Modern ILI tools record vast amounts of data. This data is provided by indirect measurement methods such as with MFL or EMAT technology and thus needs to be interpreted by subject matter experts. The field of pipeline ILI diagnostics thus has a lot in common with that of medical diagnostics with CT or MRT tools. Both fields share the challenge to securely detect anomalous feature from often multiple indirect information channels and to consecutively classify and size these anomalies appropriately. Nowadays, it is standard that medical diagnostic experts such as radiologists are successfully supported by different AI assistance systems. It is therefore not surprising that AI systems start to play a central role in the evaluation of ILI data as well.

In this contribution, we provide an overview over AI use-cases to improve the evaluation of ILI data at ROSEN. In particular, we show how Deep Learning, a subfield of AI, is tailored to help with the core challenge of ILI data evaluation: the detection and classification of specific patterns in signal data. Using the example of EMAT crack detection services, we show how AI products help to ensure the quality and consistency of results as well as to reduce reporting times. Finally, we discuss AI-design principles for ILI data evaluation und how AI models are validated.