Issue 78

G. Palomba et alii, Fracture and Structural Integrity, 78 (2026) 233-247; DOI: 10.3221/IGF-ESIS.78.14

Figure 10: True stress–strain curves for BM and WM (HAZ similar to BM) using Kamaya method. Non-linear Finite Element Modelling and comparison with DIC results Finite element simulations were conducted via ANSYS ® Mecahnical APDL to model the mechanical behavior of the welded tensile specimen under uniaxial loading. The specimen geometry replicated the experimental configuration, with the weld region represented by a central band defined by variable hardness-based properties. Three distinct material zones, BM, HAZ, and WM, were implemented using separate materials, each defined by its respective true stress–strain curve. The FE material-zone boundaries were defined using the “as-designed” geometry of the welding as primary reference and the macro etched cross-section as “as produced” amendment, using the visible fusion boundary. The hardness profile was used as a supporting criterion to confirm the transition between BM, HAZ, and WM. These boundaries were then mapped onto the tensile-specimen geometry and implemented as separate material regions in the FE model A structured mesh of 3D 20 nodes quadratic structural solid elements (solid186) was employed, with refinement near the weld line to accurately capture strain gradients, as shown in Fig. 11. The global element size was set to 1.0 mm. To accurately capture stress and strain concentrations near the weld zone, a geometric bias was applied along the longitudinal axis, resulting in a refined mesh localized near the weld root. Due to the geometric constraints at the minimum weld thickness, the absolute minimum element length reached 0.007 mm. The finalized mesh consists of 37170 elements and 166350 nodes. Boundary conditions replicated the tensile test: one grip end was fully constrained in all degrees of freedom (Ux = Uy = Uz = 0). Tensile loading was applied at the opposite end by containing the nodes of the grip surface to a displacement control condition, matching the experimental test crosshead velocity. To model the nonlinear material behavior, a multilinear kinematic hardening model (KINH) was implemented. A mesh convergence study was performed to balance accuracy and computational efficiency, based on a solution variation of less than 1.0% within the welded region. Nonlinear geometric effects were activated. Fig. 11 reports the scheme of the applied boundary conditions, a detail of used mesh and the true stress-strain curves used.

(a) (b) Figure 11: Finite element mesh and boundary conditions (a); true stress-strain curves used in the FE model (b).

The local strain distribution from the FE model also matched DIC measurements, with strain localization observed at the HAZ–WM interface. This localization corresponds to the region of highest hardness gradient, confirming the sensitivity of strain accumulation to local material properties. In terms of strain distribution, as visible in Fig. 12, the WM exhibited higher stiffness and lower plastic deformation compared to the BM. The strain concentration observed at the boundary between these regions may represent a potential site for fatigue crack initiation under cyclic loading. Accordingly, the present results suggest that hardness mapping could provide useful information for interpreting local mechanical behavior and for formulating hypotheses regarding failure initiation, although further investigation is required to establish such relationships quantitatively. To further validate the numerical predictions, a comparison with Digital Image Correlation (DIC) measurements was conducted and the results are represented in Fig. 13. Fig. 13 presents a quantitative analysis of the longitudinal strain ( ε yy ) distribution across the welded AISI 316L tensile specimen, as extracted from the DIC measurements. The plot shows the strain profile along length of the specimen, intersecting the weld centerline, obtained from DIC measurements and from FEM analysis. Both curves clearly reveals the presence of the WM region, centred between ~30 mm and ~45 mm, which shows a sharp dip in strain intensity, indicating

244