PSI - Issue 75

Georg Veile et al. / Procedia Structural Integrity 75 (2025) 184–192 Georg Veile / Structural Integrity Procedia (2025)

187

4

This works focus is the fatigue life assessment of weld joints by implementing the real weld topology into FEA. In Fig. 2 the procedure is illustrated. For a detailed description of the scan procedure reference is made to (Lotz et al. 2025; Veile et al. 2025) . The scan was conducted with GOM’s ATOS system with an accuracy of min. 12 µm (Mendricky e Sobotka 2020). The scan was polygonised and imperfections like holes were corrected manually. Subsequently, the discontinuous polygonised surface was converted into continuous surfaces using Non-Uniform Rational B-Splines (NURBS) patches.

Fig. 2. Procedure of implementing a scanned weld geometry into a 3D FEA model with sub-modelling technique from (Veile et al. 2025).

The implementation and calculation of scanned weld topologies require significantly more computational power than FEA with fictive weld radius. Each scanned weld topology was meshed and refined until reaching numerical convergence. In order to limit numerical calculation time, the sub-modelling technique was applied. Different multilinear-kinematic elastic plastic material models were used for the nuclear grade base metals (AISI 347 and AISI 304L) and weld metal ER 347. The material data was based on experimental, stain-controlled fatigue test at 300 °C (Rudolph et al. 2024). Fatigue life assessment was conducted as described in (Veile et al. 2025). 3. Effects of smoothening the weld topology As described in (Veile et al. 2025; Lotz et al. 2025; Veile 2025), the scanned weld surface is saved digitally as discontinuous polygonised surface and then converted into a continuous surface by means of NURBS. This continuous surface is then converted into a solid geometry for FEA in Ansys Mechanical 2025 R1. The selected parameters, such as the number of NURBS or acceptable tolerance, impact the local radii of the scanned weld topology. This process can also be described as smoothening the weld topology. By changing these parameters, different versions with variations of their local weld radii were created. The “runouts” of (Veile et al. 2025) were not considered in the investigation of this work. The precision of predicted fatigue life can be best described with the analytical function of deviation (see Eq. (8)) (Veile et al. 2025; Braun et al. 2022). = log −log (8) In this work, different quantities of NURBS were used to achieve smoothening. A reduction of NUBRS should lead to a reduction in local radii. Nevertheless, if base points of these NURBS were placed in areas of small radii, this radius could remain small. An alternative way to smoothen the local radius is the manual displacement of this base points. This was necessary to achieve radii over 0.45 mm. The algorithm (open source) of (Dänekas 2024; Dänekas et al. 2025) was used to determine the weld radius in this work. The published results of this work show radii increased up to 0.55 mm. In Fig. 3 the influence of smoothening the topology by means of NURBS patch quantity and displacement of the base points is illustrated. One can observe that the deviation of gradient-based FDP is not influenced by smoothening the topology to a significant amount. In contrast, common FDP improve in accuracy. Furthermore, it can be stated that the scatter is also reduced. This confirms the hypothesis, that an increase in scatter compared to fatigue life prediction with idealized weld geometry is based on geometrical factors, such as the local weld radii. However, it must also be mentioned that this should also be true for advanced FDP. No explanation for this observation was found yet.