Fire in a refinery
A 24" pipe carrying synthesis gas (Hydrogen = 43. 5%, Nitrogen = 14. 08%, Methane = 0. 2%, Carbon Dioxide = 12. 3%, Water = 29. 3% by volume) ruptured. About 60 kg of hydrogen contained in the gas initially released formed an unconfined vapour cloud and exploded. The explosion was audible up to 2. 5 km from the factory and it caused considerable alarm to the residents. Lengths of pipe on either sides of the rupture were displaced as a result of the depressurization. An 8 m length pipe downstream rotated upon itself of about 380° while a 12 m length pipe upstream rotated of about 90°. This 12 m section damaged the pipes on an adjacent pipe-rack rupturing the vent header from the desulphuriser unit and causing the release of approximately 600 kg of methane. The hydrogen released from the upstream section of the 24" pipe subsequently burned as a jet-fire about 70 m long. The released methane initially burned as a cloud (60 m long) and then as a jet-fire from the ruptured vent header. On hearing the explosion and by the indications from the plant instrumentation of a major leak, the plant operator shut-off the supply of natural gas to the plant and shut-down the individual units. After about 5 minutes the fire extinguished. A small amount of ammonia vapours which had been released from a fitting on a vapour return line (damaged by the 12 m section pipe) dispersed harmlessly.
Event Initiating System
Classification of the Physical Effects
Nature of the Consequences
Cause Comments
Metallurgical examination of the 24" pipe showed that a fatigue failure occurred, induced by thermal cycling. Examination of the fracture faces of the ruptured seam weld showed a significant amount of fatigue cracking initiating at the bore of the pipe along a length of about 3 metres of the weld downstream the water injection point. At the upstream end, the rupture propagated circumferentially around the pipe as a shear failure. At the other end, the rupture arrested at a circumferential weld and then propagated as a circumferential tear in the pipe material. The spray pattern from the nozzle was such that water at 80°C was intermittently hitting the inner surface of the pipe and quickly cooling the surface. The process gas stream at a temperature of 230°C quickly heated up the pipe wall again only for it to be quickly cooled by water impingement at the start of the next cycle. This phenomenon which affected a number of areas in the pipe wall downstream of the water injection nozzle caused a number of small fatigue cracks which ultimately developed so that the material near the longitudinal weld failed. Subsequent examination of the spray nozzle showed that this may have been distorted for some time prior to the accident causing preferential impingement possibly enhancing the failure mechanism although it is not concluded that this distortion was the main cause of the failure. There was a two year history of previous leaks in welds further downstream on this section of pipe which it is now recognized were due to the same cause. At the time the company wrongly diagnosed the cause as external stress corrosion cracking.
