PSI - Issue 54

428 6

Jenny Köckritz et al. / Procedia Structural Integrity 54 (2024) 423–430 J. Köckritz / Structural Integrity Procedia 00 (2019) 000 – 000

Membrane material data

Bending material data

Fatigue coefficient variation

Fatigue exponent variation Fig. 4. Fatigue coefficient and fatigue exponent variation and influence on assessed fatigue life with the SWF method of bending material data and membrane material data

3. Results and discussion The welds of the cargo bicycle are multiaxially loaded with different amounts of torsion and bending, as well as in-phase and out-of-phase. Here, only in-phase multiaxial loading was considered, but for two different ratios of bending and torsion to evaluate the assessment methods for deviating conditions. The comparisons of all three numerical assessment methods with the experimental results of the simplified specimen subjected to constant amplitude loading are displayed in Fig. 5. The experimental results of load case 1 in Fig. 5 (a) are scattered with a logarithmic standard deviation of =0.24 . The HSS method leads to very conservative results for load case 1, while the EN method is less conservative, but still below the scatter range of 90% survival certainty just as the SWF method with SWF Mat2. The SWF method with SWF Mat1 displays the best correlation with experimental data in this load case. Especially apparent is the good correlation of numerical and experimental fatigue exponent. In contrast, the experimental results of load case 2 with increased torsion in Fig. 5 (b) show a smaller scatter with =0.12 and do not correlate as well with the SWF method for either material data set. For load case 2, SWF and HSS lead to very conservative assessments. The HSS method for shear loaded welds and the EN method show a good correlation to the experimental results, but the low fatigue coefficients of the corresponding FAT classes cause a declining correlation with decreasing load. The results of load case 2 demonstrate that the SWF method with its reference bending and membrane S-N curve overestimates the damaging effect of shear stresses. However, if bending or normal stresses are prevalent and material data for the welded alloy are available, relatively reliable results can be reached with the SWF method. The experimental results under the load collective and the corresponding weld assessment with the SWF method are displayed in Fig. 6 (a). The experimental results show a much higher scatter of s log = 0,34, as can be expected for tests under a load collective. Surprisingly, the weld assessment with SWT showed a better correlation for this load scenario than with constant amplitude loads. The SWF Mat1 even slightly overestimated the fatigue life. The SWF Mat2 leads to slightly conservative, but good correlation.