Continuous management of Stress Corrosion Cracking (SCC) on buried pipelines is an important element of an Integrity Management Program. Crack detection inspection tools have been widely used in the pipeline industry for locating SCC. There are some potential limitations with these tools; therefore estimating the probability of detection (POD), probability of identification (POI) and sizing accuracy is essential to validate the inspection results. The relationship between field-identified and ILI-reported features is a key component.

This paper probabilistically assesses detection, identification, and sizing accuracy of a crack detection ILI tool. The paper also describes a procedure for estimating the true defect population from the number and severity of SCC features reported by an ILI using statistical analyses. As part of the analyses, transition matrices were developed.

The ILI data set was generated from 26 pipelines with a known history of SCC defects. Excavation data was collected from field reports where over 300 crack-like features were identified by Nondestructive Examination (NDE). The results obtained from the field were compared against the ILI reported features and ILI detection limits.

Statistical models using three-parameter (shape, scale, and threshold) Weibull distributions were found to best model the depth and length distributions. The statistical analyses were based on results from True Positive features (i.e., reported by ILI tool and found in the field), False Negative features (i.e. not reported by the ILI tool and found in the field to be above the detection threshold of the ILI tool), and Unreported features (not reported by the ILI tool and found in the field to be below the detection threshold of ILI tool).

To determine the impact of an excavation program after an ILI run, an approach was developed to analytically grow the defects over time. Probabilistic models were then used to predict ILI feature lists and simulate prior and future ILI results from the defect population. This task was performed by creating an inverse transition probability matrix. The methodology is the basis for the development of line-specific SCC Management Programs.

This content is only available via PDF.
You do not currently have access to this content.