Flashback arrestor installation mistakes: ten errors that leave a gas line unprotected and how to avoid them

Most flashback arrestors that fail in service were never given a chance to work. They were missing, fitted backwards, fitted in the wrong place, built for a different gas, choked with oil or water, or left in the line after they had already stopped a flame. The documented incidents point the same way: the dominant failure is absence or misuse, not a correctly fitted device letting a flame through. This page lists ten mistakes that recur in the standards, codes and incident reports, what each one does, and the rule that prevents it.

The simplest mistake is the most dangerous. In an Energy Institute hot-work case study, a cutting torch lacked the arrestors it should have had, the job checklist did not require them at both the torch and the cylinder end, and trained operators did not notice they were missing. The investigation after the first fire did not fix the cause, and a similar fire followed about seven weeks later. An arrestor that is required but not fitted is invisible until the day it is needed. Check for it before every job, as part of the daily safety checklist.

IS 11006:2011 cl 14.1 requires every body to be marked with the direction of nominal flow, along with the gas abbreviation (A for acetylene, O for oxygen) and the maximum operating pressure in bar. A device with a non-return valve fitted backwards normally stops the gas altogether, and the usual response on site is to remove it or swap it for a plain adaptor. A device fitted backwards that still passes gas is worse, because its non-return valve, temperature cut-off and pressure cut-off now face the wrong way. IS 11006 cl 11.5.1 tests the non-return valve in its most disadvantageous position, with gravity acting to open it, so a correctly oriented device does not depend on how it hangs.

Some quenching elements are bidirectional by design. Stephenson's 1930 sintered arrestor (BP 363,628) was arranged so that an explosion from either direction met the outer face of the element first, and packed Raschig-ring towers of the type tested by Schmidt and Haberl in 1955 are designed to arrest explosions travelling in either direction. That is a property of the element in a purpose-built plant arrestor. A welding-line arrestor is a stack of functions, and the valves in that stack are one-way. Treat every regulator and torch arrestor as directional.

A non-return valve stops gas flowing backwards; it does not reliably stop a flame. The American Welding Society puts it plainly in Fact Sheet 28: check valves alone will not stop a flashback. The C.P.I. said the same of dry non-return valves in 1968. Both devices are needed, in the right places. The British Acetylene Association's 1962 code for supply systems put a non-return valve between each cylinder and the header on the high-pressure side, and one or more flashback arrestors after the regulator on the low-pressure side. The non-return valve guide explains why each is there.

An arrestor is type-tested with a specific fuel-oxygen mixture: 32–35 % acetylene in oxygen for acetylene devices under IS 11006:2011 Table 2, against 13–15 % propane for LPG devices. A propane arrestor on an acetylene line has never been proved against the faster acetylene flame. An oxygen device on a fuel line, or the reverse, is equally wrong. The optional colour band under IS 11006 Annex B is red for fuel gases and blue for oxygen; the gas letter on the body is the definitive check.

Oil and grease have no place on an arrestor. Gas Cylinders Rules 2016 Rule 15(2) bars oil or lubricant on cylinder valves, and an oiled oxygen fitting is an ignition risk in its own right. Inside a fuel-gas element, an oil film insulates the metal from the flame and makes quenching less effective, the subject of the guide to oil, grease and arrestor elements. Thread paste, loose tape and swarf from new pipework end up on the inlet filter or in the element, raising pressure drop. Paint over the body can hide the flow arrow and gas marking that cl 14 requires to be permanent. Leak-test connections with a product made for oxy-fuel systems; an IMCA hose-fire investigation (Safety Flash 02/14) advised against soapy water or anything containing oil or grease on oxygen equipment.

Gas straight from a carbide generator is saturated with water. The C.P.I. ruled in 1968 that dry arrestors whose packing can be affected by moisture must not be used where acetylene comes from generators unless the gas is absolutely dry, and that their pressure loss rules them out of low-pressure installations generally. Water can also block a sintered element outright (Wilhelmsen/Unitor W-66 S manual). A generator outlet takes a wet hydraulic seal, which IS 11006:2011 covers as a wet-type device.

Each arrestor costs pressure, and IS 11006:2011 cl 10.6 rates flow at a 10 % pressure drop. Two or three devices in series, each sized on its catalogue maximum, can starve a heavy cutting torch, and a starved torch backfires more readily. IS 11006 cl 13 requires the manufacturer to give an installation method that keeps total pressure drop low, and IS 15190 (Part 1):2002 cl 4.12.1.5 calls for units in parallel where one cannot carry the flow. The pressure drop and flow capacity guide works through the numbers.

Position matters on plant and pipelines. IS 15190 (Part 2):2002 cl 6.7.4 notes that arrestors serve no purpose at the closed ends of a pipe, which must still be reinforced. IS 15190 (Part 1):2002 cl 5.4 relieves a reflection point of extra wall thickness only if an arrestor sits within the predetonation distance of it, which the standard's note puts at 2.4–16.75 m at 2.8 kg/cm² absolute in tubes wider than 35.5 mm. On a filling plant, the C.P.I. called for at least one arrestor per charging rack, between the compressor and the rack inlet, and IS 15190 (Part 2) cl 8.6 now requires three in series on every charging or discharging manifold. See dead legs and predetonation distance.

An arrestor that has stopped a flame has done its job once. Soot from the event fouls the element; IS 11006:2011 cl 10.5 forbids a temperature cut-off from being resettable; and Wilhelmsen/Unitor tells users that an arrestor whose cut-off and shut-off have operated "must be disposed of". A decomposition can also creep slowly through packing and re-ignite the gas beyond it, as at Griesheim, where a detonation followed twenty minutes after the explosion seemed to have been stopped; see the twenty-minute delay.