Release without ignition from heat exchanger of a fertilisers production plant

A worker peforming an inspection round detected a hydrogen leak at a flange of a heat exchanger located between the conversion and decarbonisation stages of the ammonia processing unit.

The control room was alerted. A steam hose was installed blowing towards the flange to dilute the leak. Explosive gas measurements wer taken twice per shift to ensure that the consequences of the leak are under control. The area was placed under surveillance.

In the following weeks, several attempts to stop the leak were tried:
(1) Few days later, 2 vapour rings were placed around the flange. A marker was also placed around the flange to indicate that access to the area was prohibited. Several control points are defined.
(2) Three weeks later, a leak sealing system was put in place.
(3) Two weeks later, the injection of sealing paste into the collar did not seal the leak.
(4) Two months after detection of the initial leak, a worker noticed short term changes in the noise and direction at the leak.
The operator finally decides to work on the leaking flange. The main process gas supply was shut down more than 9 weeks after the initial leak detection while maintaining a nitrogen flow. The gas flow stopped in the pipe.
The operator replaced the flange bolts (rod and nut) and tightened the assembly. During the next scheduled shutdown the following year, they planned to replace the seals between the heat exchanger's shell and distribution box. Tis operation required a cold shutdown of the ammonia for nearly three weeks.

After the stop of the plant operation, inspection of the site revealed that the flange was loose at the time of the accident.
An ignited leak due to a loose flange had already occurred at the plant two months earlier (see HIAD_983). At that time, the technical reason for the loss of tightness was identified in thermal cycling. The reason for this event is not specified.

heat exchanger, hydrogen supply line

The normal operation of the plant continued all along the attempt to stop the hydrogen leakage.
A leak on a gasket due to a flange tightening defect had already occurred on site 2 months earlier

Several attempt to stop the hydrogen release, 9 weeks long.

The consequence of a cold shut-down would have been 3 weeks of stop of the production. The need to guarantee production rate and the negative financial consequence in failing to do so are among the most common causes of incidents.
In this case, the company did not want to shut down the plant and decided for the alternative to control and possibly reduce the leak. It succeeded in avoiding ignition. However, despite the continuous control of the leak and the surrounding area, the decision could have reduced the safety margin of the operations. The ARIA report mentions that the company installed as preventative measure, a SOFM (Single Operating Facility Maintenance). The report does not explain in what this measure consisted, therefore it remains unknown if an ad hoc risk assessment was performed to assess the hazards and risks of the temporary situation.