Does a part that fails before its rated fatigue life have defective metal?
Not necessarily: a component that fails before its stated fatigue life invites the straightforward inference that the metal was not what it should have been, and sometimes that inference is right. More often, though, the stated fatigue life was a statement about a set of loading assumptions, and the loading the part actually saw was not the loading assumed.
Fatigue life is not a property of a material in the way that yield strength is. Fatigue life belongs to a material, a geometry, a surface condition, an environment and a load spectrum considered together, and a premature-failure claim is a claim about all five.
What is the difference between high-cycle and low-cycle fatigue?
In high-cycle fatigue the bulk material stays elastic and fatigue life is governed by stress amplitude, while in low-cycle fatigue each cycle imposes plastic strain and fatigue life is governed by strain amplitude, so the distinction between the two is not merely one of cycle count. The laboratory basis for high-cycle fatigue is force-controlled testing of the kind described in ASTM E466; for low-cycle fatigue the relevant practice is strain-controlled testing, as in ASTM E606. High-cycle and low-cycle fatigue are different problems: applying a stress-life curve to a component that is actually cycling plastically will overstate the component’s fatigue life, sometimes by a wide margin.
Does every metal have a fatigue endurance limit?
No — the fatigue endurance limit is narrower than it is usually quoted: many ferritic steels show one, but aluminum alloys and many other non-ferrous materials do not. Many ferritic steels show a plateau in the stress-life curve below which laboratory specimens survive very large numbers of cycles, and that plateau is what people mean when they say a part will last indefinitely below a certain stress. Aluminum alloys and many other non-ferrous materials show no such plateau; their stress-life curves continue downward, so there is no stress that is safe without reference to a cycle count. Carrying a steel endurance-limit assumption across to a non-ferrous part is a recognizable design error.
What happens to the fatigue endurance limit under corrosion fatigue?
Corrosion fatigue removes the endurance-limit plateau entirely: where a corrosive environment acts at the same time as cyclic stress, the fatigue endurance limit effectively disappears and initiation shifts toward pits and other environmentally produced surface features. Corrosion fatigue is also frequency-sensitive in a way that purely mechanical fatigue is not. Slower cycling gives the corrosive environment more time to act within each cycle, so fatigue data generated at high laboratory frequency can be optimistic relative to service. A fatigue life prediction is not conservative merely because it used a published curve.
How are cycles counted for thermal fatigue?
Thermal fatigue counts operating events, not revolutions: start-ups, shutdowns, quench events and process upsets each impose a cycle, so the relevant count is of operating events over the life of the equipment. Constrained thermal expansion produces cyclic strain with no external mechanical load at all. The count of thermal events may number in the thousands where a mechanical cycle count would run far higher, which places most thermal fatigue in the low-cycle regime. Counting events rather than revolutions also means that operating and maintenance logs, rather than run hours, are the primary source for the thermal fatigue cycle count.
What is a rated fatigue life actually built from?
A rated fatigue life, or design life, begins with specimen data and is then adjusted for the things specimens do not represent: surface finish, size, geometry, mean stress, temperature and the required reliability. Design codes that contain fatigue rules apply their own margins on both stress and cycles, and those margins exist to cover scatter and uncertainty rather than misuse. The number that emerges is a statement about a population under an assumed load spectrum, not a warranty on an individual part under whatever spectrum it encountered.
Is scatter in fatigue life a sign of a defect?
No — fatigue data scatter widely, and that scatter is intrinsic, not a sign of a defect. Nominally identical specimens tested at the same stress amplitude routinely differ in fatigue life by a large factor, because initiation depends on the worst feature in the loaded volume rather than on average properties.
Because of that scatter, one early fatigue failure in a fleet is weak evidence of nonconformance on its own. A cluster of early fatigue failures sharing a common origin type is much stronger evidence, which is why field-return data and the condition of unfailed sister parts belong in a premature fatigue failure analysis.
How are the load cycles a component actually experienced counted?
The load cycles a component actually experienced are counted by reducing its irregular service loading history to equivalent cycles, using the cycle-counting practices set out in ASTM E1049, rainflow counting among them. Real service loading is rarely constant amplitude, and comparing it against constant-amplitude fatigue data requires that reduction. The counted spectrum is then combined with a life curve through a cumulative-damage rule. Cumulative-damage rules are approximations that take no account of the order in which loads arrive, which is a known limitation rather than a hidden one.
Where no measured loading history exists, the service loading history has to be reconstructed from process data, operating logs or strain measurement on a comparable unit in service. That reconstruction is usually the most contestable part of counting the fatigue cycles a component actually experienced.
Why is a premature fatigue failure a question about loading?
A premature fatigue failure is a question about loading because resonance and insufficient damping, misalignment, a commissioning overload, transport damage, an unanticipated transient, or simply a duty heavier than the one specified will each produce failure well inside a correctly calculated fatigue life without any defect in the metal. Once the material has been verified against its certification by composition, hardness and tensile testing, a premature fatigue failure becomes a question of whether the service loading resembled the design loading. Establishing which of those loading causes occurred is a loading investigation as much as a metallurgical one.
What settles whether a premature fatigue failure came from the material or the loading?
In practice, the answer to whether a premature fatigue failure came from the material or the loading comes from three sources read together: the physical evidence on the part, the verification of the material against its specification, and the reconstruction of what the component was actually asked to do. Where all three agree, the conclusion is stable. Where the loading history is thin, the honest position is that the metallurgy is consistent with more than one explanation, and saying so is more durable than choosing between them.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.