In a pharmaceutical company classified Seveso upper-tier establishment, a deflagration occurs at the end of a hydrogenation reaction involving a chemical raw material, acetic acid, water and a catalyst: palladium. The operator was performing hydrogen purges, consisting in sending hydrogen under pressure into the reactor via the hydrogen vent. he heard a thud and observed a light above the hydrogen vent, at the bell of the flame breaker. He triggered the emergency stop and started the emergency plan. The staff was evacuated and the other facilities on site secured. However, the Intervention team does not observe anything anomalous during the reconnaissance and do not have to intervene. The emergency status is called of f30 minutes later.Before starting the analysis, the reactor involved in the event is secured by placing it under inert atmosphere (nitrogen) and by keeping it at 10 C .A similar incident in all respects had already occurred on 04/06.By disassembling of the system, palladium trace of palladium was found in the vent line, possibly as consequence of foaming of the reaction medium which dragged the palladium up to the flame breaker. The palladium, dried out during the purge operations may have self-ignited at the outlet of the vent, in presence of oxygen. In the case of the two events, a long pause of the production campaign had took place beforehand. This has probably favoured the drying of the palladium in the vent line. Following the first event, the flame breaker was cleaned, but apparently the cleaning had been insufficient.Another contribution to the incident was the fact that, since 4 months a new procedure was put in place, requiring operators to empty the system at each batch. Before the introduction of this new practice, the presence of liquid in the system avoided the drying of the catalyst and therefore its auto-ignition.