Case Study: BP Texas City Explosion 23 March 2005
The startup procedure called for heating the raffinate splitter tower reboiler return flow to 275°F at no more than 50°F per hour to avoid excessive pressure in the tower. However, during this startup the temperature of the reboiler return to the tower rose as high as 307°F, and from 10 a.m. to 1 p.m. the temperature increased at a rate of 73°F per hour (ERROR). In the previous five years, most of the 19 startups had deviated from procedure temperatures and heat up rates.
The Day Board Operator and the Day Lead Operator agreed that the heat to the furnace should be reduced, and at 12:42 p.m. fuel gas flow was reduced to the furnace. At this time the raffinate splitter level transmitter displayed 80 percent (8 feet) but the actual tower level was 140 feet.
In the hours preceding the incident, the tower experienced multiple pressure spikes. In each instance, operators focused on reducing pressure: they tried to relieve pressure, but did not effectively question why the pressure spikes were occurring. They were fixated on the symptom of the problem, not the underlying cause and, therefore, did not diagnose the real problem (tower overfill).
From 10 a.m. to 1 p.m. the transmitter showed the tower level declining from 97 to 79 percent. The Day Board Operator thought the level indication was accurate, and believed it was normal to see the level drop as the tower heated up (what kind of training program could produce that understanding of system response?). At the time of the pressure upset, the Day Board Operator became concerned about the lack of heavy raffinate flow out of the tower, and discussed with the Day Lead Operator the need to remove heavy raffinate from the raffinate splitter tower. None of the ISOM operators knew the tower was overfilling. At 12:42 p.m., the Day Board Operator opened the splitter level control to 15 percent output, and over the next 15 minutes opened the valve five times until, at 1:02 p.m., it was 70 percent open. However, heavy raffinate out flow had not actually begun until 12:59 p.m.
The heavy raffinate flow out of the tower matched the feed into the tower (20,500 bpd) at 1:02 p.m. and by 1:04 p.m. had increased to 27,500 bpd. Unknown to the operators, the level of liquid in the 170-foot tower at this time was 158 feet, but the level transmitter reading had continued to decrease and now read 78 percent (7.9 feet). Although the total quantity of material in the tower had begun to decrease, heating the column contents caused the liquid level at the top of the column to continue increasing until it completely filled the column and spilled over into the overhead vapor line leading to the column relief valves and condenser.