Stanislaw Kowalski

2024-10

Friction stir weld (FSW) fatigue characterization — designed specimen program for FSW ductile aluminum, ran fatigue testing, derived S-N curve and compared with conventional MIG weld baseline.

Challenge: FSW specimen failure locations varied — some at the heat-affected zone, some at the weld nugget root. Mixed failure modes complicate S-N curve fitting. Segregated by failure location before fitting — FSW toe failures showed 1.5 FAT class improvement over MIG.

Learned: FSW produces a better fatigue performance than MIG at the same joint configuration because it eliminates the weld root defect and reduces residual stress. But the mixed failure modes mean that a single S-N curve oversimplifies the behavior.

2024-04

Residual stress effects on weld fatigue life — compared as-welded vs post-weld heat treatment (PWHT) fatigue lives using modified Goodman approach with measured residual stress from X-ray diffraction.

Challenge: Measured residual stresses from X-ray diffraction were only available at the surface. Sub-surface residual stress profile (needed for Goodman correction) had to be estimated from the Osgood relaxation approach — significant uncertainty.

Learned: Residual stress effects on weld fatigue are substantial and direction-dependent. Tensile residual stress (as-welded) reduces fatigue life by 30-50% compared with stress-relieved state. PWHT can recover most of this life, but the heat treatment temperature and time must be controlled carefully.

2023-11

FAT class assignment for complex weld details not covered by standard EN 13001 tables — lap joints with load-carrying fillets, partial-penetration butt welds, tube-to-tube joints.

Challenge: For weld details not in the standard tables, an engineering decision is needed on which standard detail most closely resembles the actual weld. Wrong FAT class assignment can be 2-3 FAT classes off, changing life prediction by an order of magnitude.

Learned: When weld details fall between standard FAT classes, always use the more conservative class unless test data justifies otherwise. The consequence of a non-conservative FAT class assignment in fatigue is weld failure below the predicted life.

2023-07

Hot spot stress method vs structural stress (Dong method) comparison for weld toe fatigue assessment. Applied both methods to the same set of 6 weld details and compared against test failure data.

Challenge: Dong structural stress method is theoretically mesh-insensitive but my FEM implementation still showed 12% mesh sensitivity for fine vs coarse elements. Root cause: incorrect linearization across element thickness — had to use through-thickness stress integration, not surface extrapolation.

Learned: Dong's structural stress method linearization must be done correctly over the full plate thickness to be mesh-insensitive. Surface-only linearization misses the nonlinear stress component and reintroduces mesh sensitivity.

2023-02

Weld fatigue assessment of engine test stand structure per EN 13001-3-1 — classified all welds by FAT class, computed hot spot stresses from FEM using surface extrapolation, checked against FAT S-N curve.

Challenge: EN 13001 hot spot stress extrapolation requires specific mesh refinement rules at weld toes that most engineers are not aware of. Initial mesh gave hot spot stresses 35% lower than a correctly refined model — non-conservative. Re-meshed all critical weld toes to IIW guidelines (element size ≤ 0.4t).

Learned: Hot spot stress FEM requires strict adherence to the IIW meshing guidelines at weld toes. Coarser meshes underestimate the hot spot stress significantly. Document the mesh quality at every weld toe assessed.