Systematic statistical characterisation of stress-life datasets using 3-Parameter distributions

•Demonstration of a statistical characterisation process for stress-life datasets.•Combination of probability plotting and maximum likelihood estimates.•Construction of design S-N curves and safe-life design case study.•Challenge of commonly-assumed 2-Parameter Log-Normal distribution for S-N curves.•3-Parameter Weibull distribution can increase estimated safe-life by 20%. The variability present in S-N datasets is typically characterised using probability distributions to enable the construction of Probability-S-N curves for design. 3-Parameter Log-Normal and Weibull distributions have been proposed as alternative distributions to the commonly assumed 2-Parameter Log-Normal distribution. This paper performs statistical characterisation of a 4340 steel S-N dataset from the Engineering Sciences Data Unit using a systematic methodology. The 3-Parameter Weibull distribution provided improved characterisation of the S-N dataset. Using a case study, it was also demonstrated that use of a 3-Parameter Weibull distribution can increase component safe-life values by 20% when compared to the 2-Parameter Log-Normal distribution.