One of the time-honored conventions of aviation journalism is that every article about metal fatigue has to start with mention of the "ill-fated Comet."
That requirement has now been met.
Metal fatigue did not begin with the Comet. By the middle of the 19th century ferrous metals-irons and steels-were in common use in industrial machinery, and it had become apparent that components that were amply strong for the work they were required to do were nevertheless liable to break after long use. In 1849 researchers conducted the first systematic investigations of what would come to be termed metal fatigue, and confirmed what countless shop workers already knew: that metal parts that undergo many repeated strains eventually fail at a loading far below their original ultimate strength. By the 1870s, systematic tests of various alloys had been conducted under repeated loads of different types, and the behavior of metals in fatigue-producing conditions had been well documented. What was happening was apparent; why and how it was happening was not yet understood.
The basic facts were simple. There is, to begin with, a level of loading below which no appreciable fatigue takes place in ferrous metals. Called the fatigue limit, it is around a third to a half of the tensile yield strength, depending on the material. Repeated applications of higher stresses lead to eventual failure, with the intensity of the stress, number of cycles, order of high- and low-stress events, alloy types, and many other factors, including even surface finish, corrosion and minor manufacturing variations, determining the "fatigue life" of a component. Unlike iron and steel, aluminum does not have a clear fatigue limit; it is assumed to accumulate fatigue at any level of stress.
Paradoxically, fatigue does not cause a progressive loss of strength; a partly fatigued part remains just as strong as a new one until, at some point, cracks develop. Furthermore, until cracking begins there are no detectable signs of fatigue, which, according to the current hypothesis, build up at the invisible atomic or molecular level, not, as was originally guessed, at the microscopic (but visible) level of the metal's crystal structure. Fatigue tests typically involve inflicting millions, or hundreds of millions, of load reversals upon a sample part of standard shape and surface finish, and generating a curve of stress levels against cycles-to-failure. Data collected from generic test items are applied to all sorts of parts, using computational "finite-element" stress analysis to predict the stress at any location. Stress concentrations in complex parts can be predicted with reasonable accuracy, but since uncontrollable variables like bolt fits and the tightness of nuts and rivets can significantly affect stresses, the judgment and experience of the structural engineer remain an essential part of design.
Fatigue could be minimized simply by making load-bearing parts large enough. But flight vehicle structures need to be light. Instead, assumptions about service life and the frequency of various kinds of loadings are used to decide what fraction of its maximum strength a given part can be allowed to use, and how often.
