Failure of a diaphragm is not infrequent, but the seizure of the main nut threads is very rare. The manufacturer claimed this had never occurred before with this type of unit. The broken plunger is likely the result of poor instructions/communications between the vendor and the user. The detail and quality of drawings provided by the vendor were poor given the level of investigation and repair needed for this occurrence. It is possible that the proprietary nature of some equipment information may have been a factor.

The visiting intern had several years of experience in research projects and tasks that were similar to those in progress at this laboratory. This led to the incorrect assumption that formal introductory training for this laboratory was unnecessary. Steps should always be taken to ensure that anyone working in an unfamiliar laboratory setting is properly trained.

It is important to have written operating procedures for the use of laboratory-scale equipment involving hazards or risks of this nature. Operating procedures should also document any equipment parameter limits. Such procedures should be reviewed with all personnel as part of their laboratory training before they perform any experiments.

More information on management of change can be found in the Lessons Learned Corner and also in the Hydrogen Safety Best Practices Manual.

After the aforementioned incident, a rigid cage was designed to protect the reactor from external conditions, and to protect the contents of the hood and any experimenter from the reactor, in the event of a pressure burst from the reactor cell. Additionally, the experimental setup was redesigned to include only electrical distribution that has been verified to be in compliance with National Electric Code and that is located outside the cage. Further, the experimental setup has been redesigned to include a hydrogen supply line shutoff, so that if the pressure reactor cell integrity is compromised, the hydrogen supply is shut off. The fume hood has been fitted with a hydrogen sensor. Lastly, researchers were reminded that hydrogen experimental setups should be verified by another person, and hydrogen-involving experiments should be carefully monitored.

Hydrogen development work is inherently hazardous, and greater precautions than the norm need to be taken to ensure a safe work environment.

High-pressure fueling hoses should be examined daily for signs of external damage, including corrosion, abrasion, cuts, and kinking. High-use fueling hoses should be replaced every six months.

1. Revise fire protection operations surveillance tests to include a maximum of 110% of the agent net weight to prevent over-filled cylinders from being placed in service.
2. Identify cylinders currently in service that exceed the 110% maximum allowable limit, remove them from service, have them inspected and refilled to the new criteria, and then return them to service.
3. Purchase new compressed gas pilot control cylinders.
4. Revise the plant industrial safety program to enhance cylinder storage requirements by eliminating the use of ropes as restraints.
5. Purge the gas storage shed of all non-related storage, secure all cylinders with chain, fabricate cylinder racks for small cylinders, and eliminate storage of items next to the hydrogen system emergency manifold.
6. Conduct a probabilistic safety assessment of the possibility of missile hazards from compressed gas cylinders.
7. Change from the current CO2 vendor to another qualified service vendor.
8. Investigate whether a breakdown of the fire protection QA program occurred.

• The fuel cell vehicle that was involved in the accident has been retired. The fuel cell power plant from that vehicle has been removed and is being used in another fuel cell vehicle.
• The fuel cell vehicle accident reinforced the need for training of drivers, supervisors and emergency response personnel. As an action item, this project team will conduct refresher training courses for the drivers and local emergency response personnel. The project leads conducted training classes on hydrogen safety and incident response for local emergency response personnel; including the local fire department and the police prior to vehicle deployment and the station opening. A significant learning by this project team is that emergency response agencies are subject to frequent personnel changes. As such, training should be repeated periodically.

An investigation team was formed to determine the cause of this anomaly. The team performed a fault tree analysis which identified the following five main paths or gates: fault in the original weld, age, component not built per design intent, overstress, and design error. This led to 48 events on the fault tree, all of which were investigated. The investigation, analysis and testing determined that the first path, fault in the original weld, was the primary source of the failure.

The fault tree analysis pointed to three probable and two possible contributors of the failure. The probable contributors included incorrect filler rod selected by welder, insufficient fusion in the weld, and insufficient weld penetration. The possible contributors included incorrect power level setting and incorrect heat level applied during welding.

Materials laboratory work confirmed that improper filler metal (4043 Al) was used to weld the 5083 Al alloy vent pipe. The 5083/4043 combination is not recommended for Al welding per either current or past standards. In addition, the failed weld was a very poor weld, most likely a field weld. NDE data indicated a higher number and severity of defects in this weld compared to other welds in the vent line. According to the material lab's failure analysis report, the weld failure resulted from tensile overload originating at the bottom surface of the vent line during tanking operations and was likely the result of a single overload event.

Vent line inspection was performed to identify other suspect welds. Inspection methods included the following: visual inspection, conductivity tests on both sides of each weld to identify the pipe metal (5000 or 6000 series Al), helium leak check of welds, non-destructive evaluation (NDE) using a combination of x-rays and ultrasonic testing, and silicon etch tests to identify weld material. The silicon etch test procedure was successfully developed in the course of this effort to identify 4043 weld filler material in the field.

Instrumentation attached to vent lines during subsequent operations has provided data for the ongoing weld stress analysis and screening criteria to determine which welds required clamshell repairs. Weld defect growth/propagation can now be monitored at periodic intervals.

An unexpected temperature difference of ~200 deg F between the top and bottom of the piping was found to occur during tanking operations in the same area of the failed weld. This temperature difference was responsible for the high thermal stress seen there. Data indicates that the peak stress in that pipe section occurred at about the same time that the weld failed and was higher than predicted.

The combination of the cold water temperature (reducing the fatigue strength of the bolt), and the abnormally high number of cyclical stresses imposed by the imbalance from the hydraulic system check valve failure resulted in the failure of the fasteners.

Proper installation of check valves and other equipment should be visually inspected prior to pressurization.

Maintenance on the low-pressure venting system was not occurring at regular intervals. Ventilation integrity is now checked before starting an experiment.

Excessive venting of hydrogen from the tank due to lower facility consumption, in combination with extreme temperature conditions, placed thermal stress on the gland nut, causing a leak. The low consumption of hydrogen resulted from the shutdown of some production equipment and the delay of additional production equipment coming online. The tank size is too large for the facility's current hydrogen demand.

The hydrogen supplier will conduct annual training on handling all types of gases used by the facility and will include the local fire department in this training. The facility will continue daily rounds to look for visual evidence of leaks, and the hydrogen supplier will exchange the 9,000-gallon tank for a smaller 4,500-gallon tank to significantly reduce pressure build up from lower usage.

Follow CGA G-5.5 vent end system designs.

Several recommendations were outlined by the investigating committee to govern future operations of the hydrogen compressor in the synthetic liquid fuels laboratory:

1. Install a cutoff that will shut down the compressor when the suction pressure drops to a positive pressure of one to two inches of water.
2. Install an oxygen analyzer with an alarm as close to the compressor suction inlet as possible.
3. Attach all cylinders to the filling rack, discharge the contents to the atmosphere, and evacuate before filling.
4. Analyze as soon as possible the contents of at least one cylinder of each rack of cylinders charged.
5. Eliminate the recycle connection from the suction side of the compressor to the blowdown pot.
6. When compressing hydrogen, disconnect all interconnecting lines from the hydrogen suction line except the inert gas line.
7. Establish more definite procedures and responsibilities for maintenance repairs, construction, and operations.

1. Consider periodic testing of breakaway device.
2. Consider determining shear force limit of vehicle receptacle adapter fittings.

Mechanical pressure gauges tend to be imprecise if only used in a narrow portion of the full scale. Digital transducers, although slightly more expensive, offer much more precision. The event happened because the set pressure was only 10% of full scale, and the error of the mechanical gauge was over 5%.

While this is not the reason for the event described, control of the charging pressure is one of the most crucial parameters. Although the storage is at low pressure, the pressure increase upon temperature increase can be much steeper than the ideal gas law would predict, depending on the charging conditions.

The chosen trigger point of 35 bar for the pressure relief valve is very low. Based on the tank design, at least 60 bar would be acceptable.

Additional discussion about working with reactive metal-hydride materials in the laboratory can be found in the Lessons Learned Corner on this website and in the Hydrogen Safety Best Practices Manual.

Corrective actions included replacing the breakaway with a new one, which restored normal operation of the dispenser.

Verify and periodically inspect the pull/separation force adjustment if the breakaway is so equipped.

Additional information on equipment maintenance and inspection can be found in the Hydrogen Safety Best Practices Manual.

A gas detector was added in close proximity to the compressor shaft and a vibration switch is under consideration. Additional predictive measures are being considered to predict bearing failure. In addition, the manufacturer has been contacted and the bearing design is being analyzed to see if it can be improved.

Included inspection on monthly preventive maintenance plan and evaluated alternate materials for better cold-weather performance.

The fitting was an SAE straight thread and was likely loosened by torque applied to the fueling hose. After the incident, these fittings had additional means applied to restrict loosening, a cover installed to deflect any leakage, and means taken to restrict hose torque by using a different style nozzle. In addition, different fittings have now been deployed.

Metallurgical examination of the two failed disks by light optical microscopy (LOM), scanning electron microscopy (SEM), and energy-dispersive x-ray spectroscopic analysis (EDS) found them to be fabricated from pure nickel with evidence of extensive fracture. Each of the 24 tubes in the system is protected by a burst disk. Examination of another disk in the system that had not given way found that it, too, possessed surface fracture features, and they extended around the entire periphery of the rupture disk. Such defects are indicative of hydrogen embrittlement. An inspection of all vent circuits found that each of the 24 disks in service was made from nickel. Nickel is a material not recommended for hydrogen service in rupture disks.

Prior to the attempted use of the tube bank for hydrogen service, the vessel had been employed for helium service. The pressure vessel documentation accompanying the system indicated that the burst disks were made of stainless steel and rated to 10,000 psig. Careful physical inspection of system hardware is recommended on any system being adapted to hydrogen service. In this instance, inspection conducted prior to the transfer in service could have alerted operators to the need to install disks with the proper material, and therefore, have prevented the incident.

Relief of hydrogen gas should not lead to movement of the vent line sufficient to cause system damage. Corrective actions included increasing the line diameter and adding bracing between the lines and the system bulkhead to strengthen the components should other releases occur. The hardware that failed was of a commercial origin. Caution should be exercised to insure that all hardware is adequate for its designed purpose, even when procured from a commercial source.

More information on management of change can be found in the Lessons Learned Corner and also in the Hydrogen Safety Best Practices Manual. A web-based resource developed by Sandia National Laboratories to provide data on hydrogen embrittlement of various materials is available at Technical Reference for Hydrogen Compatibility of Materials.