Why AM fatigue scatter is bigger than wrought, and how to plan for it

Wrought aerospace alloys typically show a fatigue-life scatter factor of 3–5× at a given stress level. AM coupons of the same nominal alloy commonly show 10–25×, and worst-case as-built surfaces can exceed 50× (Beretta & Romano, 2017; Tammas-Williams et al., 2018). This is the single biggest reason A-basis AM allowables run ~30% lower than wrought.

1. Defect population. Lack-of-fusion pores, keyhole pores, and entrapped gas pores act as initiation sites. Their size distribution is heavy-tailed, so a single coupon's life is dominated by the largest defect within ~500 μm of the surface. HIP closes most pores but does not heal lack-of-fusion above ~200 μm.

2. Surface state. As-built L-PBF surfaces have Ra ≈ 8–15 μm, with notch-like valleys that act as fatigue initiators. Machining the gauge surface typically halves the scatter and lifts the mean by 40–80%.

3. Build location and orientation. Coupons near the build plate edge see different thermal history than center coupons; build-direction-aligned columnar grains create anisotropic crack paths. Z-built (vertical) coupons are typically the weak axis in fatigue.

4. Heat-treatment variability. HIP cycle, solution time, and aging temperature vary between providers. Inconel 718 aged at 720 °C / 8h vs 760 °C / 10h shows a measurable shift in the mean S-N curve.

We compute basis values per MMPDS-2024 §9 procedure: minimum n = 75 coupons per condition, drawn from ≥ 3 builds and ≥ 2 providers when available. The k-factor for A-basis on n = 75 is 2.336 (one-sided 95/99 normal-tolerance); we additionally apply a non-parametric floor on small samples.

When a material × condition has fewer than 30 coupons in the database, the wizard returns a typical value rather than a basis value and flags it as 'design-allowable not derivable yet'. This is conservative on purpose — small samples on heavy-tailed defect distributions systematically under-estimate the tail.