PSI - Issue 12

Venanzio Giannella et al. / Procedia Structural Integrity 12 (2018) 499–506 V. Giannella/ Structural Integrity Procedia 00 (2018) 000 – 000

503

5

Figure 3: DBEM model used for the numerical analyses with highlight of the crack insertion area.

4. Results

In the first experimental test, the static load was set at 24 kN and the maximum value of the dynamic load was set at 8 kN (stress ratio R = -1). Experimental and numerical crack propagation paths are compared in Fig. 4. A proper setting of the calibration parameter w made numerical and experimental Crack-Growth Rates (CGRs) to overlap; the experimental and FEM results in Fig. 4 are taken from literature (Dhondt et al., 2018). For such numerical analysis, neither initial notch shape (approximated by a sharp crack) nor contacts on the crack faces were considered; such simplifications allowed to decrease the runtimes with a negligible impact on the results since the combination of static and dynamic loads led to completely opened crack faces during the entire fatigue cycle. Various simulations were also performed at different levels of static loadings while the dynamic load amplitude was kept as fixed at 8 kN with a load ratio R=-1. In particular, the main goal was to assess the ratio of static load magnitude vs. dynamic load amplitude causing the switch of crack propagation angle, from a direction perpendicular to the dynamic load to a direction perpendicular to the static load. SIFs calculated for the initial crack are reported in Fig. 5 for various combinations of the horizontal static load and the vertical dynamic load. From the “KI” chart (Fig. 5a), it can be seen how modelling contact between the crack faces prevents their intersection, therefore providing nearly null KI values for LC2 (static – dynamic load case). Moreover, the pressure exerted between crack faces for load case LC2 leads also to a reduction of the KII values: this can be attributed to both pressure and friction between the two crack faces that limits the sliding. As instance, such reduction can be seen in the two “0 ± 8 kN” curves of KII chart (Fig. 5b): they exhibit a different absolute value, with a difference that would disappear in case of a null friction coefficient. In Fig. 6, the crack paths as well as CGRs calculated with different static loads ranging between 0 kN and 4kN are reported. It can be seen that: • there is a decrease in fat igue life when switching from the load case “0 + 8 kN” (null static load with added fatigue load at R=0) to “0 ± 8 kN” (null static load with added fatigue load at R=-1); moreover a change in the crack path can be envisaged; • the fatigue life increases with the static load up to a static load lower than 2 kN; • on the contrary, for static load greater than 2 kN (e.g. “4 ± 8 kN”), the tendency was opposite . A possible explanation of such behavior come from the fact that the crack started to propagate along the dynamic load direction for static loads greater than 2 kN and the fatigue life increased as a consequence. However, keeping on increasing the static load, the fatigue life decreased since the path did not change anymore.