1989–2009 commercial aircraft accidents attributed to NDT failures

Tom cites an Air Force study (approximately 20 years prior to the lecture, so circa early 1990s) establishing that flaws at 1/16 inch are detected with less than 50% probability, while flaws at 1/8 inch are detected with ~95% probability. The structural welding code had already settled on the 1/8-inch threshold from accumulated experience before the Air Force study formally validated it.

The reason for going through observables is — you can talk about accuracy and precision, but you ultimately have to bring it back to reality. To give you a reality check: a lot of the boiler and pressure vessel code and the structural welding code were built up by 80 to 100 years of experience and committees of people working together in good faith. There are certain things in the welding code — they're never looking for a flaw smaller than an eighth of an inch. Why? The Air Force did a big study, and you can't find flaws with more than 50% accuracy at a sixteenth of an inch. You have like a 95% probability of finding a flaw at an eighth of an inch. Before the Air Force did this study twenty years ago, the code had already decided: don't look for flaws smaller than an eighth of an inch because they're not going to be harmful. They didn't actually say it that way, but no one had ever had a problem with a vessel that might have had flaws of that size.