The flame you drive past is the plant's safest moment

One pipe, from the valves to the flame

A relief valve is a spring-loaded door: a spring holds a disc onto a nozzle, and when the pressure under the disc beats the spring the door opens. The question this page is about is where the gas goes after that. It goes into the relief header, the one pipe every valve on the site discharges into, and from there it has three things to get through before it can be burned.

The knock-out drum is sized to drop out droplets above about 300 to 600 microns, and it is the reason a flare does not rain fire. The water seal is the part people forget: it is a one-way valve made of water, and it is what stands between a lit tip and an air path back into a header full of fuel gas. The purge does the same job by flow rather than by liquid, and both are running when nothing at all is happening.

The three rules it keeps every hour

A flare is a control device, and in the United States it is held to 40 CFR 60.18 whenever it is being credited as one. Three of those requirements are the ones you can see from the road.

• A pilot flame present at all times Monitored, usually by thermocouple. Each pilot burns on the order of 70 scf/h of gas, every hour, for the life of the plant, whether or not anything is relieving.
• No visible smoke beyond five minutes in any two hours Smoke means the flame is short of air, so steam or air is injected at the tip to pull more in. That is the roar you hear and the plume you see on a trip.
• Exit velocity under the limit, and never an empty stack 60 CFR 60.18 caps exit velocity so the flame cannot be blown off the tip. A continuous purge keeps the stack full of fuel gas, because a cold stack that has drawn in air is a flammable mixture with an ignition source bolted to the top of it.

The destruction efficiency a flare is credited with in a permit is a regulatory assumption that comes with those operating conditions attached, not something measured on your stack. It is worth knowing which of the two you are quoting.

The height is the heat

The stack is not tall for dispersion and it is not tall for show. It is tall because of how much heat a person standing on the ground is allowed to feel. API 521 puts continuous exposure where people work at about 1.58 kW/m² including solar, and every metre of stack is bought to meet it.

Two things fall out of it that are worth keeping. The setback goes to zero before the radiation does: past a certain height nobody standing at grade is inside the contour at all, and that is the height the fence line is really buying. And the load moves the answer much harder than the height does, because D goes with the square root of Q while the height only has to beat it in one dimension.

The drum that was not a flare

On 23 March 2005 a refinery in Texas restarted an isomerisation unit. The raffinate splitter was overfilled, the liquid went out of the relief valves, and the relief path ended at a blowdown drum of 1950s design whose stack vented straight to the atmosphere. There was no flare on the end of that line.

• the fill

  The column was filled far past its normal level with the level instrument reading on scale, so the board showed a tower that was not full.

• the relief

  The relief valves lifted as designed and discharged to the blowdown drum. The drum overfilled and a geyser of liquid and vapour left the top of its stack.

• the cloud

  The release fell back to grade and spread as a vapour cloud across the unit, where contractor trailers had been sited close to the stack.

• the ignition

  The cloud found an ignition source. Fifteen people were killed and about a hundred and eighty were injured.

The investigation board's recommendation was the plain one: send relief streams to a flare rather than to an atmospheric blowdown drum, and keep people out of the blast path. The picture at the top of this page is what that answer looks like as equipment. The flame on the tip is the last part of it, and it is the part you can see from the road.