Magnetic Flux Leakage (MFL) continues to be the most used ILI technology since its widescale introduction in the 1980’s. Thousands of MFL inspections are completed each year by pipeline operators as a reliable and effective assessment method to determine current pipeline condition for monitoring, future integrity assessments and repair programs.

Over the years, technology has advanced with caliper and IMU data gathered efficiently in the same inspection. Advances have also been made in both resolution and measurement of the three leakage fields that occur in response to corrosion defects. Measurement improvement has benefited detection, characterization and sizing of smaller, more challenging corrosions that otherwise might lead to outliers from specification leading to potential safety risks.

This paper will briefly explain some of the advantages tri-axial measurement has brought to standard MFL analysis, but importantly it will look deeper into how multiple measurements taken at defects have enabled a step change in the application of AI to improve corrosion measurement sizing accuracy. This AI capability has introduced the possibility of defect-by-defect tolerance prediction and improved corrosion growth estimation. Together, these advances will reduce conservatism, leading to unnecessary repairs, predict where tolerances may be greater and remove unknown risk.

It covers technical development work Baker Hughes has conducted by applying deep learning techniques to large areas of raw MFL data to predict complex defect morphologies and improve burst pressure estimations in situations typically challenging for standard analysis. Additionally, how AI is being used to predict new corrosion during run-to-run assessment to consider corrosion cluster growth more accurately as a basis for future integrity assessment.

Finally, techniques being adopted to effectively monitor ILI performance to specification, how it's used to identify when inspections may have not met operator expectation and how that data can be monitored to target continuous improvements to system performance.