At 4:50 am, an explosion and a fire engulfed one of the 3 reactors at a pharmaceutical plant. Since the employees were on break in another building, no injuries were reported. First responders installed a safety perimeter, escorted employees to the exits and evacuated the toluene. The mayor activated the municipality’s disaster plan: unaware of where the smoke cloud was heading due to darkness, he asked neighbours within a 300-m radius to stay indoors; a call centre was set up. Fire-fighters extinguished the blaze around 9 am. Measurements recorded by helicopter in the smoke cloud indicated no hazard, and the disaster plan was lifted by the end of the morning.
The explosion had occurred on the 4th floor of one of the plant’s 4 buildings; the fire spread to a storage zone containing empty polyethylene barrels placed on pallets, and then to the roof. One of the building 3 manufacturing units was damaged. Employees usually assigned to this unit were made redundant for the day. The facility resumed its activities that evening around 10 and operations were back to normal by the following week.
A thermal runaway reaction of the distillation process, as it was winding down, caused the explosion. The product instability had been analysed, and the operating protocol for this infrequently-used process had been adapted in 2012 to stage the reaction within a “safe” temperature range (at 60°C for a 24-hour TMRad – adiabatic conditions – at 111°C).
The recommended temperature, measured in the reaction mass and relayed to the reactor temperature display viewed by the pipeline technician, had not been exceeded on the day of the accident. A post-accident analysis of the temperature curves recorded by the programmable controller showed a gradual rise in gas temperature inside the reactor (i.e. indicating the onset of decomposition), until surpassing the temperature at which the runaway reaction could no longer be controlled. Since this “gas” temperature data point had not been reported, the pipeline technician had no way of detecting the operational drift; the cooling it had initiated at the end of the batch run was insufficient to prevent the ongoing runaway.
This onset of decomposition can be explained by local heating: since the reactor had been heated by steam with its double shell, traces of product (projections onto the reactor’s inner walls) might have been subjected to higher temperatures, capable of triggering decomposition.
Reaction analyses conducted post-accident on the reagents and products, both pure and mixed confirmed the sequencing hypothesis of this event, which may have occurred during previous batch runs.
The plant operator suspended the production of molsidomine according to this manufacturing process. The site’s reactors containing 2 temperature probes were equipped with a 2nd (separate) display in order to relay values of the 2nd temperature measurement, as a means of detecting possible deviations. A special training session was organised for all process safety managers.
