An Analysis of Flow Strength Models for Failure Pressure Prediction of Metal Loss Features in Oil and Gas Pipelines

Abstract Failure pressure prediction of metal loss features is a vital part of any pipeline integrity management program. Assessments are typically performed using the ASME B31G based models. These models have evolved over the years, from the original to modified B31G models, to effective area / RSTRENG models, to the most recent PSQR model. Some of these methodologies are also included in API 579, BS 7910 and DNV corrosion flaw assessments. Recent work by PRCI has demonstrated these models, although increasingly refined in their approach, maintain a conservative bias in their failure pressure predictions. This decreases efficiency of the owner’s integrity programs. Some analysts have attributed the conservatism to limitations in the flow strength models used in the assessments. This work uses finite element methods to analyze the stress and strain distributions at the bases of long metal loss flaws to better understand plasticity behavior during plastic collapse. A key parameter of interest is the Ramberg-Osgood hardening exponent, which characterizes flow of material during the failure process. Materials with higher or lower hardening exponents show differences in their stress and strain distributions. In effect, there are multiple “shades of grey” between purely elastic and purely plastic behavior. This has a follow-on effect on how flow strength should be defined for different flaw geometries. In some cases, the current flow strengths models are suitable. In other cases, the definition needs refinement. Recognition of these effects can contribute to optimizing current models. The study also looks at currently available failure pressure data, from both in-service and laboratory tests. The trends observed in the flow strength modelling are considered with respect to the true failure strength of metal loss flaws. The results of the study provide insights into limitations of the current flow strength models and suggest possible improvements for improved failure pressure predictions going forward.