13th International Conference on Fracture June 16–21, 2013, Beijing, China -7- a) dmin = 3 mm b) dmin = 8 mm c) dmin = 22 mm Figure 4. AlMgSi1 fatigue data and percentile curves for 1, 50 and 99 % failure probability 5. Discussion As can be observed in Fig.4, the percentile curves for all three specimen geometries provided by the model describe the fatigue data well, both in terms of median curve and data scatter. As the estimated parameters for each data set are referred to the same surface area ΔS, they should coincide for all three data sets. Nevertheless, as can be seen in table 3, this is only the case for the threshold parameter C for the d8 and d22 specimens. According to the model, a comparison of the Weibull parameters λ, β and δ requires the parameters B and C to be coincident to compute the normalized variable V. The extrapolation from the d8 estimates to the d3 Wöhler field overestimates both the median curve of fatigue life for constant stress levels and the data scatter. A possible reason could be that for such small specimens statistical independence, as is an assumption of the model, is not fulfilled in this case. As reference for other practical cases, in [6] it was also observed that the fatigue behaviour of the shortest prestressing wires could not be described based on the estimates for the longer wires. Therefore, the statistical dependence [9] based on considerations related to the defects from which fatigue initiation arises together with experimental work should be further investigated in order to 10 4 10 5 10 6 10 7 10 8 180 200 220 240 260 280 300 320 340 N [Cycles] Δ σ [MPa] 99 % 50 % 1 % 10 4 10 5 10 6 10 7 10 8 180 200 220 240 260 280 300 320 340 N [Cycles] Δ σ [MPa] 1% 50% 99% 10 4 10 5 10 6 10 7 10 8 180 200 220 240 260 280 300 320 340 N [Cycles] Δ σ [MPa] 1 % 50 % 99 %
RkJQdWJsaXNoZXIy MjM0NDE=