| Abstract: | A concept to describe the fatigue lifetime by using statistical methods In order to spend time and money, cyclic loading tests have been usually undertaken with a relative small number of specimens. On the other hand, it is well established that fatigue life data, i.e. the cycle number of crack initiation Ni as well as the cycle number of total failure Nf show a scatter for a given load level. Therefore, fatigue live predictions must be undertaken using statistical methods. In this case it has to be taken into account that the quantity Nf consists of two parts, i.e. crack initiation and propagation. The first one has to be calculated in a statistical manner, whereas the second one with deterministic methods. Therefore, a superposition of statistical and deterministic parts has to be taken into account. Up to now it was estimated that in all cases the failure results from the same damage mechanism. In this case a scatter of fatigue life may be described by the Weibull distribution and characterised by the location parameter and the Weibull exponent. By using the correct density function, all fatigue life data Ni can be plotted in a failure probability Pv(Ni) vs. Ni plot. In case of some Al-alloys no linear dependence between Pv and Ni has been found, indicating that more than one type of defect spectrum may initiate the failure. Indeed, in AlMgSi cold worked specimens two types of cracks, e.g. surface as well as corner cracks, have been found that initiate the failure of the specimens. If was possible to correlate each type of crack with an individual Weibull distribution. By using this procedure, fatigue life can be estimated better than before. |
