rates the suggestion of an alternative potential simplification for model representations is therefore of interest [3].
To conclude the work package a range of models were developed to represent the thermal response of the material under generalised heating [1]. While the simplified model had some merits and had no significant deficiencies it was the detailed model implemented in LMS Samcef software that was subsequently used for the coupled thermo-mechanical analysis, as described below.
Mechanical models
WP4 addressed the mechanical properties of the composite material exposed to fire. In order to achieve the modelling objectives of the FireComp project (i.e. the simulation of the time to burst of the real tank subject to an inner pressure and a fire exposure), it is necessary to understand the damage mechanisms in a wound composite material undergoing such conditions. Different test campaigns have been thus carried out so as to develop thermo-mechanical models and to determine their parameters. The elementary damage modes and the stiffness components are identified from a set of samples of different stacking sequences cut from wound vessels and subjected to tensile (or, alternatively, three-point bending) tests. Because all lay-ups necessary to the determination of the mechanical properties cannot be manufactured by winding (f.ex. 0°), bibliographic data have to complete this experimental database.
Because it has been assumed the vessel composite wall is composed of two main parts (the inner layers are affected by heat without combustion, while the outer layers undergo also combustion), three types of tests have been carried out:
• First, composite samples have been mechanically tested at different homogenous temperatures.
• In a second step, in order to determine the effect of the char on the mechanical behaviour, the same types of samples have been first subjected successively to controlled heat energy and to tension.
• At last, the effect of the coupled thermo-mechanical load (heat flux / mechanical stress due to inner pressure) is investigated thanks to a specific set-up which allows applying a sustained load (creep) and a thermal energy (see Figure below).
From these three experimental programs, the following conclusions can be drawn:
• The mechanical strength remains equal to the strength of the virgin material as long as ignition does not occur. As soon as ignition occurs, the strength and the stiffness decrease proportionally to the incident energy.
• The char can be considered as a material with very low strength and stiffness.
• Whatever the stacking sequence, no significant influence of the mechanical load (if its level is lower than the damage thresholds) on the thermal degradation is observed: the strength of a sample subjected to a coupled thermo-mechanical load is similar to the strength of a sample undergoing only a heat flux.
These conclusions are the starting point for the simulations. Two different approaches have been compared: a model belonging to the class of Continuum Damage Mechanics (the damage mechanisms are described through internal variables whose evolution is in accordance with the principles of thermodynamics), and a simpler model based on failure criteria and a progressive mechanical degradation as soon as they are reached. Combustion (which degrades the mechanical properties) is assumed to be controlled by the temperature.
From this study at the sample scale, it has been proved that:
• the set of elementary samples proposed in this part is sufficient to characterize the basic damage modes and to identify the model parameters, provided the tests are completed with data from literature
• the assumptions regarding the thermo-mechanical coupling are valid and capable of simulating accurately the thermal and mechanical degradation at the sample scale, whatever the modelling approach.
In the next part, these models are used to simulate the time to burst of the pressurized vessel subjected to fire.
FireComp Experimental Part #2: Cylinder Testing, Cylinder Modelling
The main objective of FireComp project is to better characterize the conditions that are required to prevent from bursting. Following this objective, experimental work was done in order to improve the understanding of heat transfer mechanisms and the loss of strength of composite high-pressure vessels in fire conditions. The thermo-mechanical behaviour modelling of these vessels has then been possible: experimental burst and leak results have been accurately reproduced numerically.
The main outputs of the project are recommendations for Regulation Codes and Standards regarding the qualification of high-pressure composite storage and sizing of its protections, using the safe pressure curve methodology developed within the project.
In this context, the objectives of WP5 were the following:
• define the best conditions for tests to assure reliability, reproducibility and safety of the test,
• check that the tests performed at large scale in laboratory are representative of real fire scenarii and worst case scenarii,
• compare the influence of the filling media on the cylinder behaviour,
• develop and validate the model of thermo mechanical behaviour of a storage in fire,
• optimize the hydrogen release strategy using the cylinder model developed in the project,
• make recommendation for cylinder design to reduce the risk of burst.
The present period has mainly been dedicated to the experimental campaign, performed on composite cylinders. The objective of this task was to generate data to validate the thermo-mechanical model developed in WP 4. This model is helpful regarding the methodology of optimization of safety strategy developed for high pressure storage . Additional tests have been performed on what remain of composite cylinders that did not burst during fire tests.
In order to meet the targeted objectives presented previously, a complete instrumentation of composite reservoirs was implemented to follow the behaviour of the envelope of the reservoir during and after the fire, with thermocouples located inside the envelope of the vessel, and pressure measurement inside the vessel. Tests have been performed with the defined optimized reference fires on 19 and 36 L composite cylinders at nominal service pressure and at lower pressures. At the beginning an optimization of the experimental setup defined in task 5.1 has been performed with steel cylinders. Finally, Hydrogen gas fire has been retained for thermal aggression of composite cylinders, with the following characteristics:
• Use of 4 injectors,
• 1.5 g/s Hydrogen flow rate per injector.
• 0.5 g/s Oxygen flow rate per injector.
• Use of a confinement.
Following this optimization step, the experimental campaign performed on composite cylinders has been divided in 3 phases:
• radiant panel tests,
• tests performed to investigate the influence of filling media with real gas fire,
• tests performed to study of the behaviour of composite cylinders with real gas fire under various conditions.
The first phase with radiant panels allows to supplying data with a perfectly controlled thermal aggression at different levels.
The second phase led to consider that the filling media has not a real influence on the behaviour of the cylinder, in comparison with the impact of the variability of cylinders strength due to normal uncertainties linked with their fabrication process. Following tests were therefore performed with Nitrogen, which also presents the advantage to be non flammable, increasing by the way the safety of the experimental campaign.
The last phase, in addition to providing important data to validate the thermo-mechanical model by the study of various scenarii, has allowed defining a safety depressurization strategy after detection of the fire. This safety area has been defined with only 4 initial pressures, but it could be much more accurate by multiplication of curves, obtained experimentally, or thanks to modelling.
All the results of tests performed have been used to calibrate and validate the models developed during WP4. The models identified and validated at sample scale are used to simulate the thermo-mechanical behaviour of the cylinders tested in real conditions. The inputs needed for these simulations were:
• the thermo-mechanical behaviour identified in the previous part (stiffness and strength vs. temperature, ignition temperature, conductivity, specific heat, emissivity,...)
• the boss’ and liner’s geometry
• the composite lay-up (layer thickness and orientation in the cylindrical part); in the absence of data about the domes, it has been assumed the lay-up is the same in all parts of the vessel.
The following computation was performed: a vessel is pressurized at 700 bar and subjected to homogeneous heat fluxes of 100kW/m2 and 200kW/m2 (these values are the lower and the upper bounds of the actual flux, according to the experiment). This vessel has been modelled in two different ways: a 3D geometry (allowing the simulation of general fires) and an axisymmetric one (less time consuming, but limited to engulfing fires). The simulated temperature field is found to match the experimental measures recorded by the thermocouples inside the composite shell (the result is less satisfactory close to the outer surface because of the uncertainty of the thermocouple position). Regarding the time to burst, the difference between simulation and experiment is less than 2%.
In a second step, simulations at different pressurization levels have been carried out (525 bar, 250 bar, 100 bar). The test program showed a competition between layer degradation and temperature increase of the liner, leading to burst (for high inner pressure) or leak (lower pressures). A leak criterion (involving the temperature at the interface liner / composite) is chosen, and in spite of its simplicity, it allows capturing, in a numerical way, the transition between a burst regime and a leak one, and accurately simulating (without parameter recalibration, once the models are identified at sample scale) the times to burst or leak, according to the test conditions. The numerical tools implemented in this project are then a substitute for complex and costly bonfire tests. They are capable of drawing a pressure relief curve, providing, for a given value of the inner pressure, the time to burst or leak.
References
1. A framework for evaluating the thermal behaviour of Carbon fibre composite materials. J.P. Hidalgo, P. Pironi, R.M. Hadden and S. Welch. School of Engineering, The University of Edinburgh, Proceedings of the 2nd IAFSS European Symposium of Fire Safety Science, 2014.
2. Hidalgo, J.P. Pironi, P., Hadden, R.M. & Welch, S. (2016) “Effect of thickness on the ignition behaviour of carbon fibre composite materials used in high pressure vessels”, Proc. 8th Int. Sem. Fire & Explosion Hazards (ISFEH8), Hefei, China, 25-28 April 2016.
3. Hidalgo, J.P. Pironi, P., Hadden, R.M. & Welch, S. (2016) “Experimental study of the burning behaviour of a commercial carbon fibre composite material used in high pressure vessels”, Proc. ECCM17 - 17th European Conf on Composite Materials, Munich, Germany, 26-30 June 2016.
Potential Impact:
The purpose of this project is to develop public available knowledge and guidelines that will lead to safe introduction of high-pressure hydrogen composite vessels in early markets. The effective use and dissemination of the results of the project in the hydrogen safety community is a key objective.
The exploitation of the results will be successful when they will be integrated in the standards[OGY(F)1] . FireComp has directly or indirectly produced results and recommendations/proposals RCS-committees as listed below:
· ISO/TC 58/SC 3 has agreed to split up ISO/TR 13086-2 under development by ISO TC 58/SC3/WG24 into three parts. The first part is dedicated to 'Bonfire test issues'. Both AirLiquide and Hexagon is participants in the WG this Technical Report, and by that we as members of the WG we have directly forward any result/proposals generated in the FireComp project. This ISO Technical Report is intended to be a general guideline for all Cylinder Standards developed under ISO TC 58/SC3, both for new standards and standards due for revision. HEX is the convenor of the Working Group.
· ISO TC58/SC3/WG35 is responsible for developing a new standard (ISO DIS 17519) for permanently mounted cylinders from 450L-10.000L and for working pressure up to 100MPa. Elements of the new approach on fire testing is already integrated 2nd DIS version which is in now in for a new official ballot. There is still possibilities for updates based on results from FireComp. HEX in convenor or this WG.
· ISO TC197 WG15 has under development a standard for cylinders intended for Gaseous hydrogen — Cylinders and tubes for stationary storage (ISO 15399). HEX has been given the task related to fire protection. The methodology proposed by the FireComp project has been submitted to the Working Group, and new text including the option to follow new and alternative test strategies has been proposed by HEX. Convenor is AL and HEX is well represented in the WG.
FireComp has followed up any possibility within ISO for implementation of any results from the project by our participation in all relevant RCS-committees related to composite cylinders, including directly or indirectly GTR 13, ECE 134, SAE. In addition the results and recommendations/proposals by FireComp have been presented and discussed in different conferences like WHEC, ECCM, JITH, MECHCOMP, Int Sem Fire and Explosion Hazards, Congrès Français de Mécanique, IChemE Hazards, ICHS.
The results and recommendations/proposals has been well received, and will contribute to a gradually change/improvement on how potential risk related to storage of high pressure hydrogen in composite cylinders can be controlled. We expect an implementation by 2017.
The risk methodology developed will be a reference for industrial developments or integration for industrial. HSL will utilize the Risk Methodology while working with industrial partners in order to assess their high hazard systems, barriers and safe working parameters.
The results of the experimental set-up for composite vessels from WP 5 will be utilized by HSL to influence the set-up of their tank testing facility. This will provide a facility for industry to test their composite vessels.
The developments on modeling will be disseminated through conferences and publications.
Regarding the final modeling, it could be used in a potential project as a tool for vessel design optimization.
List of Websites:
The public website is www.firecomp.info
Relevant contact details can be found on the following webpage: http://www.firecomp.info/consortium.html
The use of hydrogen presents an opportunity for a carbon-neutral fuel source. There are a number of challenges that need to be overcome in order to exploit the benefits of hydrogen at large scale, in particular the storage of hydrogen must be secured. Successful large scale storage of hydrogen has been managed at high pressure with the use of composite vessels. Under normal working conditions a burst in service of a composite vessel is very unlikely, however when exposed to a fire the integrity of the composite material can be compromised and presents safety challenges. Thus research into protection strategies for the vessels in fire conditions is needed to combat the potential for a burst and limit hazardous the potential effects of hydrogen release.
The main objective of the FireComp project has been to better characterize the conditions that need to be achieved to avoid burst of composite vessels for CGH2 storage. To this aim, experimental work has been done in order to improve the understanding of heat transfer mechanisms and the loss of strength of composite high-pressure vessels in fire conditions coupled with modelling of the thermo-mechanical behaviour of the vessels. Both stationary and mobile applications have been considered including; automotive application, transportable cylinders, bundles and tube trailers.
Key activities undertaken during the FireComp project:
• Development of a risk methodology and proposed an approach for comparison with metallic cylinders.
• Modelling studies and large-scale experimental campaigns
• Comparison between experiments and modelling resulting in a good correlation with respect to time to burst.
• Development of a methodology to establish a safety pressure curve
• Development of modelling criterions for leak or burst scenarios
Scientific dissemination of the research has been achieved through presentations at conferences such as IChemE Hazards 26, ICHS and WHEC and public webinars.
Dissemination towards Regulations, Codes and Standards (RCS) has been achieved through the relationships established with experts from ISO TC58/SC3/WG24 committee and the new ISO Technical Report TR 13086-2 on topics regarding the qualification of high-pressure composite storage and sizing of its protections. FireComp has presented an excellent possibility to contribute to updating RCS on fire safety.
Project Context and Objectives:
Today’s challenge
Hydrogen is expected to be highly valuable energy carrier for the 21st century. To exploit its benefits at large scale, further research and technological developments are required. In particular, the performance of the storage of hydrogen under high pressure must be assessed. Keeping the integrity of composite pressure vessels in case of fire is necessary to protect people, infrastructure and the environment. More efficient protection strategies are then required to increase societal acceptance and consolidate the hydrogen energy business.
Better characterization of the conditions for safety.
The main objective of FireComp project was to better characterize the conditions that need to be achieved to avoid burst of composite cylinders in case of fire.
• Experimental work has been carried out to improve the understanding of heat transfer mechanisms and the loss of strength of composite high-pressure vessels in fire conditions.
• Then the modelling of the thermo-mechanical behaviour of these vessels has been set up. The model has been validated by full scale fire tests.
Different applications have been considered: automotive application, stationary application, transportable cylinders, bundles and tube trailers.
Perspectives of FIRECOMP project
Outcomes of the project are recommendations for the Regulation Codes and Standards regarding the qualification of high-pressure composite storage and sizing of its protections.
The FireComp project brings together partners with diverse expertise: a gaseous compressed hydrogen technology integrator as a coordinator (Air Liquide); a manufacturer of composite pressure vessels (Hexagon Composites); leading actors in international Regulation, Codes and Standards (HSL, Hexagon Composites, Air Liquide); experts in industrial risks (INERIS); experts in thermal radiation and mechanical behaviour of the composite (CNRS Pprime & CNRS LEMTA and SIEMENS); experts in thermal degradation, combustion and numerical simulation of composites (Edinburgh University, SIEMENS and CNRS Pprime) and an expert in European R&D collaborative project management (AYMING).
More information can be found on the internet website www.firecomp.info
Project Results:
FireComp Risk analysis (methodology)
The aim in this field was to describe the hydrogen systems currently being deployed by the partners of the consortium (Hexagon and Air Liquide) and to describe their fire safety strategies to avoid burst. This information is critical to define representative fire scenario depending on applications.
As a starting point, the possible fire scenarios were listed and classified according to first quick calculations of the heat levels involved. Various hydrogen systems with their possible environments were considered (see Figure 3)
Figure 3 Hydrogen systems considered
The risk analysis underlined three possible types of fire:
- Large engulfing fires
- “Partial” fires exposing an important part of the cylinder (especially for large cylinders), for example a car burning next to a hydrogen trailer
- “Local” or “punctual” jet fires, for example hydrogen jet from the pressure relief device of another burning cylinder
Punctual jet fires are known to lead to rapid burst of the cylinders, while leaving blind the TPRDs, due to the very high momentum of the flame. It was decided in the project to focus on large fires (partially or completely engulfing).
The quantitative study showed that for all applications, an acceptable level of safety can be reached, provided the pressure relief system has a low probability of failure on demand and 100% fire detection capacity. A new methodology was proposed for the design of safety strategies according to the conclusions of the QRA:
- The system integrating a composite cylinder should protect it efficiently against both local and engulfing fires
- Whatever other protections are used, the pressure relief system shall activate for all types of fires which are severe enough to lead to a burst
- The performance of the cylinder alone (without any protection) should be assessed in order to provide information to the integrator.
- The representative test should expose a large part of the cylinder, including a dome area, to a heat flux of 80 kW.m-2
Then, the integrator should design and test his safety devices, using
a) The end user’s own risk analysis, which depends on how and where the cylinder is used
b) the information received on the cylinder behaviour in fire (e.g. time to burst or safe pressure relief curve)
c) other requirements for the pressure relief (e.g. flame length, orientation, time, number of openings...)
The description of the hydrogen systems developed by Hexagon and Air Liquide allowed defining the representative fire scenarios (by simulation): the bonfire should reproduce the thermal heat fluxes observed in accidents involving fire. Similar levels of aggression should be reached in bonfire tests. Data from the fire safety barriers has been also gathered (essential for the risk analysis). The definition of methodology and the preliminary quantitative risk analysis of hydrogen systems have been conducted.
The objectives of the second phase of the study were to:
• Provide input data for the experimental tests campaign: level of thermal aggression to be reached in bonfire tests on composite vessels
• Define the methodology and gather the input data to conduct a Quantitative Risk Analysis (QRA) for the hydrogen systems developed by Hexagon and Air Liquide
• Quantify the bow ties of the risk analysis for each application
• Compare the results with traditional technologies (metallic cylinders)
The detail of the methodology that comes out of the project is gathered, with all supporting scientific arguments, in the public deliverables D6.4 and D6.5. It is based on obtaining (either experimentally or numerically) a safe pressure zone such as displayed onFigure 4, i.e. an area of pressure vs. time in fire within which there is no burst of the cylinder; and using it to design the safety devices for the structure considered (car, trailer, forklift or any other).
Figure 4 Example of safe pressure zone which can be used to design the fire safety strategy
FireComp Experimental Part #1: Sample Testing, Thermal & mechanical models
Thermal models
WP3 addressed the thermal properties of the composite material exposed to fire. In order to achieve the ultimate modelling objectives of the FireComp project (i.e. the simulation of the time to burst of the real tank subject to an inner pressure and a realistic arbitrary fire exposure), it is necessary to be able to sufficiently characterise the different aspects of the material thermal response. A significant part of this characterisation pertains to the pre-ignition conditions, when it is possible to applying some simplifying assumptions. The subsequent solid-phase degradation and combustion behaviours are inevitably much more complex and their generalised representation is challenging. Nevertheless both aspects are amenable to study via fire experimental techniques in conjunction with an array of modelling approaches.
The main vehicle of the experimental work was the cone calorimeter, a test apparatus with a heated coil which can generate different radiative heat fluxes in order to study the reaction-to-fire behaviours of the materials, shown below along with a degraded specimen. A controlled atmosphere version of the device was also used to explore the impacts of vitiated (oxygen starved) environments.
[Figure 5 Left: cone calorimeter apparatus. Right: degraded specimen after fire exposure]
The obtained values or correlation expressions for some of the essential thermal parameters, including those related to ignition itself, are summarised in Table 1:
[Table 1 Essential thermal parameters]
These include all of the relevant parameters needed to estimate the progression of the thermal response via heat transfer analysis (note that the thermal inertia has a single constant value, as it is an effective parameter, it cannot be temperature dependent). Whilst there are formal temperature limits on applicability of the thermal conductivity and specific heat, the impression of linear trends up to the bounds set by capability of the instrument suggests support for some extrapolation beyond – such representations are likely to be much more accurate than assuming constant values.
It is important to note the uncertainties in the measured parameters, in particular the ignition temperature which is obtained indirectly from the critical heat flux [2]. The latter value requires careful interpretation as it will vary according to the specimen thickness, orientation and rear-face boundary condition. The results reported above were obtained with the assistance of detailed numerical heat transfer analysis which described departures from more simplified treatments in the classical theory. A remaining source of uncertainty relates to the convective heat transfer coefficient. In equivalent cone calorimeter tests using a vertical orientation the measured critical heat flux rose to 15.0 ± 0.5 kW.m-2 representing the impact of the slightly enhanced convective heat transfer in the vertical flow. It seems likely that the heat transfer coefficient will have a very modest impact on results under many practical conditions.
Besides these macroscopic or engineering scale phenomena it is important to characterise the degradation behaviour at the material scale. Thermogravimetric analysis was used to establish the kinetics of the composite degradation under a) an inert atmosphere (100% nitrogen, so no reaction with air) and under b) air (21% oxygen). The reaction steps deduced by Genetic Algorithm optimisation are shown below, with parameter values for a modified Arrhenius expression A (Pre-exponential factor [s-1]); E (Apparent activation energy [kJ·mol-1]); n (Reaction order); v (Stoichiometric coefficient of product of reaction i) presented in the table:
[table]
Again there will be uncertainties in the representations but they provide a basis for prediction and testing.
Another aspect of the characterisation was examining the evolution of the emitted radiative spectrum, both pre and post ignition, and in isolation or in conjunction with mechanical testing. Identification of flame properties was achieved via spectrum reconstruction and it was shown that, as expected, there was no influence of mechanical loading or fibre orientation on the heat transfer processes.
To explore the post-ignition behaviours the flammability and combustibility properties of the composite material were evaluated for boundary conditions representing a range of real fire exposures, spanning heat fluxes of 15 to 65 kW.m-2 and allowing for the effects of vitiated fire environments by imposing different oxygen concentrations, i.e. nitrogen (~2%), normal air (21%) and an intermediate level of 11%, all by volume. It was demonstrated that at lower fluxes the influence of oxygen level can be dominant while the influence may be small or negligible at high heat fluxes. The same trends in mass loss behaviour were observed at each oxygen concentration, with a single peak at 20 kW.m-2 but a second peak appearing at higher fluxes, representing the pyrolysis of the char. The second peak is delayed at lower oxygen concentrations.
The evolution of gas mole fractions was consistent with the observations pertaining to the mass loss rates, with very low yields of the major species at the reduced oxygen levels. By contrast, the lightweight hydrocarbons (e.g. CH4, C2H2 and C2H4) showed more strongly under these conditions but were diminished at the intermediate level and very low under air, where they have been effectively eliminated by oxidation. The same trend was observed for NH3. NOx production follows the pattern of the major species concentration.
Flame spread was studied in a Lateral Ignition and Flamespread test (LIFT), shown right. The ignition time was in broad agreement with that achieved in the cone calorimeter at the same heat flux (50 kW.m2). There was a minor influence of the fibre orientation on the time to ignition, and the rate of flame propagation was also moderately influenced, being quickest for the EC90 orientation, intermediate for ECiso, and slowest for the EC45. There was also an influence on flame duration.
A final research element under WP3, not specifically mentioned in the original task breakdown, was an analysis of the charring rates achieved in the cone calorimeter. Two different specimen thicknesses were used and the rear face of the sample was either insulated or backed with an Aluminium block which served as a heat sink. For the latter case the results showed reasonably constant charring rates after a brief initial peak, and a reducing sensitivity to the imposed flux. Thus at longer time periods there was a convergence to upper values in the range 0.3-0.6 mm/min, presumably because the insulating behaviour of the char limits regression rates at higher fluxes. While higher rates are observed for insulating rear surfaces, with much stronger dependence on the incident flux, these conditions are expected to be of less relevance to many practical scenarios. Naturally, these results cannot be taken as definitive or assumed to apply in general, but it is noted that other aspects of the fire problem are subject to significantly larger uncertainties. In particular the main driver of degradation, the heat flux to the exposed surface(s), depends strongly on the nature of the fire and varies in space and time. Despite uncertainties in the charring rates the suggestion of an alternative potential simplification for model representations is therefore of interest [3].
To conclude the work package a range of models were developed to represent the thermal response of the material under generalised heating [1]. While the simplified model had some merits and had no significant deficiencies it was the detailed model implemented in LMS Samcef software that was subsequently used for the coupled thermo-mechanical analysis, as described below.
Mechanical models
WP4 addressed the mechanical properties of the composite material exposed to fire. In order to achieve the modelling objectives of the FireComp project (i.e. the simulation of the time to burst of the real tank subject to an inner pressure and a fire exposure), it is necessary to understand the damage mechanisms in a wound composite material undergoing such conditions. Different test campaigns have been thus carried out so as to develop thermo-mechanical models and to determine their parameters. The elementary damage modes and the stiffness components are identified from a set of samples of different stacking sequences cut from wound vessels and subjected to tensile (or, alternatively, three-point bending) tests. Because all lay-ups necessary to the determination of the mechanical properties cannot be manufactured by winding (f.ex. 0°), bibliographic data have to complete this experimental database.
Because it has been assumed the vessel composite wall is composed of two main parts (the inner layers are affected by heat without combustion, while the outer layers undergo also combustion), three types of tests have been carried out:
• First, composite samples have been mechanically tested at different homogenous temperatures.
• In a second step, in order to determine the effect of the char on the mechanical behaviour, the same types of samples have been first subjected successively to controlled heat energy and to tension.
• At last, the effect of the coupled thermo-mechanical load (heat flux / mechanical stress due to inner pressure) is investigated thanks to a specific set-up which allows applying a sustained load (creep) and a thermal energy (see Figure below).
From these three experimental programs, the following conclusions can be drawn:
• The mechanical strength remains equal to the strength of the virgin material as long as ignition does not occur. As soon as ignition occurs, the strength and the stiffness decrease proportionally to the incident energy.
• The char can be considered as a material with very low strength and stiffness.
• Whatever the stacking sequence, no significant influence of the mechanical load (if its level is lower than the damage thresholds) on the thermal degradation is observed: the strength of a sample subjected to a coupled thermo-mechanical load is similar to the strength of a sample undergoing only a heat flux.
These conclusions are the starting point for the simulations. Two different approaches have been compared: a model belonging to the class of Continuum Damage Mechanics (the damage mechanisms are described through internal variables whose evolution is in accordance with the principles of thermodynamics), and a simpler model based on failure criteria and a progressive mechanical degradation as soon as they are reached. Combustion (which degrades the mechanical properties) is assumed to be controlled by the temperature.
From this study at the sample scale, it has been proved that:
• the set of elementary samples proposed in this part is sufficient to characterize the basic damage modes and to identify the model parameters, provided the tests are completed with data from literature
• the assumptions regarding the thermo-mechanical coupling are valid and capable of simulating accurately the thermal and mechanical degradation at the sample scale, whatever the modelling approach.
In the next part, these models are used to simulate the time to burst of the pressurized vessel subjected to fire.
FireComp Experimental Part #2: Cylinder Testing, Cylinder Modelling
The main objective of FireComp project is to better characterize the conditions that are required to prevent from bursting. Following this objective, experimental work was done in order to improve the understanding of heat transfer mechanisms and the loss of strength of composite high-pressure vessels in fire conditions. The thermo-mechanical behaviour modelling of these vessels has then been possible: experimental burst and leak results have been accurately reproduced numerically.
The main outputs of the project are recommendations for Regulation Codes and Standards regarding the qualification of high-pressure composite storage and sizing of its protections, using the safe pressure curve methodology developed within the project.
In this context, the objectives of WP5 were the following:
• define the best conditions for tests to assure reliability, reproducibility and safety of the test,
• check that the tests performed at large scale in laboratory are representative of real fire scenarii and worst case scenarii,
• compare the influence of the filling media on the cylinder behaviour,
• develop and validate the model of thermo mechanical behaviour of a storage in fire,
• optimize the hydrogen release strategy using the cylinder model developed in the project,
• make recommendation for cylinder design to reduce the risk of burst.
The present period has mainly been dedicated to the experimental campaign, performed on composite cylinders. The objective of this task was to generate data to validate the thermo-mechanical model developed in WP 4. This model is helpful regarding the methodology of optimization of safety strategy developed for high pressure storage . Additional tests have been performed on what remain of composite cylinders that did not burst during fire tests.
In order to meet the targeted objectives presented previously, a complete instrumentation of composite reservoirs was implemented to follow the behaviour of the envelope of the reservoir during and after the fire, with thermocouples located inside the envelope of the vessel, and pressure measurement inside the vessel. Tests have been performed with the defined optimized reference fires on 19 and 36 L composite cylinders at nominal service pressure and at lower pressures. At the beginning an optimization of the experimental setup defined in task 5.1 has been performed with steel cylinders. Finally, Hydrogen gas fire has been retained for thermal aggression of composite cylinders, with the following characteristics:
• Use of 4 injectors,
• 1.5 g/s Hydrogen flow rate per injector.
• 0.5 g/s Oxygen flow rate per injector.
• Use of a confinement.
Following this optimization step, the experimental campaign performed on composite cylinders has been divided in 3 phases:
• radiant panel tests,
• tests performed to investigate the influence of filling media with real gas fire,
• tests performed to study of the behaviour of composite cylinders with real gas fire under various conditions.
The first phase with radiant panels allows to supplying data with a perfectly controlled thermal aggression at different levels.
The second phase led to consider that the filling media has not a real influence on the behaviour of the cylinder, in comparison with the impact of the variability of cylinders strength due to normal uncertainties linked with their fabrication process. Following tests were therefore performed with Nitrogen, which also presents the advantage to be non flammable, increasing by the way the safety of the experimental campaign.
The last phase, in addition to providing important data to validate the thermo-mechanical model by the study of various scenarii, has allowed defining a safety depressurization strategy after detection of the fire. This safety area has been defined with only 4 initial pressures, but it could be much more accurate by multiplication of curves, obtained experimentally, or thanks to modelling.
All the results of tests performed have been used to calibrate and validate the models developed during WP4. The models identified and validated at sample scale are used to simulate the thermo-mechanical behaviour of the cylinders tested in real conditions. The inputs needed for these simulations were:
• the thermo-mechanical behaviour identified in the previous part (stiffness and strength vs. temperature, ignition temperature, conductivity, specific heat, emissivity,...)
• the boss’ and liner’s geometry
• the composite lay-up (layer thickness and orientation in the cylindrical part); in the absence of data about the domes, it has been assumed the lay-up is the same in all parts of the vessel.
The following computation was performed: a vessel is pressurized at 700 bar and subjected to homogeneous heat fluxes of 100kW/m2 and 200kW/m2 (these values are the lower and the upper bounds of the actual flux, according to the experiment). This vessel has been modelled in two different ways: a 3D geometry (allowing the simulation of general fires) and an axisymmetric one (less time consuming, but limited to engulfing fires). The simulated temperature field is found to match the experimental measures recorded by the thermocouples inside the composite shell (the result is less satisfactory close to the outer surface because of the uncertainty of the thermocouple position). Regarding the time to burst, the difference between simulation and experiment is less than 2%.
In a second step, simulations at different pressurization levels have been carried out (525 bar, 250 bar, 100 bar). The test program showed a competition between layer degradation and temperature increase of the liner, leading to burst (for high inner pressure) or leak (lower pressures). A leak criterion (involving the temperature at the interface liner / composite) is chosen, and in spite of its simplicity, it allows capturing, in a numerical way, the transition between a burst regime and a leak one, and accurately simulating (without parameter recalibration, once the models are identified at sample scale) the times to burst or leak, according to the test conditions. The numerical tools implemented in this project are then a substitute for complex and costly bonfire tests. They are capable of drawing a pressure relief curve, providing, for a given value of the inner pressure, the time to burst or leak.
References
1. A framework for evaluating the thermal behaviour of Carbon fibre composite materials. J.P. Hidalgo, P. Pironi, R.M. Hadden and S. Welch. School of Engineering, The University of Edinburgh, Proceedings of the 2nd IAFSS European Symposium of Fire Safety Science, 2014.
2. Hidalgo, J.P. Pironi, P., Hadden, R.M. & Welch, S. (2016) “Effect of thickness on the ignition behaviour of carbon fibre composite materials used in high pressure vessels”, Proc. 8th Int. Sem. Fire & Explosion Hazards (ISFEH8), Hefei, China, 25-28 April 2016.
3. Hidalgo, J.P. Pironi, P., Hadden, R.M. & Welch, S. (2016) “Experimental study of the burning behaviour of a commercial carbon fibre composite material used in high pressure vessels”, Proc. ECCM17 - 17th European Conf on Composite Materials, Munich, Germany, 26-30 June 2016.
Potential Impact:
The purpose of this project is to develop public available knowledge and guidelines that will lead to safe introduction of high-pressure hydrogen composite vessels in early markets. The effective use and dissemination of the results of the project in the hydrogen safety community is a key objective.
The exploitation of the results will be successful when they will be integrated in the standards[OGY(F)1] . FireComp has directly or indirectly produced results and recommendations/proposals RCS-committees as listed below:
· ISO/TC 58/SC 3 has agreed to split up ISO/TR 13086-2 under development by ISO TC 58/SC3/WG24 into three parts. The first part is dedicated to 'Bonfire test issues'. Both AirLiquide and Hexagon is participants in the WG this Technical Report, and by that we as members of the WG we have directly forward any result/proposals generated in the FireComp project. This ISO Technical Report is intended to be a general guideline for all Cylinder Standards developed under ISO TC 58/SC3, both for new standards and standards due for revision. HEX is the convenor of the Working Group.
· ISO TC58/SC3/WG35 is responsible for developing a new standard (ISO DIS 17519) for permanently mounted cylinders from 450L-10.000L and for working pressure up to 100MPa. Elements of the new approach on fire testing is already integrated 2nd DIS version which is in now in for a new official ballot. There is still possibilities for updates based on results from FireComp. HEX in convenor or this WG.
· ISO TC197 WG15 has under development a standard for cylinders intended for Gaseous hydrogen — Cylinders and tubes for stationary storage (ISO 15399). HEX has been given the task related to fire protection. The methodology proposed by the FireComp project has been submitted to the Working Group, and new text including the option to follow new and alternative test strategies has been proposed by HEX. Convenor is AL and HEX is well represented in the WG.
FireComp has followed up any possibility within ISO for implementation of any results from the project by our participation in all relevant RCS-committees related to composite cylinders, including directly or indirectly GTR 13, ECE 134, SAE. In addition the results and recommendations/proposals by FireComp have been presented and discussed in different conferences like WHEC, ECCM, JITH, MECHCOMP, Int Sem Fire and Explosion Hazards, Congrès Français de Mécanique, IChemE Hazards, ICHS.
The results and recommendations/proposals has been well received, and will contribute to a gradually change/improvement on how potential risk related to storage of high pressure hydrogen in composite cylinders can be controlled. We expect an implementation by 2017.
The risk methodology developed will be a reference for industrial developments or integration for industrial. HSL will utilize the Risk Methodology while working with industrial partners in order to assess their high hazard systems, barriers and safe working parameters.
The results of the experimental set-up for composite vessels from WP 5 will be utilized by HSL to influence the set-up of their tank testing facility. This will provide a facility for industry to test their composite vessels.
The developments on modeling will be disseminated through conferences and publications.
Regarding the final modeling, it could be used in a potential project as a tool for vessel design optimization.
List of Websites:
The public website is www.firecomp.info
Relevant contact details can be found on the following webpage: http://www.firecomp.info/consortium.html
