Numerical methods for the fatigue assessment of welded joints: influence of misalignment and geometric weld imperfections

The fatigue design life of welded joints in steel structures is increasingly assessed by using numerical models and methods, such as the structural (hot-spot) stress method and the effective notch stress method. When compared to the classical design approach using nominal stress S-N design curves, these methods offer the advantage of flexibility and a wider scope of application. However, a number of questions arise when these methods are used to assess geometrically "imperfect" welded joints, such as joints with plate misalignments or excessive weld convexity or concavity. In these cases, the classical S-N curves are known to cover imperfections up to the common tolerance classes for fatigue-prone welded joints (e.g. in accordance with ISO 5817 class B). For the numerical methods, differing and conflicting recommendations exist on how to account for the geometric imperfections in the welded joints, with little or no background to these recommendations available. In this paper, a study is presented in which two standard welded joints (butt welds between plates of equal and unequal thickness; T-joints with fillet welds) are analysed with the help of the structural (hot-spot) stress and the effective notch stress approach, considering various levels of geometric imperfection up to the tolerance limits, and the resulting fatigue life predictions are compared to test results from the literature and the nominal stress approach predictions. Since the nominal stress approach curves are based on reliable statistical data and desired survival probabilities for these known, standard cases, this methodology allows one to determine the correct application of the numerical methods to cases with geometric imperfections. This information may be used for a pertinent refinement of design recommendations for these methods, as well as for cases where these methods are applied to fitness-for-purpose assessments - e.g. because the nominal stress approach is not applicable.