Improving Pipeline Integrity with Artificial Intelligence

Introduction

Informed machine learning (ML) has been used successfully across a broad range of industries to teach computer programs to consider input and historical data, calculate probabilities, assess potential scenarios, and act on assumptions. The use of ML is not always straightforward or successful, and many ambitious projects have failed to deliver. However, it is fair to say the ML process, when properly used, can be rewarding. The results have the possibility to not only enhance decision-making but they can improve project economics and reduce risk exposure for workers.

Now adapted for pipeline inspections, the technology is being used in the field to predict material loss and provide uncertainties around material loss predictions, enabling more accurate estimates for scheduling of inspection and maintenance and improving the ability to ensure pipeline integrity.

Improving on Traditional Inspection Technology

Pipeline inspection programs are an important part of operations because they help owners maintain equipment, increase uptime/reliability, protect the environment, and ensure asset and worker safety. Smart pigs (aka intelligent scrapers) have been used for nearly 60 years to gather material loss and damage data from pipelines. Outfitted with sensors that transmit data, intelligent scrapers can accurately detect corrosion, dents, cracks, gouges, and areas of the pipe where wall loss has occurred. Sensors record data for each anomaly, which is tagged with coordinates to enable asset owners to identify locations where a pipeline is damaged or where a defect has the potential to become a safety or operational concern.

This technology has served the industry well, but the delay incurred in processing and analyzing the data can be considerable, and that can reduce the usefulness of the data.