I have, for the first time, actually thought about cylinder failure in detail, and even managed to do a few sums. Going through the posts above, I would comment as follows (sorry if it is too “techno-babbly”):
• The hoop stress (greatest stress) on a 12l cylinder at WP is ~410 MPa (assuming a wall thickness of 5mm). This is barely at the lower limit for high cycle fatigue, which would require you to fill and empty a cylinder every day for more than 28 years! And then it only just counts as high cycle fatigue. Even at the test pressure (610 MPa) fatigue on its own is still not really an issue.
• The only way to affect the crystal structure is to really belt the cylinder (drop it down a pitch), and that will generate local deformation, or heat it to >750°C or so. Heating it to >500°C for a period will temper it back and drop the strength. Then all bets are off.
• The hydrostatic pressure test is designed to work within the elastic region of the steel’s properties, and as such there will be no lasting effect from the test. Basing on a consistent test pressure (from one test to the next) the only way you can get a change in the degree of expansion during the test is by a thinning of the wall (either local or general).
• ‘Stretching’ (i.e. plastic deformation) can only occur if the yield strength of the steel is exceeded. The hoop stress on the neck (threaded region) of the cylinder is lower than the rest of the cylinder, and so circumferential stretching is even less likely. Over-tightening of the tap in the threads seems a little strange to me for modern, parallel threaded valves. They have a shoulder which I would expect to prevent over-tightening. This is not the case with old taper threads. I do not doubt that this is the main cause of hydro test failures, but I do wonder if it is sometimes associated with manufacture – I had two taps that failed the go / no-go gauge on their first test.
The problem with all of the above is that it relates to a cylinder in good condition with no localised wall thinning, corrosion or defects. Manufacturing techniques are not perfect; corrosion is usually localised at an inherent weakness in the steel (and there will always be inclusions); a localised weakness (whether corrosion induced or residual from steelmaking / cylinder manufacture) will act to concentrate the stress, and may significantly reduce the pressure at which the cylinder with fail hydro test / go bang. As Andrew pointed out (and we all should know), the tap is the weakest point anyway.
Also, from a safety point of view, I think all of us would prefer a cylinder to fail an over-cautious hydro test than actually go bang underground.
Just as an aside, using the same assumption as I used for the above calcs, a 12l blown to 600 bar would approach / exceed the yield point of the steel and fail catastrophically.
Sorry, but this is very much the “short version”.
Marcus