PSI - Issue 39

R. Yarullin et al. / Procedia Structural Integrity 39 (2022) 364–378 Author name / Structural Integrity Procedia 00 (2021) 000–000

375

12

conditions. Good agreement between these results indicates the correctness of loading conditions and applied crack growth criteria.

Fig. 14. Numerical and experimental CMOD comparison under (a) tension and (b) tension/torsion loading conditions.

4.3. Fatigue life prediction The numerical simulation of 3D fatigue crack growth under complex stress state represents a crucial factor in fracture mechanics. The synergic use of the modern software ABAQUS and FRANC3D allowed to predict the residual fatigue life of components on the base of fully automatic procedure, without the need for relevant user intervention. In this part of our study the numerical crack length vs cycles curves are presented for tension, tension/torsion and torsion loading conditions of a hollow cylindrical specimens with initial semi-elliptical notch. The first experimental crack detected was taken as reference for the numerical cycles counting. Therefore, the numerical cycles have been offset in such a way to get the same initial point, in a diagram size vs cycles, between numerical and experimental results. In the Fig. 15, the numerical and experimental fatigue lives are compared, according to their loading conditions. The tension tests and the related numerical analyses were carried with a tensile load of 35 for D16T, and 80 for the B95AT Al-alloy. A very good agreement between numerical and experimental fatigue life is evident in Fig. 15a for D16T Al-alloy. On the contrary, a higher experimental crack growth rate / with respect to numerical simulations is observed for B95AT Al-alloy (Fig. 15b): a possible reason for such a mismatch is that the Linear Elastic Fracture Mechanics (LEFM) hypotheses might not be strictly valid with such high load magnitudes. The tension/torsion tests and numerical analyses have been carried with a tensile load of 40 and a torsional torque of 250 , for both materials. An almost perfect correlation between numerical and experimental data is observed for both alloys during all propagation stages (Fig. 15c and Fig. 15d). The numerical/experimental crack length vs. cycles for D16T and for B95AT Al-alloy specimens, undergoing a pure torsional load of 450 and 250 respectively, is shown in Figs. 15e and 15f. In the early propagation stages, the numerical analyses perfectly retrace the experimental data for both alloys. The final numerical crack propagation steps, starting from nearly 120000 , show a growth rate higher than the experimental one. In general, by considering all load conditions, it can be observed that the misalignment between numerical and experimental data increases when the crack front intersects the specimen’s central hole and splits in two shorter fronts. This phenomenon could be explained by the fact that a larger plasticity zone may be developed and the free surface effect interests a wider portion of the crack front in that circumstance, leading to exceed the LEFM validity limits. Any way this discrepancy is relegated to the final crack growth stage.