Refinery turnaround

Texas City, March 23, 2005: BP restarts its isomerization unit after a maintenance outage, the raffinate splitter tower overfills, liquid escapes through the vent stack and ignites. Fifteen dead, one hundred eighty injured. The Chemical Safety Board found a faulty level indicator and a startup procedure that was not followed. Every loss control engineer knows the pattern: refinery accidents happen more often in transient phases than in steady operation. A turnaround brings hundreds, sometimes thousands, of outside contractors on site for a few weeks, opens equipment that is seen once per cycle, and multiplies hot work permits and lockouts. The property underwriter reads it from two angles. During the outage the unit is empty and inerted, but the work itself creates its own perils: hot work, heavy lifts, isolation errors. At restart, everything that was taken apart goes back under pressure at once, and that is where the risk window is shortest and densest. The risk survey questions follow: the schedule, what was inspected and what was deferred, the pre-startup safety review, the findings and how they were closed. A turnaround pushed back to protect margin lengthens the run and wears the equipment; the market notices. Then there is business interruption. The downstream policy carries a deductible expressed in days, and the indemnity is calculated on what the site would have produced without the loss. If the accident happens during the outage, the days already planned as downtime are not a loss: only the gap against the schedule counts. A schedule overrun without physical damage stays with the operator.

• Chemical Safety Board, BP America Texas City refinery explosion investigation (2005)