By Gregory Hale
Just under 10 years ago, a vapor cloud explosion (VCE) occurred at the Caribbean Petroleum Corporation (CAPECO) site in Bayamón, Puerto Rico on October 23, 2009.
An above-ground gasoline storage tank was being loaded with gasoline from a cargo ship shortly before the incident and a tank ended up over filled which resulted in gasoline flowing through tank vents at the top of the storage tank which formed a large vapor cloud that extended off the plant into the surrounding area covered with thick jungle like vegetation.
The resulting VCE caused blast damage to on-site and off-site structures. Fires ended up ignited in 20 storage tanks within the tank farm, which burned for several days. In terms of injuries, there were several workers hurt, but no fatalities. The tank farm got lucky.
The ensuing investigations resulted in certain conclusions, but J. Kelly Thomas, vice president and blast effects section manager at Baker Engineering and Risk Consultants, Inc. and his team have a few conclusions of their own.
“They were offloading gasoline from a ship to a series of tanks,” Thomas said during his talk entitled “Caribbean Petroleum Case Study” at the American Institute of Chemical Engineers 2019 Spring Meeting and 15th Global Congress on Process Safety in New Orleans, LA. “Tanks 405 and 504 and 411 were just filled. Next up was Tank 409.”
“Tank 409 was built about three years before the incident,” Thomas said. “It was 63 feet high. Tank 409 was not communicating with the control room. The local indicator was not working properly. They began filling tank 409 at 10 p.m. expected it to be filled at 1 am. Operators checked Tank 409 at 11 and when they came off the tank at 11:23 and it was not full.
“They overflowed the tank through elevated vents; there were level detection issues. They created large flammable cloud.
“Operators entered the tank farm from west and noticed a fog and they figured out what it was and they called the ship to stop pumping the gas. They overflowed about 4,600 barrels. The tank overflowed from elevated at 60 feet and the overflow created a waterfall effect that generated a mist vapor mixture.”
At 12:12 a.m., the ship stopped pumping the gasoline.
“Ignition occurred at 12:22 a.m. about 10 minutes after the flow of gasoline stopped,” Thomas said.
Based on a detailed material balance, accounting for the material off-loaded from the ship and received at the facility, approximately 4,600 barrels (193,000 gallons, 731,000 liters) of gasoline overflowed from Tank 409. The tank was overflowing for about 24 minutes before flow was stopped, and approximately 10 minutes later the VCE occurred. The combined time for gasoline vapor/mist to spread before ignition was therefore 34 minutes, the investigator said.
“The initial blast came from an electrical enclosure that services pumps, a secondary blast ignition came from an electrical enclosure,” Thomas said.
“The VCE was a deflagration,” Thomas said. “Assuming a detonation (explosion) grossly over predicts observed damage.”
On top of that, the rapid and extensive secondary fires were due to gasoline mist wetting vegetation, he said.
The following conclusions were drawn from the investigation.
• The VCE at CAPECO was the result of overfilling of Tank 409 during gasoline unloading from a ship.
• The volume of fuel estimated to have spilled was 4,600 barrels.
• The vapor-mist cloud dispersed in all directions in the calm wind conditions due to momentum and density of the vapors and mist, with the cloud spread generally following the topography. The cloud width was approximately 2,000 ft (610 m).
• The VCE was a deflagration. Thorough site inspection by the authors revealed no evidence of a deflagration-to-detonation transition (DDT), even on a small, localized basis.
• Blast loads were consistent with a deflagration. By contrast, blast loads for the detonation of a flammable cloud size per HSL guidance for an overflowing gasoline tank were over-predicted by at least 5 times and by as much as two orders of magnitude. The over-prediction of the blast pressure was, on average, a factor of 40.
• Using a cloud volume consistent with the HSL methodology and applying the BST VCE blast load methodology would be expected to yield reasonable consequence predictions for similar events.
• Extensive secondary fires are suspected to have been rapidly ignited due to gasoline mist wetting vegetation, tanks and equipment. Such rapid fire spread is uncharacteristic of VCEs involving flammable gases.