Vapor Cloud Triggers Historic U.K. Blast
Vol 20 Issue 3
EDITORS NOTE: The following article is taken from a presentation delivered March 16, 2005, by Allen to the Worldwide Firefighters Conference in Manchester, U.K.
The Guiness Book of World Records still lists it as the worst industrial explosion in U.K. history. On June 1, 1974, a rupture in a 20-inch bypass pipe dumped nearly 40 tons of flammable cyclohexane at a chemical plant near Flixborough, creating a colossal vapor cloud that soon found an ignition source.
Beside the human toll, the Flixborough disaster is important in that it awoke management to the threat that a single vapor cloud release could destroy an entire industrial facility. Engineers began spending endless hours developing accurate ways to estimate the worst case scenario, i.e., the maximum foreseeable loss, from vapor cloud loss so that management could take the financial steps to protect the company.
Given the size and complexity of industrial activity in the U.S. alone, vapor cloud releases are unfortunately inevitable. Aside from oil refineries, a multitude of petrochemical plants use flammable gaseous materials as feed or intermediate stock. However, a better understanding of the forces at play in such events reveal that steps can be taken to actually limit the size of the monetary cost from a plant-killing disaster to a large but sustainable loss.
STUDYING THE AFTERMATH
In the days after Flixborough the loss prevention engineers came up with a misleading new term to cover the a vapor cloud scenario -- Unconfined Vapor Cloud Explosion. The problem is unconfined vapor cloud releases involving saturated hydrocarbons such as propane and butane have historically very seldom caused explosion damage. Ignite these vapors in open air and what results is a simple flame front that passes from one side of the cloud to the other.
This appearing to be true, where does the blast damage come from when a vapor cloud release ignites in your plant? Does it come from the pockets of blast damage within the area or is most of the damage done by the ensuing fires fed from broken pipes, valves and the resulting flame impinged vessels and piping BLEVEs? Instead of measuring the overall effect, lets break the incident down into its quantitative parts.
With an explosive such as TNT, a detonation will always occur with or without containment, creating tremendous heat and force which, in turn, creates a sharp blast wave. With hydrocarbon vapors after ignition the flame front compresses the unburned gases in front of it. The faster the flame front moves the more pressure is generated until it reaches supersonic speeds, causing both blast and overpressure damages. The generated blast overpressures are much slower than overpressure waves generated by explosives such as TNT. Think of an overpressure caused by the ignition of a hydrocarbon in a confined space as a slower moving but still powerful tsunami with pressurized air pushing the wave. In a vapor cloud explosion the wave passes through the plant, dissapating its energy much more rapidly than an overpressure generated by an explosive.
Studying the debris in the aftermath of a disaster like Flixborough is very difficult because the entire facility was severely damaged. The best place to analyze the damage done by a vapor cloud release is in an area that has limited ensuing fires. Examing events in such an environment clearly indicates that the explosive overpressures happen as a result of the vapors entering enclosures before the ignition. Then this enclosure causes the vapors to produce an explosion that creates a blast wave, resulting in damage to adjacent and sometimes distant equipment. The easiest examples of this to see are the cabs and engine compartments of vehicle. Doors are popped open and there is a a telltale arch in the roof or hood which indicates an overpressure trying to reshape the square interior into a perfect sphere.
Enclosures need not be four sided to cause an overpressure. Three sided and two sided enclosures can also cause blast effects under the right conditions as can be seen by the photograph of the railroad boxcar laying on its side with the floor and roof blown out by an explosive overpressure. Notice that the tank cars next to it were relatively undamaged by comparison. Three months after Flixborough a massive leak from an LPG tank car in Houston's Englewood rail yard had a delayed ignition that then set off a series of explosions and fire that burned for 31 hours. More than 600 railcars were destroyed or burned across 20 tracks of the 2?-mile-long rail yard. A coupler punched through a tank car releasing 38,000 gallons of butadiene in a matter of minutes. The resulting vapor cloud ignited from a switch engine almost a half a mile away.
About 20 or 30 open box cars showed evidence of overpressure explosion either inside or beneath the cars. In some cases, the floors of the box cars were blown through the roof, leaving the walls intact. But in most cases the walls and roof were blown out.
In review of overpressure events involving gasoline and heavier vapor clouds the blast damage characteristics change. Investigations have shown that vapors from gasoline type materials from a tank overfill or process release tend to create a lot more explosive pressure from much less containment on the outer edges of the vapor cloud than LPG. At Englewood, the vapor cloud was so big that only the outer layer or skin of the cloud appeared to be mixed with air enough to create a flammable mixture.
Inside the cloud the mixture was much too rich to allow the vapors contained in boxcars to explode. They appeared to only have an interior flash fire without causing explosive overpressures. Only in a specific area perpendicular to the wind direction both upwind and downwind from the point of release was found a band of cars that were damaged from explosive overpressures inside them. The reason for this is the cloud was deep enough to allow the flammable outer layer to enter the box cars that had open doors.
"Bluff bodies," i.e. congestion of equipment in a process unit will also add speed to the flame front but it is unclear if this will cause enough turbulence to generate blast overpressure waves within the unit. If workers can not easily pass between vessels, pipes and reactors, the likelihood of faster flame speeds and some additional damage is higher than if the process unit were well spaced or open.
PREVENTING THE WORST
Enough evidence exists to support the theory that explosive overpressures do originate with enclosures. The overpressure then wrecks pipes and valves, contributing to the fire that follows. So how do we use this information to our advantage? Pressurizing buildings inside process units is an old story at most hydrocarbon processing plants. In most cases we pressurize to keep the vapors away from electrical ignition sources. The mostly unrecognized phenomenon of explosive overpressures from building containing flammable vapors has not been widely recognized as a reason to pressurize enclosures (See Figure 1 from the API RP500 electrical classification paper). In some cases another containment area is born. To get massive electrical cables in and out of switch gear rooms it is not unusual to raise the room three feet off the ground, creating another area for a potential overpressure. This area should be skirted and pressurized also to prevent the vapors from exploding under the building. It is also time to look at compressor sheds, operator shed and control room instrumentation shacks inside the process area as potential containment areas. Any area or building that can trap vapors can turn into a bomb.
Pressurization guidelines generally call for an air inlet 20 feet above the roof of the building, to pull in fresh air. The building should maintain a positive pressure around .15 inches of water to sustain more pressure inside than outside. When you open the door a rush of air pushes out of the building making sure outside gases are kept out.
Distance of the buildings from the source of a released vapor cloud is good protection but there is no predicting what a vapor cloud might do or where it will go.
It now appears that Flixborough was destroyed not because of a single vapor cloud detonation, but overpressures owning to confinement and congestion. These overpressures damaged other pipes and valves, triggering further releases that continued to burn. Typically in these large events fire fighting efforts are generally abandoned after the first large vessel BLEVE, dooming the plant to destruction.
The extent of the initial destruction and ensuing fires can be greatly reduced if steps are taken to identify and either eliminate or further protect these potential areas of vapor confinement. Being proactive will lessen the financial risk and potential for harm.