The use of probabilistic fracture mechanics to assess the reliability of engineering structures requires knowledge about the statistical scatter of the parameters used in the models.

A review is given of the statistically defined tanh reference toughness curves. This method is then used to aid an investigation into the effect of tighter chemical specifications. The predicted curves and small quantity of available published data do support the hypothesis of improved brittle/ductile transition temperatures with controlled chemistry. An experimental program has been undertaken to obtain plain plate data that are then used to derive an empirical correlation between K_Ic and Charpy energy, assuming the distributions of crack tip opening displacement and C_v can be related for lower shelf temperatures. Tolerance limits and sample size estimates are also obtained for specified confidence levels.

Data on fatigue crack growth rates have been reviewed and the effects of parameters, such as stress ratio, temperature, and irradiation, studied qualitatively. A statistical analysis of published data has enabled a distribution for C in the Paris law to be derived assuming a linear ℓn C versus m relationship.

Finally, an illustrative example calculates the change in failure probability with the number of accumulated cycles.