A novel pipeline inspection tool based on inductive sensing and machine learning is developed to detect and characterize internal corrosion in metallic pipelines with high performance. The tool leverages inductance-to-digital converters (LDCs), eliminating the need for transmitter coils or magnets. This design offers a new, low-power, and cost-effective solution capable of detecting defects as small as 8 mm in diameter size and 2 mm in wall-loss depth. The tool’s robustness allows it to function effectively in harsh environments, including high-temperature conditions.

To enhance the characterization of defects, a hybrid neural network (HNN) is employed with novel and advanced data processing techniques. The HNN combines convolutional and recurrent subnetworks to extract the multitude of spatio-temporal features, enabling the generation of highly accurate visual representations of defects shapes and penetration depths and therefore their full characterization along the pipe’s entirety. A test loop setup is constructed, where field used spools with natural corrosion are examined using the developed tool. Results from these test runs validate the system's accuracy and efficiency, showcasing its high potential for real-world pipeline inspection applications.

This tool provides a scalable and reliable method for early wall-loss detection, with the potential to significantly reduce the risks of product leak and operations downtime.