Explosion and fire on a compressor in a petrochemical plant

The explosion occurred in a benzene production unit (Litol process), which was being restarted following a scheduled shutdown for maintenance. About 30 kilograms of hydrogen gas were relesed into a compressor shed from a burst flange operating at 4.8 MPa (700 psig). The following explosion caused two injuries and two fatalities.

The investigation is reported in the referred scientific article, and arrived to these conclusions:
1. the hydrogen was released through a failed flange gasket.
2. it accumulated and ignited under the roof of the compressor shed. It also escaped from this semi-confined structure and accumulated under the roof of an adjacent building.
3. Probably the cloud under the roof of the compressor shed ignited first, and deflagrated to the rest of the cloud.
4. the total estimated cloud mass was 30 kg, however the force of the blast suggest a much higher energy.
5. This leak could not have been detected, because the leak test procedure use gas only up to 0.7 MPa.

The INITIATING cause was the failure of a gasket on hydrogen compressor. Flange and gasket were of the correct type. Perhaps the gasket was displaced and/or the bold improperly tightened. According to reference MacDiarmit et al 1989, the real cause of the gasket failure could not be found, despite establishing that the gasket was not catered. This could have been caused by the use of a lubricant when installing the gasket, and the use of a gasket type not corresponding to the original specification (but which demonstrated good performance for many years).

A ROOT CAUSE were unauthorised or improperly considered modifications, which clearly played a major role in this incident.
There had been several modifications carried out in the Litol unit which were not adequately documented or for which documented authorization was not available. For example the installation of a valve drain and the addition of gables and a wall to the compressor shed originally consisting of only a roof over the compressors. The walls were likely added to aid in winter operation and maintenance, but hindered the dispersion of the gas.

Additional ROOT or CONTRIBUTING CAUSE were (1) a risk assessment which did not consider the possibility of a full gasket failure, (2) a leak test procedure performed at a pressure value well below the operative pressure.

gasket, hydrogen compressor

After a maintenance outage, the plant was starting up for a new production run. The accident occurred after the hydrogen system pressurisation past 600 psig.

The incident occurred on a plant holiday, and all the offices were empty.

DESCRIPTION OF THE PROCESS: styrene manufacturing is based on the alkylation of benzene and ethylene. Ethylbenzene is formed and then dehydrogenated to produce a product–styrene, with by-products: Toluene and hydrogen.

For the benzene needed for a styrene production process, the plant used a single-step hydro-de-alkylation process on coke oven light oil by means of Litol a catalyst . This complex process achieves desulphurisation, removal of paraffin and naphthenes, and saturation of unsaturated compounds, in addition to de-alkylation and dehydrogenation.

The nature of the explosion damage is noteworthy. The damage insider the perimeter of plant was considerable. Piping, ladders, and other light steel work were buckled and distorted. At the control house, exposed walls were bulged inward slightly and doors were blown in, but columns and beams were unaffected. Inside, ceilings fell and instruments were damaged. Flying glass was a major hazard in this explosion, and wired glass proved not to be shatter-proof.
The office building of neighbouring company sustained damage consistent with exposure to a peak overpressure of about 1.1 psi. At a distance of 150 meters (500 feet), this is equivalent to the peak overpressure from a ground burst of 1 ton (0.91 tonnes) of TNT, according to commonly-used references.

The sound was heard at several km distance. In the outside area, neighbouring office buildings were also damaged, doors, particularly Shipping entrance doors, were buckled and broken. There was significant window damage at a distance of 365 meters (1200 feet) and random breakage as far away as 900 meters (3000 feet).

The two release of hydrogen ignited within a couple of minutes. The fire was extinguished by the operators.

One general conclusion coming from the investigation (MacDiarmid et al.) is that the use of the TNT-equivalent method for quantifying hydrogen explosion must be done with great caution. “The prediction of blast effects can be complicated by real-world factors".

Another important lesson learnt is the importance to be able to manage changes and assess the new risks introduced by these changes. The originally semi-open construction around the hydrogen compressor had been closed later with a wall hindering the dissipation of hydrogen and creating the conditions for an explosion in confined space.

According to MacDiarmid et al (see references), following the accident and the investigation, the following safety measures were implemented in the plant:
(1) Operative procedures were reviewed and improved, (design and engineering safety checklists, material specifications)
(2) Blast and Fire Resistance was improved (separation standards for buildings and facilities, more resistant components for building components)
(3) Management of change procedure was adopted, with a focus on ‘closure’ technologies
(4) To avoid full emptying of storage unit during accidents, emergency shut-off valves were installed
(5) Safety awareness was increased by better communication and training.