Abstract
In-Line Inspection (ILI) tool performance is a key input parameter to the pipeline integrity reliability models. The ILI performance is measured through metrics, established in the pipeline industry, that quantify the ILI tool’s detection, sizing, and identification capabilities. The minimum expected performance values are specified by the ILI vendor based on the vendor’s internal testing. However, the actual observed ILI tool performance by individual pipeline operators may vary, and thus, it is important to include the actual tool performance in the reliability model to properly estimate the pipeline Probability of Failure (PoF).
In addition to estimating the PoF for the list of ILI detected features across the pipeline, the pipeline operator needs to estimate the PoF for features that might have not been detected by the ILI tool due to ILI performance limitations. These undetected features may pose integrity threat and should be accounted for in reliability modelling for a comprehensive risk assessment. This paper provides a practicable implementation of a methodology developed in a Pipeline Research Council International (PRCI) project for characterizing the population of undetected ILI corrosion features. The methodology estimates the density of undetected features based on a function of ILI’s probability of detection (POD), probability of identification (POI), and probability of false call (POFC) while utilizing the ILI sizing accuracy to estimate the severity of undetected features.
In the present study, the PRCI undetected feature methodology was extended to include crack ILI features and was adapted to provide an accurate quantification of ILI sizing error from linear regression using matched ILI and field-measurement data. In addition, the paper demonstrates, (1) the impact of using feature morphology specific ILI performance values compared to generic performance values, and (2) effect of feature depth and length correlation in undetected feature severity modelling. The practicable implementation will be presented through two case studies that will simulate the population of undetected features across two pipelines and estimate their PoF values along with the PoF values of detected features population.
