Pipelines are huge infrastructural facilities that are set up to transport hydrocarbon products over long and short distances from the point of production to the end user. Most pipelines are made of carbon steel, therefore, making corrosion a critical integrity threat of these systems. Corrosion is a process that is time dependent, which reduces the pipeline’s wall thickness gradually. This may result to rapture failures and leakages that have considerable consequences for the environment, society, and economy.
The integrity of corroded pipelines should be managed throughout the life of the pipeline in a bid to avoid failures and service interruptions. Therefore, in-line inspections have been adopted as non-destructive methods to monitor the structural health, detect and size pipelines’ corrosion features. In-line inspections are normally done within a few years. In order to establish feature-specific corrosion growth as well as time to failure, features from a minimum of two in-line inspections must be matched in relation to their location on the pipeline.
Generally, matching can be realized by using raw signal data from the inspection tools or using individual corrosion features already identified and sized as the result of the raw signal processing. The former has an advantage of relying on raw signals that contain more information as opposed to reported locations of the corrosion features. This therefore increases the chance of getting more precise matching outcomes.
Feature matching implementing the sized features is normally done manually. Normally, this time intensive process involves comparing feature locations from one inspection to another and identifies manually the features belonging together. Therefore, this process is susceptible to errors.
Markus Dann at University of Calgary in collaboration with Christoph Dann at Carnegie Mellon University introduced a framework for automated matching of corrosion features. This was in a bid to realize a time efficient and a more reliable matching replacing the manual process. Their research work is published in journal, Reliability Engineering and System Safety.
The proposed framework consisted of five steps. The first three steps entailed the overall problem of feature matching from a number of in-line inspections being transformed into several instances of a 2-dimensional matching problem. The inputs of the framework were the locations of all corrosion features that were detected along with girth welds from every in-line inspection. Implementing a multi-step approach, the size of a number of in-line inspections with a large number of features could be reduced to an assembly of independent smaller problems to effectively match the corrosion features.
Reducing the multi-in-line inspection 3D matching problem to an assembly of 2-D matching problems made solution computation fast for complex matching problems as well. The outcomes of the matching process could be visualized implementing the transformation obtained from the matching evaluation for improved verification of the final solution. Matching uncertainty that is inherent in a matching process was quantified in the correspondence matrix. The uncertainty could be included directly in the corrosion growth evaluation for a consistent propagation
Detection and false call uncertainty still applied to the results obtained. This meant that a separate analysis was necessary to establish if the detected outliers were false calls or the missing sets had not been detected. It was likely that perfect matches were identified correctly, but a few pairs could have been false.
Markus R. Dann and Christoph Dann. Automated matching of pipeline corrosion features from in-line inspection data. Reliability Engineering and System Safety, volume 162 (2017), pages 40–50.
Go To Reliability Engineering and System Safety