A fabricator proposes a larger fillet to cure a fatigue problem at a stiffener end
A runway girder for an overhead crane has a transverse stiffener welded to a 12 mm web with a 6 mm fillet on both sides, terminating 50 mm above the tension flange. The contract is EN 1090-2, execution class EXC3, and the fatigue verification to EN 1993-1-9 uses a detail category appropriate to a welded stiffener end, with a stress range of 65 MPa at 2 × 10^6 cycles. On a previous crane of similar design, cracks started at the stiffener ends after about four years in service. The fabricator proposes increasing the fillet leg from 6 mm to 10 mm on both sides, arguing that "more weld metal means lower stress and the fatigue problem goes away". The designer has already signed the drawing with the 6 mm fillet, and the first two girders are tacked together.
1. Evaluate the fabricator's proposal technically. Does increasing the fillet size improve the fatigue resistance of this detail?
Read the worked answer
No. Fatigue at a stiffener end is governed by the local geometry of the detail - the abrupt change of stiffness where the stiffener terminates, the weld toe profile, the angle of the attachment and whether the weld is continued around the end - not by the nominal throat thickness.
The detail category in EN 1993-1-9 is assigned to the detail, not to the weld size. Increasing the leg from 6 mm to 10 mm makes the stiffener connection stiffer and the transition harsher, and adds more restraint and residual stress at the very location where the crack initiates. In practice it can reduce life, not extend it.
There is also a procedural problem: the designer has signed a drawing with a 6 mm fillet. Changing a weld size on a fatigue-critical detail is a design change, requiring the designer's approval and, under EXC3, an update of the drawing and inspection and test plan. The fabricator cannot make it on the shop floor.
The correct reasoning is: the crack in the previous crane confirms that this detail is fatigue-driven; the response must change the detail, not thicken the weld.
2. What detailing or design changes would actually improve fatigue performance here, and what approval and documentation would each need?
Read the worked answer
Options, in rough order of effect and practicality:
- Soften the stiffener end: taper the stiffener or add a rat-hole so the stiffness change is gradual. This changes the detail category and is a design change.
- Grind the weld toe to a smooth transition, which is a recognised fatigue improvement technique, but only if it is specified, executed and inspected as such. It cannot be assumed by the inspector.
- Change the connection type, for example a full-penetration stiffener-to-web weld if the detail demands it, or a bolted detail.
- Maintain the good feature already present: the stiffener stops 50 mm short of the tension flange, keeping the weld out of the highest stress region. This should not be lost in any redesign.
Each option is a design change requiring the designer's approval, revised drawings, a re-run of the fatigue verification with a detail category taken from the standard's catalogue, and an updated inspection and test plan. If toe grinding is adopted, the procedure, the profile acceptance and the inspection method must be written down, because the benefit depends entirely on execution.
3. Is the 6 mm fillet adequate in static terms? Give a rough check, and say why this does not settle the matter.
Read the worked answer
Throat a = 0.7 × 6 = 4.2 mm. Using EN 1993-1-8 for a fillet weld in S355: fvw,d = fu / (√3 · βw · γM2) ≈ 470 / (1.732 × 0.9 × 1.25) ≈ 241 MPa. For two 300 mm long fillets: resistance ≈ 241 × 4.2 × 600 ≈ 607 kN, far above the design force a stiffener normally transfers (typically tens of kN).
So the 6 mm fillet is statically ample with a large margin. That is exactly why the fabricator's argument fails: the member is not sized by static strength, it is sized by fatigue. Adding weld metal addresses a static concern that does not exist, while worsening the fatigue detail and adding distortion.
Practical points: weld balanced on both sides, use a back-step or skip sequence to limit angular distortion, and check girder straightness after welding. A 10 mm fillet would increase distortion and residual stress for no fatigue benefit, and would depart from the approved drawing.