One element of probabilistic pipe fracture analyses is an assessment of the stability of cracks under applied load. If the load at some time step is greater than the cracked section can sustain, the state of the pipe is reported as ruptured. Another element of probabilistic pipe fracture analyses is the calculation of the crack opening displacement (COD) value which is used as an input to a leak rate calculation. Both of these analyses typically require pipe material properties that are sampled from a probabilistic distribution characterized by a mean value and some variance about the mean. Such properties include yield strength, ultimate strength, Ramberg-Osgood stress-strain curve coefficients, and fracture toughness curve parameters such as initiation toughness and tearing parameters. From a procedural perspective, it is tempting to sample all of the relevant parameters independently. However, in some instances, the assumption of independence leads to nonsensical conditions such as yield being greater than ultimate; yield and ultimate being completely disconnected from the underlying Ramberg-Osgood description of the stress-strain curve; J-R toughness curves that are convex instead of concave; or initiation J values as low as zero. This paper provides a suggested approach to sampling material properties for probabilistic pipe fracture analyses. The approach maintains consistency between yield, ultimate and the underlying Ramberg-Osgood stress-strain curve, and sets reasonable cutoff limits on the properties that are sampled. Equations for ensuring consistency among the related parameters are given along with justifications for the sampling limits, and an example of application of the proposed methodology.
