This incident occurred in the cooling medium system for subcooling liquefied natural gas (LNG). This medium comprises hydrocarbons and nitrogen.

The leak occurred during removal of insulation from a one-inch pipe segment upstream of a safety valve. This was a planned action before taking out the safety valve for recalibration. The safety valve was connected to the cavity of a manual valve in the cooling circuit.

After detection of the leak by the insulation personnel and confirmation by the area technician, the leak was notified to the central control room (CCR) and the plant evacuated. Ignition-source disconnection was also initiated from the CCR.

No automatic actions were initiated by the gas detection system during the incident, since none of the detectors close to the safety valve initiated an alarm.

A modular valve (compact double block and bleed) is installed in the line upstream of the safety valve to isolate towards the cooling system when the safety valve is removed.

The direct cause of the leak is that the bleed plug in the bleed line for the modular valve was in the open position, while the actual bleed valve also stood slightly open. The latter had probably been moved slightly during the process of removing the insulation.

An anti-blowout type, the bleed plug involved in the incident is left-hand threaded and differs from the plugs normally found at the plant. It has to be screwed out to close.

Since the valve arrangement and bleed plug were enclosed in insulation, the natural conclusion is that the plug has been open since the most recent calibration of the safety valve in 2021. With the plug screwed in, detecting that it remains open is not easy for someone not familiar with its type.

Through a search of its own incident database, Equinor has identified several earlier incidents related to bleed plugs in the wrong position.

Actual consequence

The actual consequence of the incident was a gas leak lasting about 6.5 hours, with an estimated quantity in the order of 9 300 kilograms and an initial rate of 0.8 kilograms per second.

One person was hit by spray from the leak and followed up by a nurse. No permanent harm has been identified.

Production at HLNG was shut down for eight days as a result of the incident.

Potential consequences

The Havtil investigation has concluded that ignition or explosion was unlikely. Both when the incident occurred and during the response to it, the potential existed for further exposure of personnel.

The open bleed plug was a hidden fault which could have had a different outcome had the leak not occurred. This plug is included in the activity for verifying depressurisation before opening the system and removing the safety valve.

Unfamiliarity with the plug’s function could have led to the system being opened before it was depressurised.

Results of the investigation

The investigation has identified several factors which were or could have been significant for the incident taking place and the scale of the leak. These include:

  • knowledge about and documentation of the bleed-plug type involved in the incident
  • managing gas leaks
  • safety system design (positioning of gas detectors).

Identified lessons

The investigation has identified the following lessons to be learnt in relation to bleed-plug functionality and design of gas detection systems which Havtil wants to highlight.

  • Bleed plugs can vary in design and functionality. When closing, some are screwed in and others out. And it can be difficult for people not familiar with a specific type to determine that it is in the right position. Various types of plugs might be present in a plant, particularly ones which have been in operation for a while and subject to modifications. Training and documentation are important factors for ensuring familiarity with plug functionality.
  • Liquefied gases, in this case a cooling medium, will initially act as a heavy gas when they leak, regardless of their actual composition. Positioning of gas detectors must be assessed on the basis of how the gas will behave if it escapes.

Nonconformities and improvement point

Havtil’s investigation has identified nonconformities from the regulations related to:

  • knowledge about and documentation of equipment components
  • procedures and training
  • gas detection
  • dealing with impaired barriers
  • documentation
  • inadequate radio communication in the response/smoke-diving team
  • inadequate warning system – PA system in Oddasit building.

Furthermore, it has chosen to categorise one observation as an improvement point:

  • inadequate documentation on and description of preparing risk assessments and tactical emergency response plans.

What happens now?

Havtil has requested that Equinor explains to it how the nonconformities will be tackled and provides its assessment of the improvement point observed. The deadline for these responses has been set to 19 April 2024.