As the volume of industry inspection data increases and the number of qualified data analysts stagnates, the case for an artificial intelligence (AI) data analysis aid becomes economically imperative. Even off-the-shelf AI analysis aids must be customized for specific applications based on PAUT data, typically from physical mockups containing prototypic flaws of concern.

Attend this talk to learn about the following: Beam modeling to ensure the optimal techniques were selected for inspection of welds in 6-inch Schedule 120 304 stainless steel pipe welds. Mockup fabrication using a state-of-the-art wire EDM process to create economical, realistic, and highly controllable crack-like material discontinuities (and to see their resulting PAUT indications). PAUT data acquisition, including best practices to maximize compatibility with AI aids. Procedures and policies to generate the required data to build AI aids for indication detection, identification, sorting, classification, and sizing. Human and AI-aided PAUT data analysis, including comparison with destructive evaluation (metallographic) results.

Work is ongoing, including full implementation of the AI aid and comparison to the performance of human analysts. Final results are expected to be published in 2026.