Evaluation of Corrosion Growth Prediction Methodologies Using Burst Pressure Comparisons From Repeated In-Line Inspections

At the forefront of the effort to understand and mitigate pipeline corrosion is the prediction of corrosion growth rates. It is important to understand the effect of corrosion growth estimates on integrity management decisions. An overly conservative approach results in unnecessary digs, while removing conservatism increases the potential for a missed feature to grow to a threatening size. While approaches to feature depth growth have been well-established, there has been less investigation into the growth of feature lengths. A literature review was performed on the methodologies applicable to length growth, and their performance was compared to those that only account for depth growth using a sample analysis. For pipelines with multiple in-line inspection (ILI) runs, feature or signal matching can be used to estimate the change in feature size. These rates can be used directly on individual features, averaged across pipe joints, or compiled into a statistical distribution. Alternatively, only one ILI measurement can be used and an assumption made on the age of the defect. These approaches are more commonly applied to depth growth but could be used to predict length growth as well. To compare the growth methodologies, the study used historical ILI measurements of a liquid pipeline to predict feature sizes and estimated burst pressures determined at the time of the latest ILI. The number of defects correctly predicted to have an insufficient burst pressure safety factor for safe operation was compared to the number of defects that were erroneously predicted to not meet this criterion, and those that were predicted to be safe but later found to not meet the safety factor requirement. The number of erroneously flagged defects was found to vary the most between methodologies. For the assessed data set, using the joint average rate based on feature box-matching was non-conservative on average. It was also found that incorporating length growth did not significantly affect the accuracy of the burst pressure predictions.