A statistical investigation of fatigue behaviour according to Weibull's weakest link theory

The problem of deciding which statistical model is the "true one" in fatigue studies has occupied investigators for many years.Since the amount of data is generally not sufficient to discriminate conclusively between various models, a "defect-tolerant-approach" was used to postulate a suitable model.The two-parameter Weibull distribution was adopted since it can be deduced from a statistical distribution of crack initiating defects.Parameter estimation is carried out based on a large quantity of experimental high cycle fatigue data, which can be used in the context of reliability studies and quantification of the size effect.The Weibull shape parameter was estimated and found to be presented approximately by a Normal distribution.The Weibull scale parameter was suggested to be modelled as a deterministic value.