A polyethylene production line in a workshop, included within a vast petrochemical industrial complex, was shut down for logistical reasons tied to an insufficient raw material supply chain. It was restarted 12 days later, according to a specific, several-step procedure due to the high-pressure polymerisation process involved.

Downstream of the polymerisation reactor, this line contained a separator between the polymer (polyethylene) sent to the extruder and the monomer (ethylene), which did not react and which could be partially recycled via a heat exchanger used to cool the separator. The first steps proceeded normally: purge, test of system seal, and initial start-up phase. During the 2nd start-up phase, which began the next day around 1:50 am with a gradual rise in separator pressure, an abnormal temperature increase in the separator caused the cooling circuit regulation valve to open. A leak of pressurised ethylene was then detected at 2:26 am downstream of the valve, along with a sudden pressure drop in the circuit and an alarm triggered by the gas detection network. The technician halted the chemical reaction and placed the installation in safe mode. The non-ignited leak ceased 3 min later; 1 tonne of ethylene was released into the atmosphere. The cloud that had formed then dispersed, with no human or property consequences being observed. The operator informed the classified facilities inspectorate within hours of the accident.

According to the operator, the temperature rise was due to intense and localised heating of the separator cooling circuit pipe, with this heating being caused by “soft” decomposition of the ethylene contained in a priming volume of the cooling circuit downstream of the valve. The ethylene would have leaked through an opening a few centimetres wide on the circuit rupture disc line. This highly exothermic decomposition stemmed from a phenomenon of isentropic recompression of the ethylene generated by abrupt pressure changes when restarting the line and in the presence of a priming volume. During a maintenance intervention one month earlier, this valve had been replaced by a more leak-proof model, yet this change served to amplify pressure drops in the line upon restart. Several measures were adopted prior to restarting these installations, namely:

• reduction of the priming volume on the separator cooling circuit;
• verification of the absence of this type of volume (and risk of soft decomposition) on the other workshop lines;
• preservation of a constant flow rate on the cooling circuit section downstream of the regulation valve, in order to avoid sudden pressure drops upon opening the valve;
• study leading to modification of the cooling circuit and elimination of the priming volume.