2024 – 55

Daryl Bandstra1, Alex Fraser1, Miaad Safari2, Kai Ji2

1Integral Engineering, Edmonton, Canada., 2Enbridge Gas, Toronto, Canada

Abstract

The transmission pipeline industry is increasingly utilizing probabilistic models for assessing the probability of failure of anomalies measured during in-line inspections such as corrosion and cracks.  PHMSA (Pipeline and Hazardous Materials Safety Administration) has recently classified probabilistic models as best practice, suitable for supporting all types of decision-making, as these models effectively represent the uncertainty in input data using probabilistic distributions.  When modeling corrosion using a probabilistic model, the in-line inspection data includes measurements of individual corrosion anomalies.  When these anomalies are in close proximity, there is potential for interactions that reduce the overall burst capacity, and the effect of these interactions can be considered by modelling these individual defects as a cluster.

 Some probabilistic analyses idealize the clusters as single, large anomalies while others address the added complexity of considering all possible combinations of the individual anomalies that comprise the cluster.  This presentation will investigate the differences in the estimated probability of failure of these two approaches by evaluating a range of test cases and real-world clusters. These cases will illustrate scenarios where significant discrepancies exist and areas where both approaches yield comparable results. Additionally, we compare the effects of different approaches for modeling the correlation of measurement errors associated with anomalies within a cluster.

Read/Download PaperGrade