Oxygen enrichment during oxy-acetylene cutting in a vessel
Inside a decommissioned 8 m³ stainless-steel mixing vessel a fitter is using an oxy-acetylene torch to cut out a seized agitator shaft. The vessel is 2.4 m tall with a 450 mm side manway, and ventilation is a single 300 mm flexible duct from a blower hung at the manway. The gas tester stands at the manway and takes readings every 20 minutes. At 10:20 oxygen reads 20.9% and LEL 0%. At 10:40 LEL still reads 0% but oxygen reads 24.3%. The fitter, in cotton overalls and leather gloves, has noticed nothing, though there is a smell of acetylene near the torch body. The oxygen and acetylene cylinders are outside with 10 m of hose; the hose crosses the manway lip, where there is a kink wrapped in tape and a shiny patch on the oxygen line. The permit requires the atmosphere to be kept between 19.5% and 23.5% oxygen. The fitter says the reading is just over the line and asks for 15 more minutes to finish the cut.
1. Interpret the trend from 20.9% to 24.3% oxygen. What is the likely cause, and why is this a stop-work even though the excursion looks small?
Read the worked answer
Stop the cut now.
- The trend is a signal, not drift. A rise of 3.4 percentage points in 20 minutes inside an 8 m³ space means an oxygen source is discharging into it. The only pressurised oxygen in the vessel is the torch line, and a taped kink with a shiny patch on the oxygen hose is exactly where it will be leaking. The acetylene smell indicates a second, separate leak at the torch body - a fuel leak that on its own could produce an explosive mixture.
- Why enrichment is a stop-work at only 24.3%. The flammable range of most materials widens and the minimum ignition energy falls as oxygen concentration rises. Materials that will not sustain burning in air - cotton overalls, leather gloves, hair, grease and oil residues on threads and fittings - will burn vigorously in an oxygen-enriched atmosphere and can be lit by a small spark or a flake of hot slag. Clothing that has sat in 24% oxygen for 20 minutes may be saturated, and the hazard travels out of the vessel with the wearer.
- The permit limit is 23.5%. The atmosphere is outside it and the AGT has no authority to accept an excursion.
- The 0% LEL is not reassuring. In an oxygen-enriched atmosphere a catalytic bead sensor over-reads and can be permanently damaged, so its zero is not proof that there is no fuel, and it says nothing at all about the acetylene leak.
2. Set out your actions in the correct order and justify the sequence.
Read the worked answer
- Stop the work. Signal the fitter to shut the torch at the torch body first, then at the cylinders. Avoid anything that can spark inside the space - no grinding, no switching electrical equipment, no hot slag.
- Get the fitter out into the open air and keep him away from ignition sources, including smoking and synthetic clothing. Oxygen-soaked clothing should be removed if practicable and aired for at least 15 minutes well away from ignition, then washed before reuse; a garment saturated with oxygen is a fire risk long after the person has left the vessel.
- Isolate and repair the source. Close the cylinder valves, bleed the lines, remove the damaged hose, replace it, and leak-test the whole run with soap solution or a proper detector, including regulator, torch body and flashback arrestors. The acetylene smell means a second leak to find.
- Ventilate with clean air until oxygen is back at 20.9% and stable, then re-test at the work location.
- Withdraw the instrument for a check. A catalytic LEL sensor exposed above 23.5% oxygen may be out of specification; calibrate before it is used again.
- Report and record. The permit is suspended, not amended, by the AGT. Re-issue needs the permit issuer, a fresh gas test, and a change to hose routing so the manway lip cannot abrade the line again - a sleeve, a proper pass-through, or a different entry point.
The order matters: killing the ignition source and removing the person come before diagnosing the leak, because the next spark is the one that matters.
3. How should the monitoring regime have been set up so that this was caught earlier, or prevented?
Read the worked answer
The regime was not matched to the hazard.
- Spot checks every 20 minutes cannot catch a hose failure. A line can split in seconds and the atmosphere can move from normal to dangerous in a couple of minutes. The entrant should wear a personal monitor with a continuous display and an alarm set at the permit limit (23.5% oxygen) with a pre-alarm below it. The AGT's readings at the manway are an addition, not a substitute.
- Position matters. The reading was taken at the manway, near the ventilation outlet, not at the work location. Sample where the fitter is breathing and where the torch is burning.
- Ventilation can mask the problem. Clean air from the duct was diluting the leak, so the true release rate was higher than the number suggests, and any interruption to the blower would have made conditions worse very quickly.
- The trend matters more than the snapshot. Two readings 20 minutes apart with nothing in between hides the shape of the rise; a minute-by-minute log or a datalogging instrument shows the leak starting.
- Pre-task checks would have prevented it: hose condition and leak test before entry, protection of the hose run, flashback arrestors at both ends, and a rule that hoses do not cross sharp or load-bearing edges.
The record should show the reasoning, not just the numbers: fixed-interval spot sampling suits a static, well-mixed space and is inadequate for a space containing a live, pressurised gas source.