THE MAIN GOAL of in-line inspection (ILI) is (according to the seven basic ILI quality metrics ) to correctly detect, locate, and identify all types of defects present in a pipeline and to size them in a fashion which allows statistical assessment of their true sizes. If achieved, the last fact opens the door widely to meaningful usage of the most sophisticated methods of structural mechanics (which were developed spending worldwide billions of $$$ but not used yet to its fullest in the pipeline industry) and obtaining most-accurate values possible of pipeline residual strength, probability of failure, and residual lifetime. This, in its turn, permits using predictive-maintenance technology in pipeline operation, introducing optimal inspections and repair logistics, and maximizing the long-term utility of the asset (in our case, the pipeline system).
The paper describes the methodology developed by the authors of a holistic innovative approach to ILI data generation and data management, which dramatically increases ILI inspection capabilities. This methodology, to a large extent, decreases the existing uncertainties and minimizes scatter of the input parameters and, thereby, makes predictions based on ILI data less conservative. As a result, this permits creation of safe solutions and avoidance of dangerous errors in predictions which include assessment of pipeline inspection frequency and safety margins.
According to the API 1163 Standard , the ILI measurement results are characterized by three parameters of statistical nature: tolerance, certainty, and the confidence level. Tolerance is the range with which an anomaly dimension or characteristic is sized or characterized, and certainty is the probability that a reported anomaly characteristic is within a stated tolerance. Confidence level is a statistical term used to describe the mathematical certainty with which a statement is made, and indicates the confidence with which the tolerance and certainty levels are satisfied. The paper discusses the statistical sources of these probabilities and how they should be interpreted and handled. The paper contains recommendations on how to approach different practical problems, and illustrates each case with real-life examples.
How to Cite
pipelines, defects, in-line inspection, measurement errors, statistical analysis.