Consistently Inconsistent: Lessons From Operators Comparing Multiple IMU Data Sets

Abstract Operators and regulators have demonstrated an increased awareness of the geohazard threat over the last decade. In response to this threat, many pipeline operators have begun integrating bending strain assessments based on IMU data into their integrity management programs. Bending strain assessments, which provide a direct measure of the pipe condition, can be a valuable tool in any geohazard integrity management program. Unfortunately, many operators encounter challenges or even confusion when they receive results from a bending strain assessment. The results from a bending strain assessment can be difficult to understand and even more difficult to properly sentence, especially if the assessment identifies hundreds of features. These challenges can be compounded if the operator acquires a second IMU data set and elects to perform another bending strain assessment or a strain change assessment comparing the two data sets. In these situations, operators often observe significant variations in the number of reported bending strain features and the reported strain magnitudes associated with those features. These challenges can rarely be solved by comparing the final reports as the graphical information for the bending strain features is often rendered differently between vendors. This paper presents the results of two case studies where each case study had multiple bending strain reports based on different IMU data sets. Both case studies demonstrated common challenges experienced by operators. Each assessment had significant differences in the number of reported bending strain features and the locations of those features. In both case studies, the information supplied with the final reports was insufficient to reconcile the differences between the two assessments. After demonstrating these challenges, the paper presents the results of the strain change assessments when these assessments are performed according to the same criteria and the identified features are fully reconciled. Based on these results, the paper identifies and discusses the causes of the variation in both case studies which includes gauge length, signal noise, and analyst judgment. The paper demonstrates how those variables can impact the results of a bending strain assessment, and the lessons learned from this case study. The paper concludes by identifying the necessary reporting requirements for bending strain or strain change assessments and makes recommendations on how results from these assessments can be standardized to help operators minimize future variations.