Burst Pressure Prediction for Axial Cracks in Pipelines With Non-Ideal Depth Profiles

Abstract Crack-like anomalies in pipelines are complex features that typically have an irregular profile. However, due to modelling complexity, the common practice is to approximate such profiles with a semi-elliptical crack whose length is equal to the total flaw length and whose depth corresponds to the maximum depth of the flaw. This approach can result in significant underestimates of burst pressure. The PRCI MAT-8 pipeline fracture model has become widely used in the industry since it was developed in a PRCI project in 2015 [2]. The original version of the MAT-8 model was derived from over 200 3D elastic-plastic finite simulations of pipe joints with axial cracks. The PRCI study considered three crack types: semi-elliptical surface cracks, constant depth (canoe-shaped) surface cracks, and through-wall cracks. In 2019, the Modified MAT-8 model was introduced to address non-ideal crack depth profiles [3]. An extensive finite element study was undertaken to model numerous non ideal longitudinal crack profiles in pipe joints. The output from these analyses led to a procedure to convert an arbitrary crack profile into an equivalent semi-elliptical crack. The depth of the equivalent crack is equal to the maximum flaw depth, but the length is typically much less than the total flaw length. Accounting for actual depth profiles, as opposed to the semi-elliptical assumption, affects computed burst pressures by as much as a factor of 2. The procedure described in this study improves upon an earlier phase of work [3] that developed an approach which was accurate for cracks with a single isolated deeper section that dominates the fracture demand. Since then, more real-world profiles have been assessed and several cases identified for which the previous procedure underestimated the fracture demand. The new procedure resolves this while ensuring a slight conservative bias and maintaining an accuracy that is comparable to typical inline inspection tool length sizing accuracy. The algorithm considers interactions between deeper sections of the crack profile to increase the length of the equivalent semi-elliptical crack. While the algorithm was developed and calibrated using machine learning techniques, it can be implemented without any specialized machine learning tools. To apply the algorithm in a probabilistic framework, the in-line inspection (ILI) measurement uncertainty must be characterized at all points along the crack profile. ILI vendor specifications on measurement uncertainty however only cover the uncertainty in reported maximum depth and total length. In this study a method to automatically align over 17,500 ILI profiles collected from repeated measurements of over 8,800 crack-like features was developed so that the repeatability of the reported maximum depth and total length could be quantified and the point-to-point error along the profiles characterized.