Failure mechanism
Fatigue failure in welded steel structures
Fatigue is damage accumulated by repeated stress, and what drives it is the stress range — the difference between maximum and minimum stress in a cycle — together with the number of cycles, not the peak stress on its own. A detail can be safe under a static load that never varies and fail under a much smaller load that repeats a few million times. Welded details are the usual origin, because a weld toe is a built-in stress concentration and a built-in crack starter rather than a smooth surface.
How this failure works
Fatigue is not a strength problem, it is a damage-accumulation problem. A member held at a constant stress, however high within its static capacity, does not accumulate fatigue damage; a member subjected to a smaller stress that varies, repeatedly, eventually cracks. The governing quantities are the stress range — maximum minus minimum within the cycle — and the number of cycles applied at that range, so a detail with a few very large cycles and a detail with millions of small ones can both be critical, for the same reason. Mean stress matters secondarily and most obviously when part of the cycle goes into compression, because a compressive range closes a crack instead of opening it and contributes little or nothing to growth.
The design method is a family of S-N curves: for each detail, a relationship between stress range and the number of cycles to failure, derived from test data and expressed as a detail category. The consequence that surprises people is that the curve is not a property of the steel grade. A higher-strength steel has the same fatigue detail category as a mild one, because fatigue life is set by the detail and the crack it develops, not by the yield strength. Two structures in the same grade, carrying the same load, can differ in fatigue life by a large factor because one detail is several categories below the other.
A welded connection is a continuous path for stress, and it contains at the toe of every fillet or butt weld both a geometric discontinuity and a metallurgical one. That is where cracks initiate: the weld profile concentrates stress, the heat-affected zone has a different microstructure, and the process leaves intrusions, undercut and slag lines that behave as starter defects. A bolted connection interrupts the path, has its own concentrations at the holes, and distributes load across a group of fasteners; more importantly it can be inspected, tightened and replaced. A welded detail has none of that recourse. It can only be found before the crack outgrows the section, and by then the inspection interval has to have been short enough.
A fatigue crack initiates at the toe, grows a little with each cycle at right angles to the principal stress, and leaves a surface that looks different from a static fracture: smoother where the crack propagated under repeated opening and closing, rougher where the remaining section finally tore. Inspection intervals therefore have to be set against the crack growth that the detail would tolerate, not against the convenience of the calendar — a crack that can reach a critical size between two inspections is not being managed by them. This is why fatigue-critical details are identified at design stage, why an inspection method has to be able to find a surface-breaking crack at a toe rather than a change of section, and why a crack found in one detail is a finding about every similar detail in the same structure.
Illustration
Conceptual illustration of a fillet weld with a crack initiating at the toe and growing through the section, shown beside the S-N curve that sets its life.
Typical causes
- A detail with a low fatigue category — a fillet weld across the line of stress, an unstiffened attachment, a cope or a notch — placed where the stress range is high.
- Stress range underestimated because the load model took the maximum load but not the number of cycles at each level.
- Cyclic loading from cranes, machinery, vehicles or vibration that the original design did not consider at all.
- Weld defects — undercut, slag, porosity, an abrupt toe profile — at a location where stress is already concentrated.
- Out-of-plane distortion or misalignment at a joint, which raises the local stress at the toe well above the nominal stress.
- Load applied repeatedly at the end of a stiffener or attachment, where the detail terminates and the stress has nowhere to flow.
- A modification that welded a new attachment or bracket onto a member already carrying cyclic stress, without re-checking the detail.
Warning signs
- A hairline crack running away from a weld toe at right angles to the stress, easily mistaken for cracked paint.
- Rust staining in a fine line along a weld toe, which is often a crack holding moisture rather than a coating failure.
- Cracks beginning at the end of a stiffener, at a cope, or where an unstiffened attachment terminates.
- Movement or vibration that can be felt at a member or a connection, which is the cyclic stress the detail is actually seeing.
- A crack in the parent metal beside a weld rather than in the weld itself, which points to the toe or the heat-affected zone.
- A crack that reappears in the same place after being ground out and repaired, which means the detail itself was never changed.
Inspection methods
- 01Identify the fatigue-critical details from the design intent — every welded attachment, termination and cope on a member seeing cyclic load — and inspect those first and closest.
- 02Inspect the weld toe specifically, since that is where initiation occurs, rather than the weld face or the member in general.
- 03Clean the surface and remove coating locally where a crack is suspected; a hairline crack at a toe is not visible beneath an intact film.
- 04Use a method capable of detecting a surface-breaking crack at a toe, such as magnetic particle or dye penetrant inspection by a competent party, and record findings against the detail.
- 05Monitor any known crack with a fixed gauge, marked ends and a date, so that growth is measured over time rather than compared with memory.
- 06Compare findings across the structure, because fatigue is a detail problem: one cracked toe makes every identical detail a candidate for the same inspection.
Preventive measures
- Design the detail and not only the member: select details with a fatigue category adequate for the stress range and cycle count, and record which details are fatigue-critical.
- Avoid attachments welded across the line of stress, and terminate stiffeners and gussets away from regions of high stress range.
- Control weld quality and toe profile, because an abrupt or undercut toe is where the crack will start.
- Control fit-up, since misalignment and out-of-plane distortion raise the local stress above the value the design assumed.
- Consider post-weld treatment of the toe where the detail category is marginal, and treat it as a designed measure rather than a repair.
- Set inspection intervals from the crack growth the detail could tolerate, and state the method that will actually find it.
- Re-check the fatigue detail whenever a change adds load, cycles or a welded attachment to an existing member.
Relevant standards
The codes and specifications a design is normally judged against. Naming them is not a claim that AKEN certifies, inspects or approves against them — the design authority for a project does that.
- IS 800
- AWS D1.1
- EN 1993
- AISC 360
Related AKEN services
Scope note
This article explains a failure mechanism for engineering education. It is general guidance, not a design check, an inspection report or a certification for any particular structure. Assessments of an actual structure must be carried out by the responsible design authority against the project’s own specifications.