Abstract

Electromagnetic Acoustic Transducer (EMAT) is a non-destructive inspection technology that uses guided acoustic waves to detect planar flaws in a metal structure. When deployed via in-line inspection (ILI), it is an effective way to detect cracks in a pipeline. EMAT has thus become a staple of crack management programs throughout the world since its introduction to the market over a decade ago.

As with all technologies, challenges remain with the inspection process. One such challenge with EMAT is classification. While it is possible to determine that a defect is “crack-like” (a property determined by its tendency to reflect incident waves), it is difficult to determine the nature of the defect from the EMAT measurement alone. Indeed, similar reflections are obtained for many different types of defects, from relatively benign manufacturing and construction abnormalities, to more concerning anomalies such as stress corrosion cracking (SCC).

To compensate for the difficulties in classification, it is good practice to follow up an EMAT inspection with a number of in-field verifications. These investigations allow for a more direct observation of classification and size, and provide valuable information about the nature of cracks. They are, however, expensive — meaning that avoiding unnecessary digs is a top priority.

In this paper, we document a developing approach to post-ILI crack management, whereby the results of an EMAT run are combined with those from field verifications to maximize the amount of information gained from costly field work. This approach — which relies on supervised machine learning — leads to a marked improvement in the classification of crack-like indications from EMAT, and allows future investigations to be prioritized according to the likelihood of finding a concerning defect.

The method was trialed on a pipeline system with extensive SCC, leading to an improved success rate in finding SCC, and a more cost effective crack management plan.

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