Just as a railcar was starting to be filled at 8:50 am, ammonia (NH3) was released for 20 to 25 min via a spherical tank vent at a site containing both nitrogenous products and fertilisers.
A filling step was completed by removing the rigging screw to shut once again the railcar’s bottom valve and close the fill valve on the loading arm. A manual drain valve was opened and the arm, before being disconnected, was depressurised via a special hose hooked up to a pipe reaching at the top of the spherical tank a vaporization pot whose vent opened to the outside atmosphere. Recently-installed NH3 sensors around the vent enabled detecting eventual anomalies. During transfer of a small NH3 volume 48 hours earlier, under weather conditions not favourable to gas dispersion, the detectors triggered inexplicably at first by the operator and later due to a defective pump trap; the automatically-secured station was placed back into service and the loading continued.
An NH3 transfer was thus ongoing when subcontractors, some of whom stationed aboveground, fell ill while working at the site periphery around 9 am. Peripheral detectors measured 15 ppm of NH3 on the ground (alarm threshold: 10 ppm). The workforce was confined; 35 workers experienced illness, 12 of whom were hospitalised for observation: 10 returned home that afternoon, the 2 others discharged a couple hours later.
The detectors surrounding the vent did not activate. The malfunction was located on the hose (white frost on the drain valve); liquid NH3 spread to the loading arm in the hose and then partially vaporized, at least in the pot, so as to continuously escape via the vent to the top of the spherical tank. Favourable weather conditions pushed the toxic cloud higher, resulting in the initial non-detection. A portion of the NH3 then fell back to the ground, upsetting a few minutes later the onsite subcontractors, while at the same time being detected by the ground NH3 sensors installed in this sector. The operator informed the classified facilities inspectorate, the municipality, emergency services and the closest residents, in addition to issuing 2 press releases.
The operator had foreseen a drain valve deficiency or a material handling error. An expert evaluation of the manual valve, of a common design and replaced in May 2009, revealed no anomaly. The technician supervising transfers had been monitoring 2 railcars simultaneously and was not near the arm when the incident occurred; the complete valve closing after degassing of a previous car could not be confirmed.
Several measures were adopted: use of a 2nd valve on the drain network (already existing, yet not included in the procedure); installation of counterweight valves (or “dead man” valves) along with an abnormal leak detection system in the drain recovery network (analogue pressure switch with alarm); revision of loading procedures; and list of loading operations with technicians’ approval and memorandum.
