Fatigue is a slow, quiet failure - the kind that arrives years after handover, at a detail nobody flagged, under loads that never once came close to the static capacity. It is driven not by how large a stress is, but by how many times it cycles. For any structure that sees repeated loading - a crane runway, a bridge, a vibrating support - fatigue can be the check that actually governs the connections.
Stress range, not peak stress
The single most important shift in thinking is that fatigue cares about the stress range - the swing between maximum and minimum in each cycle - not the peak value. A detail can sit far below yield and still crack, simply because it is cycled millions of times. This is why a member that passes every static check with ease can still have a fatigue problem hiding in one weld.
The detail category
Every weld geometry is mapped to a detail category - a number such as 80, 71 or 56 - that defines its S-N curve, the relationship between stress range and the number of cycles to failure. The number is, in effect, the stress range (in N/mm²) the detail can survive for a reference number of cycles. A higher category is a better detail.
The categories follow geometry and workmanship, and the differences are large. A ground full-penetration butt weld is a high category; a fillet weld toe, where stress concentrates sharply, is a much lower one; a poorly accessible weld with a likely root defect is lower still.
The detail category is set by the geometry and the quality of the weld, not by the grade of the parent steel. Choosing S355 over S275 does nothing for fatigue life.
That last point surprises people: you cannot buy fatigue resistance with a stronger steel. You buy it with better detailing - smoother geometry, full penetration where it matters, grinding to remove the sharp toe.
Where the curves and reality part ways
The S-N curves come from controlled laboratory specimens. Real structures rarely match those conditions, and that gap is where field cracks live:
- Undercut at the weld toe creates a sharper notch than the test specimen assumed.
- Porosity or slag inside the weld seeds an internal crack.
- Misalignment at the root adds a secondary bending stress the nominal calculation never included.
This is why inspection of fatigue-critical welds matters as much as the calculation. A detail designed to category 71 but executed with a 1 mm undercut is no longer a category 71 detail. The number on the drawing is a promise the fabrication has to keep.
Designing against it
Three habits protect fatigue-loaded connections:
- Identify the cyclic load paths early and route them through high-category details where you can - butt welds over fillet welds, smooth transitions over abrupt ones.
- Specify the weld quality the category assumes, and make the fatigue-critical welds accessible for inspection.
- Do not rely on a strength check alone for repeatedly loaded members; the static section and stability checks can all pass while fatigue quietly governs.
Fatigue rewards the engineer who thinks about geometry and the inspector who looks closely - and punishes the one who trusts a static utilisation ratio on a member that never stops moving.