Short Cracks, Crack Closure & Threshold Behaviour

Technical provenance

Applicable standards / specifications

• ASTM E647 — Standard Test Method for Measurement of Fatigue Crack Growth Rates
• ASTM E1820 — Standard Test Method for Measurement of Fracture Toughness
• BS 7910 (2019) — Guide to methods for assessing the acceptability of flaws in metallic structures — Fitness-for-service / flaw-assessment reference; project-specific acceptance requirements govern.

References

• Anderson, T. L. — Fracture Mechanics: Fundamentals and Applications — Background reference for LEFM, elastic-plastic fracture mechanics, crack driving force and fracture assessment.
• ASTM E647 — Standard Test Method for Measurement of Fatigue Crack Growth Rates — Test-method reference for fatigue-crack-growth data.

Why Short Cracks Are Different

Conventional long-crack growth data are measured using cracks large enough for fracture-mechanics assumptions and closure behaviour to be established. Small physically short cracks can propagate at nominal ΔK values below the long-crack threshold and can grow faster than long-crack correlations predict. Microstructural scale, notch fields, residual stress and limited crack closure all contribute. This matters when a damage-tolerance calculation starts from a very small assumed flaw.

Physical, Microstructural and Mechanically Short Cracks

A crack can be short relative to grain size, relative to the plastic zone, or relative to the surrounding stress gradient. These are different reasons for conventional long-crack similitude to fail. A crack at a notch root, for example, may experience a rapidly decaying local stress field and significant plasticity even when its absolute length is not microscopic. The analyst should identify which short-crack mechanism is relevant rather than using one generic correction.

Crack Closure

Plasticity, roughness, oxides and residual deformation can cause a fatigue crack to remain closed through part of the nominal load cycle. The effective crack-driving range is then smaller than the nominal ΔK range. Closure evolves with stress ratio, overload history and crack length. Models that use ΔKeff can represent this physics, but closure should not be double-counted if the material growth data or empirical growth law already incorporate the relevant stress-ratio effects.

Threshold Is Not a Universal No-Growth Boundary

The measured long-crack threshold ΔKth depends on stress ratio, environment, frequency, load history and test procedure. Near-threshold growth rates are extremely sensitive to small changes in these conditions. Treating one handbook threshold as a hard endurance limit can therefore be unsafe, particularly for corrosion-assisted growth, high R-ratio loading or small cracks.

Starting-Flaw Selection

If the assumed initial flaw is below the range for which long-crack data are valid, the analyst has several options: choose a larger justified initial flaw, use a short-crack model, use conservative no-threshold growth, or demonstrate that the small-crack phase is negligible relative to total life. The choice should be linked to inspection/manufacturing evidence and the governing damage-tolerance philosophy.

Notches and Stress Gradients

At a notch, the local elastic Kt can produce severe initial crack driving force, but the field decays as the crack moves away from the notch root. Short-crack growth should therefore account for the evolving local stress rather than applying a constant notch-amplified nominal stress indefinitely. Weight functions, local stress distributions or dedicated cracked-notch models can be used to capture this transition.

Overloads and Retardation

An overload can enlarge the crack-tip plastic zone and increase subsequent closure, producing retardation after the overload. For a small crack or a crack in a steep stress gradient, the retardation length scale may be comparable with the crack size itself. Sequence-sensitive models therefore need adequate crack increment resolution and a load history that preserves event order.

Sensitivity Strategy

• No-threshold versus measured-threshold assumptions
• Initial flaw size and aspect ratio
• Stress ratio and residual stress
• Environmental reduction of threshold
• Closure/retardation model on and off
• Local notch-stress gradient representation

Verification

Short-crack predictions are difficult to validate directly, so engineering conservatism and bounding become important. Compare against long-crack data at larger crack sizes, verify smooth transition between regimes, check that the model does not create an artificial growth discontinuity and use relevant test or service evidence where available.

Microstructure and Small-Crack Scatter

At crack sizes comparable with grains or other microstructural features, local crystallography and barriers can dominate growth. Individual small cracks may therefore show substantial scatter, temporary arrest or bursts of growth that are smoothed out in long-crack data. Structural assessments normally do not model grains explicitly; instead they use conservative short-crack correlations, larger assumed initial flaws or scatter factors that avoid taking credit for microstructural arrest.

Residual Stress Can Remove Apparent Threshold Margin

Compressive nominal loading does not guarantee a low crack-driving force if tensile residual stress exists at the crack location. Residual stress raises Kmax and effective stress ratio and can suppress closure, allowing growth at applied ranges that would appear sub-threshold using nominal load alone. This is especially relevant at welds, cold-expanded holes and locally plastically worked features.

Environmental Sensitivity Near Threshold

Near-threshold behaviour is particularly sensitive to humidity, corrosion products, hydrogen and loading frequency. An air-laboratory ΔKth may therefore be inappropriate for long-duration service in a damaging environment. When threshold assumptions dominate predicted life, use environment-representative data or perform a conservative no-threshold/low-threshold sensitivity so the decision does not rest on an unverified laboratory boundary.

Transition to Long-Crack Modelling

A short-crack model should transition smoothly into the conventional long-crack relation as the crack becomes large relative to the microstructure, plastic zone and local stress gradient. Check both growth rate and driving-force continuity at the transition. An artificial jump in da/dN is a modelling artefact and can distort life materially if the transition falls in a slow-growth portion of the history.

Initiation Life Versus Propagation Life

When short cracks dominate, the conventional boundary between fatigue initiation and fracture-mechanics propagation becomes blurred. A notch-strain method may count microstructural crack formation as initiation while a short-crack model starts propagation much earlier. Avoid counting the same physical period twice. Define clearly where the fatigue-initiation calculation ends and where the crack-growth calculation begins, and check that the chosen transition flaw is consistent with both methods.

Inspection Relevance

Very small cracks may be analytically important but well below practical NDT capability. Damage tolerance therefore cannot rely on detecting them directly. The analysis should demonstrate adequate growth time from an assumed small flaw to a reliably detectable size and then from detectable to critical size. If those intervals collapse because short-crack growth is rapid, inspection-based damage tolerance may not be viable and a safe-life or design-change response may be needed.

Key Engineering Principle

The threshold region is where small modelling assumptions can produce very large changes in predicted life. A robust damage-tolerance argument therefore avoids relying on a single precise ΔKth unless its applicability is well supported and shows how the conclusion changes when short-crack and closure assumptions are varied.

Key takeaways

• Short cracks can propagate below conventional long-crack thresholds.
• Closure changes effective crack driving force and is history-dependent.
• ΔKth is conditional on R-ratio, environment and test method.
• Starting flaw assumptions should be consistent with the growth model validity.