PSI - Issue 75

Carl-Fredrik Lind et al. / Procedia Structural Integrity 75 (2025) 519–529 Carl-Fredrik Lind et al./ Structural Integrity Procedia (2025)

520

2

1. Introduction Structural durability and strength are critical in predicting the life of components across industries. In the automotive sector, where numerous welds are present, fatigue assessment of weldments is essential. Manual evaluation is time-consuming and prone to errors, making automation in FEM-based fatigue assessment both necessary and beneficial. Several methods exist for stress-based weld evaluation, including the nominal, hot spot, and effective notch stress methods, as outlined in the IIW recommendations (Hobbacher & Baumgartner, 2024). Nominal stress methods are sensitive to joint geometry and load type. Hot spot extrapolation improves this but remains mesh-dependent and limited in capturing localized stress (Dong, 2001). To address this, the Master S-N curve method introduces a mesh insensitive structural stress parameter (Dong et al., 2003), aligning with far-field stress concepts in fracture mechanics and reliably capturing stress concentrations across various joints and loadings (Dong & Hong, 2004). The effective notch method offers accuracy but requires fine meshing and complex modeling (Baumgartner et al., 2020). Meanwhile, machine learning (ML) methods are increasingly applied to fatigue prediction (Kalayci et al., 2020; Nasiri & Khosravani, 2022; Rohani Raftar et al., 2024; Wang et al., 2017). Unsupervised learning, such as K-means (MacQueen, 1967) and DBSCAN (Ester et al., 1996), enables data-driven grouping without labeled input, which can be useful for identifying weld features in FEM data. This study implements the Master S-N curve method and benchmarks it against other approaches. The DBSCAN algorithm is used to detect weld toe lines automatically. The resulting method improves fatigue assessment efficiency by identifying critical welds directly in the FEM post-processing phase. 2. Methodology 2.1. Effective notch stress method The Effective Notch Stress (ENS) approach has been widely implemented for the fatigue strength evaluation of welded structures as it captures local stress at notches, weld toes and roots. The underlying principle of this method derives from Neuber’s concept (Radaj et al., 2013), where stress is averaged along the crack path by integrating theoretical crack tip stresses on a defined length. In this method, the weld notches, which are prone to fatigue crack initiation, are modelled using a fictitious notch radius, as shown in Fig. 1. This radius is variable and changes according to the thickness of the load-bearing plates (Hobbacher & Baumgartner, 2024). The stress distribution is then analysed within the notch ligament, where the maximum local stress, known as the effective notch stress, is identified. This maximum notch stress reflects the local effects of geometry and loading, and it can be directly used in fatigue assessment.

Figure 1.Effective notch stress method (a) Fictitious rounding of weld toe and roots (b) stress distribution (c) Recommended meshing at weld toe (Ghanadi et al., 2024)

2.2. Master S-N curve method The Master S-N curve method was selected for its ability to handle coarse shell meshes, which are common in large FEM models where fine meshing is impractical. It allows for direct calculation of structural stress from nodal