A fatigue failure is two processes sharing one name. A crack has to start, and then it has to grow. In most components the growth phase is comparatively brief and the initiation phase consumes the bulk of the service life, which has a practical consequence for anyone trying to establish what went wrong. The answer is usually sitting at the origin, in a feature a few hundred micrometres across, rather than in the broad fracture surface that gets photographed at the scene.
Two processes sharing one name
Fatigue life is conventionally divided into the cycles spent nucleating a crack and the cycles spent propagating it to final separation. The split is not fixed — it depends on stress level, material and geometry — but for a component operating at modest nominal stress, the crack spends most of its existence very small. A part can be within a small fraction of its remaining life while carrying a crack that no routine inspection method would find.
This is why the fracture surface, dramatic as it looks, is often less informative than the few square millimetres at one end of it. The propagation record explains how the crack advanced; the origin explains why there was a crack at that location at all.
Finding the origin
Origins are locatable by convergence. Radial marks, ratchet marks and the curvature of successive crack-front positions all point back toward the initiation site, and low-magnification examination under raking light usually narrows it to a small area before any sectioning is contemplated. Multiple ratchet marks along an edge indicate several origins that started independently and later joined, which is itself a finding: it generally implies a high local stress or a broadly damaged surface rather than a single isolated defect.
What sits at the origin
The origin feature usually falls into a limited set of categories. Machining and grinding marks transverse to the stress direction, forming laps and seams, non-metallic inclusions, porosity, corrosion pits, fretting scars, plating cracks, weld toes, and mechanical damage from handling, stamping or assembly. Each of these implicates a different stage of the part's life, which is why identifying the feature matters more to a dispute than confirming that fatigue occurred.
Distinguishing them requires magnification. At the scale that matters, a tool witness mark and a corrosion pit are not confusable, and an inclusion has a composition that energy-dispersive analysis will report directly. What is confusable is an origin that has been rubbed, oxidised or wiped before anyone looked at it.
Fatigue is a surface phenomenon
Because the maximum stress in bending and torsion occurs at the free surface, and because a surface grain has fewer constraints on slip than an interior one, surface condition carries disproportionate weight. Roughness, grinding burn, decarburisation and tensile residual stress from machining all shorten initiation life; shot peening, rolling and nitriding lengthen it by putting the surface into compression. A surface treatment specified but not performed, or performed to the wrong parameters, is a recognisable and testable proposition rather than a matter of opinion.
Subsurface origins
Not every origin is at the surface. In high-strength steels cycled at low amplitude, and in parts carrying an effective compressive surface layer, initiation can occur at a subsurface inclusion, producing a characteristic facet around the initiating particle. That shifts the question from surface processing toward melting and cleanliness practice. Inclusion content can be rated on polished sections and compared against the material specification and against unused stock from the same lot, which is often the only way to establish whether the population was atypical.
Geometry decides where, stress decides whether
Stress concentrations do not create fatigue on their own, and neither does a surface defect on its own. Initiation happens where the two coincide: a machining mark inside a fillet, a pit at a section change, a fretting scar under a clamped joint. Where the origin sits at a geometric feature with no defect present, the geometry is the finding; where it sits at a defect on an otherwise unremarkable surface, the defect is.
Built in or acquired in service
The categories separate along a useful line. Inclusions, porosity, machining marks, decarburisation and plating defects were present when the part entered service. Corrosion pits, fretting scars, impact damage and thermal cracking arrived later. That distinction frequently maps onto the dispute more directly than the failure mechanism does, and it is testable against exemplars: an unfailed sister part from the same batch will carry the built-in features and, if it saw comparable service, some of the acquired ones as well.
What destroys an origin
The initiation site is small, shallow and easily damaged. Bringing the mating faces back into contact grinds it, wire-brushing removes it, light rusting obscures it within days in humid storage, and solvents or preservatives leave deposits that interfere with elemental analysis. Sectioning through the origin before it has been documented intact removes the option of examining it whole. Separate the pieces, keep them dry, keep anything off the fracture faces, and photograph before anything else happens.
What the origin does not settle
Identifying the initiating feature establishes where the crack started and what gave it somewhere to start. It does not establish the stress that was acting, the number of cycles applied, or whether the feature was within specification. Those come from the loading history, the drawings and the material records.
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.