PSI - Issue 4

Author name / Structural Integrity Procedia 00 (2017) 000–000

7

J. Maierhofer et al. / Procedia Structural Integrity 4 (2017) 19–26

25

3.3. Load amplitudes below the fatigue crack growth threshold

For common fatigue experiments small loads are rather not interesting because they are expected to be only a waste of testing time. Of course, loads smaller than the threshold of crack propagation do not lead to crack propagation; on the contrary, they can lead to crack retardation. This retardation effect is neglected if small loads are omitted in block program testing. The reason for crack retardation in this case, which has so far been rarely investigated, is that small loads near the threshold lead to a build-up of oxide layers on the fracture surface and to wear debris formation behind the crack tip. To investigate the influence of such small loads, again experiments with SENB specimen were done. Similar to the overload experiments after pre-cracking a constant load amplitude was applied until a crack extension of ~4 mm was reached. Then the load amplitude was decreased to a stress intensity factor ∆ K smaller than the threshold of crack propagation ∆ K th . A certain amount of load cycles with small loads was used to build up a measurably thick oxide layer before again the primary load amplitude was applied. In Fig. 6 the influence of these small loads is shown and compared to a reference specimen were no small loads were applied. Similar to the overload effect also small loads can lead to a significant drop in the crack propagation rate.

Fig. 6. Crack retardation due to build-up of oxide layer.

Due to time-saving reasons, small loads are very often omitted in full-scale tests, and when the test is finished the number of omitted load cycles is added to the result. But, as shown in Fig. 6, the small loads influence also the crack propagation rate and thereby increase the residual lifetime. We have tried to approximate the zone of influence of small loads Z ox depending on the number of load cycles N ox and the corresponding amplitude ∆ K ox of the small load: ( ) ox ox ox ox ox ox ox 1 q p s Z L R K N ⋅ ⋅ + ⋅ ∆ = (6) To consider the influence of small loads – i.e., the oxide induced crack retardation – in computational models, we suggest to increase the crack growth threshold by ( ) ox ox ox ox ox ox th,ox 1 n r m R K N K K ⋅ ⋅ + ⋅ ∆ = ∆ . (7)

Again, the model parameters K ox , L ox , m ox , n ox , p ox , q ox , r ox , s ox are determined by statistical regression. The simple