by John C. Aldrin

Be sure to read the July 2023 Materials Evaluation special Technical Focus Issue on artificial intelligence and machine learning in NDT. This open-access issue was guest edited by John Aldrin. Following is Aldrin’s introductory letter to this issue.

.

It is my pleasure to share with you this technical focus issue on the subject of artificial intelligence and machine learning (AI/ML) in nondestructive (NDT).

I have had the great opportunity in my career to work actively on this subject over the past 25 years, going all the way back to my graduate work at Northwestern University. That work, performed in collaboration with so many important mentors of mine and others in the NDT community (Professor Jan Achenbach, Glenn Andrew, Charlie P’an, Bob Grills, Tommy Mullis, Floyd Spencer, and Matt Golis) resulted in the successful demonstration of making calls on complex ultrasonic data through a probability of detection study of a neural network–based approach.

There exists great potential with the application of AI/ML in NDT. Such tools can excel at repetitive tasks, performing complex data review faster than inspectors. The vision of AI/ML has been to reduce the burden of laborious data review and ideally eliminate missed calls, ensuring greater reliability. However, the widespread application of AI/ML in NDT has not yet been adopted for a number of reasons. Training deep learning neural networks requires very large, well-understood data sets, which are not typically available for many NDT applications. Also, many promising research demonstrations have run into issues with overtraining or robustness to variability found outside of the laboratory. In addition, while human factors are frequently cited as being sources for error in NDT applications, humans are inherently more flexible in handling unexpected inspection scenarios and are better at making judgement calls.

AI/ML clearly has provided us with so many advances to our daily lives in recent years. For example, Google Translate can translate text well between English and more than 100 other languages, enabling broader communication throughout the world. Apps like Shazam can detect a song being played in seconds. Computers using deep learning routines can beat the best human chess players. How can the NDT community make better use of AI/ML, while ensuring we are doing what is best for our customers and members of ASNT? My goal with this technical focus issue is to highlight progress and success stories, share best practices for AI/ML use, but also discuss concerns and the value of having humans in the loop.

_______

The vision of AI/ML has been to
reduce the burden of laborious data review
and ideally eliminate missed calls, ensuring
greater reliability.

_______

In the first feature article, I write about emerging AI chatbots and explore the benefits and concerns with using them as part of our work in NDT. As well, in the NDE Outlook column, Singh and Garg highlight the importance of balancing innovation and responsibility with these generative AI tools, ensuring a safer world.

The second feature article by Lindgren highlights several successful applications of AI/ML for the Department of the Air Force, introducing best practices for development and validation, and highlighting the critical role for human inspectors to ensure NDT data quality and address outlier indications. The NDT Tutorial article by Harley and Zafar provides a number of helpful tips for effective training and testing of AI/ML for NDT applications. A Review Paper by Taheri and Zafar provides a survey of different AI/ML architectures that are used today and reviews ML progress for interpreting acoustic data acquired as part of additive manufacturing process monitoring.

Two technical papers included in this special issue highlight promising AI/ML research and applications. The first paper by Scott, Stocco, Chertov, and Maev presents progress on the development and performance evaluation of a real-time AI-driven weld process characterization routine reviewing ultrasonic NDT data. In the second paper, Huang, Elshafiey, Farzia, Udpa, Han, and Deng present a promising demonstration of using deep transfer learning with a finite element modeling–based knowledge transfer for an improved acoustic emission source localization demonstration.

I hope you enjoy this technical focus issue and learn something new about AI/ML in NDT. Please feel free to send comments or feedback.

_______

Author

John C. Aldrin, Technical Focus Issue Editor, aldrin@computationaltools.com

Papers featured in this Technical Focus issue of Materials Evaluation:

• Benefits and Concerns of Using Emerging Artificial Intelligence Chatbots with Work in NDT
• Validated and Deployable AI/ML for NDT Data Diagnostics
• Tips for Effective Machine Learning in NDT/E
• Machine Learning Techniques for Acoustic Data Processing in Additive Manufacturing In Situ Process Monitoring – A Review
• Real-Time AI-Driven Interpretation of Ultrasonic Data from Resistance Spot Weld Process Monitoring for Adaptive Welding
• Acoustic Emission Source Localization Using Deep Transfer Learning and Finite Element Modeling–Based Knowledge Transfer

.