At 2:08 pm, while walking along the main corridor of a Seveso-rated plant producing animal feed, an employee noticed a strong smell of ammonia (NH3). Upon identifying a leak, he activated the emergency shutoff switch by breaking the glass to secure the NH3 network through closing the shutoff valve; he then alerted the site’s safety unit.
The leak (gas + aerosol) was located 2 m aboveground on a 40-mm diameter pipe, with a tap on one side connected to a central, 80-mm diameter NH3 pipeline on a rack placed 4 or 5 m high, while on the other side connected (after a horizontal and vertical stretch of undetermined length) to a pressure sensor located 1 m above ground level. This sensor served to regulate NH3 pressure within the central pipeline over a length of several tens of meters.
A maintenance team unsuccessfully attempted to install a sealing collar. The internal emergency plan was activated at 2:27 pm, and 294 of the 297 employees at the facility (the other 3 had not heard the alarm siren) remained in confinement for 30 min.
The central pipeline was isolated by closing 2 safety valves; then, the 60-m segment lying between the two valves was drained and inerted with nitrogen. The corresponding intervention was completed in less than 30 min and included sprinkling of the opening throughout the operation. Shortly before 3 pm, the Director of Emergency Services considered that the situation was under control and lifted the emergency plan, yet access to the main corridor remained prohibited until further instructions.
NH3 supply to workshops was stopped while awaiting replacement of the defective pipe segment by a new connection with identical specifications. NH3 distribution and site activities gradually resumed, between midnight and 2 am on 7th January, i.e. some 10 hours after the alarm first sounded. Operating losses were estimated at several hundreds of thousands of euros.
During normal operations, the NH3 network was supplied from a 250-tonne storage sphere. The quantity of NH3 emitted was not known, but could nonetheless be evaluated at less than 200 kg. A wind (25 km/hr) from the north-west facilitated dispersion of the NH3 could that had formed and no odours were perceptible outside the plant. According to the site operator, the leak had resulted from corrosion under the pipeline’s heat insulation, with an equivalent opened surface area estimated at 5 mm².
In the safety report conducted on these installations, the operator had anticipated leaks of this type on openings equivalent to at least 1% of the pipe cross-sectional area, in considering that smaller-sized leaks would not exert a significant impact beyond the site boundary; this approach was confirmed by the present incident when correlated with the day’s weather conditions.
The operator had been inspecting the thickness of these NH3 pipes twice a year at points it had predefined. One such control point had been identified just a few metres upstream of the leak zone, yet no point had been indicated downstream given that the leak was located at the end of the tapping section. No anomaly had been detected during the most recent inspection performed at this upstream spot.
