Issue 48

R. Baptista et alii, Frattura ed Integrità Strutturale, 48 (2019) 257-268; DOI: 10.3221/IGF-ESIS.48.27 265 different modeled geometry is also responsible for the different predicted fatigue life, Fig. 8 b), as the SIF for mode I are lower on the 2D model, resulting in a lower crack driving force. a) b) Figure 8 : a) Fatigue crack growth paths, b) Crack propagation curves on the HAZ of a specimen with a longitudinal crack, using different FEM simulation models. a) b) Figure 9 : Stress intensity factors using different FEM simulation models, a) mode I, b) mode II. Fig. 9 show the evolution of K I and K II as the crack is propagating. In Fig. 9 a) one can see the evolution of K I , as expected the SIF increases as the crack length increases. While in Fig. 9 b) one can see the evolution of K II , for small cracks K II is almost zero, but as the crack length increases the absolute value of K II will increase, as the crack is propagating in mixed mode. As the crack propagation direction is calculated from the previous increment values of K I and K II according to Eqn. (2), the crack deflecting is in this case influenced by the evolution of K II . Fig. 9 b) show the existence of some numerical noise, as the value of K II can change from positive from negative, from one increment to the other. Shi et al. [13] have also reported this behavior. The 2D model is the one with less numerical noise, while the 3D model using the contour integral technique show small noise for smaller cracks, but a larger noise for larger cracks. As mentioned by Shi et al. [13], the numerical noise can be reduced by using a smaller crack length increment. In our case the increment used was 0.5 mm for all models. Shi et al. [13] reported that the best results were obtained with a fixed crack length increment of 0.3 mm. The difference, when using a 1 mm increment, is not significant. Fig. 10 show the evolution of the crack resulting deflection angle, as calculated by the MTS criterion and Eqn. (2). As mentioned initially the crack propagates in mode I, as the deflection angle is low. But as the crack grows and the value of K II increase, the deflection angle also increases, and the crack propagates in mixed mode. The increase is higher for the 2D model, justifying the obtained results. For bigger cracks, as the value of K I increases, the deflection angle decreases, and the crack will again behave as in pure mode I. As a final remark, as mentioned by Nasri et al. [21], if the crack deflection increases, the crack propagation will be delayed. Therefore the 6% increase in the fatigue life for the 2D model, is also justified by the increase crack deflection offer by this model. Crack deflection can be seen in Fig. 11. Here the 3D contour integral model is shown for the normal CT specimen (Fig. 11 a)) and for the longitudinal welded specimen (Fig. 11 b)). It is possible to see the crack propagating in the horizontal direction for the normal specimen, while the crack is deflected in the welded specimen. Fig. 12 show different snapshots