The twenty-minute delay: why an acetylene arrestor that seems to have worked can still fail

An acetylene flashback arrestor is supposed to stop a flame completely. Usually it does. But the research literature contains a failure mode that every plant operator should know: the arrestor appears to work, the bang stops, and minutes later the decomposition, having crept slowly through the packing, ignites the gas on the far side. The best-known case took twenty minutes. Partial success is the dangerous outcome, and it changes what operators should do after any flashback.

At the Griesheim works in Germany, steel-wool sponges were fitted as arrestors in an acetylene pipe 30 m long and 20 cm in diameter, working at 6 atm. Twice they arrested explosions. Twenty minutes later a vigorous detonation occurred. The decomposition had passed very slowly through the steel wool, and when it emerged it ignited the acetylene in the rest of the pipe (as compiled by S. A. Miller, 1965).

The same material behaves very differently when confined. Steel-wool plugs only 2.5 cm long, compressed to 75 % porosity, have suppressed detonations at initial pressures of 20–25 atm in 2.5 cm pipe. In a 20 cm pipe at 6 atm, loosely filled, steel wool merely slowed the reaction. The variable is not the material but the confinement, the compression and the size of the passages relative to the flame.

Hölemann and Hasselmann measured it directly (Forschungsberichte des Landes Nordrhein-Westfalen No. 526, 1956). They packed 1.9 cm tubes with 5 and 10 cm lengths of steel shot of 1 and 2 mm diameter and fired acetylene decompositions into them:

In the middle regime, the gas in the layer was pushed through by the explosion, the reaction inside the layer was incomplete, and when the decomposition emerged, the acetylene behind the layer exploded with violent pressure effects, followed by secondary decomposition of acetylene left in the packing. Hard brittle packings such as marble were split. The same two critical limits appeared with silicon carbide, petroleum coke, limestone, charcoal, silica gel, coke, cement, pumice, lime, phenolic resin and aluminium powder (Hölemann and Hasselmann, reports No. 383, 1957, and No. 690, 1959).

Three conditions moved the limits the wrong way: a film of oil, wax or water on the packing, which reduced heat transfer; coarse particles; and thin layers. Rough, adherent corrosion films moved them the right way, because they added surface. An arrestor element that is oily, wet, coarse or shallow is closer to the creep regime than the same element clean and deep. See oil, grease and arrestor elements.

The last row is the second half-success. After the Hüls pipeline explosion of 2 January 1954, Schmidt and Haberl showed that a packed tower repeatedly quenched detonations. But the gas in the long pipe had been heated to about 2,800 °C and raised to 72 atm, and it did not return to ambient at once; a pulse of more than 10 atm passed into the gas beyond the tower even though no explosion occurred there. A bursting disc set at 10 atm on a T-junction near the tower inlet, discharging upwards, cut the pulse to about 3.5 atmg. Arrest and relief work together; neither is enough alone. The packed-tower guide has the design rules, and bursting discs vs flashback arrestors explains why a disc is never an arrestor.

Webster and colleagues added a further warning in 1972 (IChemE Symposium Series No. 33): with strong initiation, detonation propagated at pressures below those needed to sustain a deflagration, and measured terminal pressure ratios exceeded 320 times. A line that is safe against a weak spark may not be safe against the flame front arriving from a decomposing component.

Partly, through repetition and margin in testing. IS 11006:2011 cl 11.6 does not accept a single successful arrest: each sample must stop five flashbacks in a static mixture of 32–35 % acetylene in oxygen at its maximum operating pressure, with time allowed between shots for conditions to recover, and the upstream gas must not ignite on any of them. Separately, cl 4.1 requires the casing to withstand ten times its maximum working pressure or 60 bar, whichever is higher. What no type test reproduces is the slow, minutes-long emergence seen at Griesheim, or an element that has become oily, wet or partly blocked in service. That gap is closed only by procedure: isolation, observation and replacement after every event.