Field failures are never the failure you predicted

Why the analysed failure mode is rarely the one that occurs.

Field failures are rarely the failure anyone predicted because a failure analysis can only examine the mechanisms someone thought to check, while a real machine operating in the real world is exposed to combinations of load, environment and small manufacturing variation that nobody sat down and deliberately imagined, and it is almost always one of those unimagined combinations that actually finds the weak point first.

A competition season is really a list of things that failed, and the list is more instructive than the result, since almost nothing on it was the failure anyone had spent time worrying about beforehand, and nearly all of it was something nobody had thought to check at all.

A picnic ruined by wasps

Planning a picnic carefully, checking the forecast for rain, packing a blanket, sun cream and enough food for everyone, and then having the afternoon ruined by wasps nobody thought to plan for shows the pattern in a lighter setting. Every precaution was reasonable and aimed at a plausible risk. The planning was sound, yet an afternoon outdoors can go wrong in more ways than any finite list of precautions can cover. The rain never came and the sun cream never mattered, because the day was spoiled by the one thing missing from the list. A more thorough planner would not have fixed that by packing more of the same things. A second umbrella and a larger tube of sun cream guard against the risks already covered, and the wasps would have arrived just the same.

Why a review's list always has gaps

A design review checks the failure modes its reviewers can think of, and a well-run review checks a great many: worst-case loads, expected fatigue cycles, known environmental extremes. Even so, the list is built from imagination and experience, and both have limits. A part can survive every scenario anyone tested it against and still fail under some combination nobody pictured. A vibration frequency nobody modelled can meet a resonance nobody calculated, or a small manufacturing variation in one batch can meet a load condition nobody thought to pair it with, or a fastener can loosen for a reason unrelated to the load it was sized against.

None of these is exotic on its own. Each simply falls outside the finite set of scenarios the original analysis covered, and a field failure is usually the visible evidence of that gap between what was imagined and what the world produced. A review's list also grows one item at a time, while the variety of ways a part can be loaded, exposed or subtly mis-made is effectively open-ended. That is why even a mature product with years of field history can still turn up a first-of-its-kind failure without anyone's standards having slipped.

Investigating with the list set aside

Once something does fail in the field, the temptation is to assume it must be a more severe version of a risk already known and analysed. That leads an investigator to look first in exactly the places a design review has already checked and cleared, while the real cause sits somewhere the original analysis never considered. A useful investigation starts by putting the list of analysed failure modes to one side and examining the failed part with fresh eyes for whatever evidence it actually shows: where the crack began, which surfaces rubbed, what was corroded or burnt, and what was not.

A mature engineering process treats every field failure, however small, as new information, and feeds each one back into future design reviews so that the list of scenarios considered slowly grows to cover ground the original imagination missed. Discovering, failure by failure, which unimagined combinations the real world produces is the only reliable way to narrow the gap between an analysed failure mode and a real one. For the picnic, it means the wasp spray goes in the basket next time, and the list grows by one line.

Diligence and imagination are different gaps

This offers no excuse for skipping the obvious, well-understood failure modes. A part that fails from a known, easily predicted mechanism, an undersized fastener or an obvious stress concentration nobody addressed, has failed from a gap in diligence, a separate problem from a gap in imagination. The unpredictability described here applies to genuinely novel combinations. The well-trodden failure modes are ones any competent review should have caught, and filing an avoidable failure under the same heading as an unforeseeable one is a comforting but inaccurate excuse for work a more careful review would have covered from the start.