Session: 07-05: Cracks Uncertainty Modelling

Submission Number: 185746

The Probability of Unreported Crack-Like Features in Pipe Segments With Degraded Ili Performance 

In-line inspections (ILI) are foundational to pipeline integrity and reliability management. While operators rely on ILI data to estimate the likelihood of failure from identified features, recent incidents have highlighted a significant analysis challenge: the presence of unreported or missed stress corrosion cracking (SCC) anomalies. Traditional risk assessment methods often account for unreported features by scaling reported feature counts by a size-dependent Probability of Detection (POD) derived from validation data or ILI specifications. However, a significant gap in current methodology exists when an ILI tool reports no features at all. In such "nothing-found" scenarios, standard scaling methods fail, potentially leading to an underestimation of risk.

This paper presents a generative Bayesian model designed to quantify the frequency and size distribution of SCC features that may exist on a pipeline despite an ILI inspection that reported zero features. The model treats the inspection as a probabilistic sampling problem without replacement, utilizing a Hypergeometric distribution to evaluate the likelihood of different combinations of feature frequency and depth sizing, and incorporating varying tool performance to address degraded POD resulting from tool speed excursions. By integrating ILI performance specifications with an Exponential size distribution and a Beta-Binomial frequency, the model estimates the most likely population and size distribution of unreported anomalies missed during the inspection.

Furthermore, the paper showcases a practical application of this methodology inspired by a real-world project and dataset. The case study examines a pipeline segment where a portion of the ILI run suffered from degraded data quality. We demonstrate how the model estimates the "missed" feature population in these degraded zones and how these results are directly incorporated into a reliability-based risk assessment. This approach provides operators with a quantitative basis to determine if the associated uncertainty in missed anomalies requires accelerated actions such as modified ILI re-inspection intervals or other mitigations.

 

Presenting Author: Ryan Stewart, Integral Engineering

Presenting Author Biography: Ryan has 6 years of oil & gas experience, focusing on measurement, statistics, and machine learning. He participated in PRCI research, like the IM-1-06 project developing the ILI validation spreadsheet. He has also authored several papers at IPC on coating performance, material verification and facility integrity.

Authors:

Ryan Stewart Integral Engineering
Jason Skow Integral Engineering