PSI - Issue 57

Alberto Visentin et al. / Procedia Structural Integrity 57 (2024) 524–531 Alberto Visentin et al./ Structural Integrity Procedia 00 (2023) 000 – 000

529

6

Figure 3 . Case study: Witt et al. (Witt et al. 2000; Yousefi et al. 2001), tube-to-flange welded joint. (a) Joint geometry (dimensions are in millimeters). (b) FE model and detail of the generated 10-nodetetrahedral mesh at the weld toe in Ansys® Mechanicalaccording to the PSM. (c) Normalized stress range spectrum applied for VA fatigue tests (Campagnolo et al. 2022). The joint was analysed according to the PSM in (Campagnolo et al. 2022), taking advantage of a 2D FE model generated adopting a free mesh pattern of 4-node harmonic elements (PLANE 25) in Ansys® Mechanical APDL. Such previous results (Campagnolo et al. 2022) have been compared in the present investigation with new results obtained from automated PSM tool in Ansys® Mechanical(Fig. 2). The geometry has been simplified to a quarterof the joint by taking advantage of the axis-symmetry and 3D 10-node tetrahedral elements (SOLID 187) have been adopted to generate the FE mesh in Ansys® Mechanical (Fig. 3b). A minimum mesh density ratio a/d ≥ 3 must be adopted at the weld toe when subjected to both mode I and mode III local stresses in order to apply the PSM (Meneghetti and Campagnolo 2020). The characteristic size a equals the tube thickness at the weld toe, namely a = t = 8 mm.Therefore, a global element size equalto d = 8 /3 = 2.67 mm has been adopted to generate the FE mesh over the entire model (Fig. 3b). For each analysed node at the weld toe, the PSM tool automatically extracted mode I Δσ θθ,θ=0,peak and mode III Δτ θz,θ=0,peak peak stresses (mode II peak stress Δτ rθ,θ=0,peak being not singular at the weld toe), evaluated mode I and mode III f w and f s parameters featured in Eq. (2) and calculated the equivalent peak stress according to Eq. (1). Eventually, the PSM tool evaluated the local biaxiality ratio λ according to Eq. (3), which resulted null for the pure bending loading case, while it was greater than zero for both pure torsion and combined bending and torsion loading cases. Due to the moving averages performed on peak stresses (Meneghetti and Campagnolo 2020), ‘Q’ is the first available node for which the equivalent peak stress can be defined (Fig. 3b). Indeed, when considering only a quarter of the joint, node ‘P ’ (Fig. 3b) belongs to a free surface of the model and is therefore neglected during the PSM analysis (see (Meneghetti and Campagnolo 2020)). However, in the case of bending loading, the mode I peak stress σ θθ,θ=0,peak calculated with a quarter of the joint on node ‘ Q ’ differs by less than 2% from that calculated with half of the joint on node ‘ P ’ , given the same global element size. In the case of torsion loading, the mode III peak stress τ θz,θ=0,peak is constant for each node along the weld toe line. Eventually, the equivalent peak stress calculated on node