Probabilistic Fracture Analysis of Pipelines

Many pipelines were built before the industry developed material specifications for fracture control. For these older pipelines an essential first step in fracture control is to estimate the existing likelihood of fracture initiation and propagation. It is also desirable for operators to know the size of defects the pipeline can tolerate without causing pipeline fracture. This paper describes a methodology developed for the probabilistic characterization of the fracture initiation and propagation susceptibility of older pipeline segments, made from line pipe exhibiting (by today’s standards) low to moderate strength and low notch toughness. It is applicable to ductile, brittle and mixed-mode fracture behaviour. A probabilistic analysis approach is ideally suited to the problem since it offers a way to quantitatively address both the inherent variability in the mechanical properties of line pipe and the uncertainties associated with the various models currently available to determine the conditions necessary to cause crack initiation or to force crack arrest. The method described addresses both of these forms of uncertainty, and also reflects the added uncertainty inherent in trying to estimate material properties for existing lines from small samples of data.