AI Before It Was Cool: How the ARIES System Outperformed Human Inspectors in Aircraft Eddy Current Testing

by Edward Korkowski - eddycurrent.com.

Long before artificial intelligence took over headlines, a team of forward-thinking engineers quietly revolutionized aircraft inspection. The year was 1991. The problem? Low-cycle fatigue cracks in aircraft compressor disk bolt holes—tiny, dangerous flaws that even skilled human inspectors routinely missed.

Their solution? An AI-powered eddy current inspection system called ARIES.

🚨 The Real-World Challenge: Tiny Cracks, Big Risks

Aircraft engine components like compressor disks endure tremendous stress cycles. Over time, this leads to low-cycle fatigue (LCF)—the gradual emergence of microscopic cracks at fastener holes. Many of these cracks are less than 0.125 mm deep.

Human inspectors scanning hundreds of bolt holes often miss the faint signals these cracks produce. Even with visual data review, the probability of detection (POD) was often unacceptably low.

🤖 Enter ARIES: Automated Real-Time Intelligent Eddy Current System

The ARIES system, developed by Chapman, Fahr, Pelletier, and Hayt, was designed to overcome these limitations. Here’s what made it revolutionary:

• Rotating differential probe (4.4 mm diameter) scanned each bolt hole at six vertical levels, operating at 660 kHz.

• An XYZ scanner ensured precise positioning (0.005 mm accuracy).

• A Hewlett-Packard computer sampled data at 7 kHz, capturing the entire waveform.

• A trained AI pattern recognition engine (ICEPAK) processed every signal in real time.

Instead of relying on threshold alarms or static templates, ARIES classified signals using a K-nearest neighbor algorithm, leveraging features like amplitude, power, and autocorrelation.

📊 The Results? Nothing Short of Groundbreaking

Inspection Method	Cracks Detected
Human Inspector	13
Human + Visual Review	17
ARIES AI System	29
SEM Confirmation (actual)	33

ARIES missed only 4 of the confirmed cracks—all less than 0.1 mm deep. The human inspector, even with post-analysis, missed 16 cracks, including some as deep as 0.5 mm.

This wasn't just a lab experiment. It was a fully operational production system that outperformed trained humans in both sensitivity and repeatability.

💡 Lessons for Today’s NDT Professionals

• Smart ECT isn't new. It has decades of history grounded in rigorous engineering and smart signal processing.

• AI works best when paired with good sensors and signal conditioning. ARIES succeeded because it combined hardware precision with intelligent software.

• Feature extraction matters. Amplitude alone is rarely enough—domain features like autocorrelation and energy made ARIES effective.

• 
  

🔍 What Happened to Systems Like ARIES?

The core principles behind ARIES have quietly shaped modern eddy current testing. Many of today’s automated bolt hole scanners and surface array tools owe a debt to early AI inspection systems like ARIES.

But here’s the kicker: most of the NDT world still hasn’t caught up.

Too many inspections still rely on amplitude thresholds or manual interpretation. If you’re in the eddy current field and not thinking about automated waveform analysis, classifier design, and signal feature fusion—you’re a few decades behind.

📎 Want to Learn More?

You can read the full historical write-up in Appendix R of the upcoming eBook Libby: A Modern Teaching Edition—a curated archive of the best thinking in electromagnetic testing history.

And for more on advanced eddy current testing, machine learning, or to find the equipment that bridges old-school physics with next-gen software, visit:

👉 eddycurrent.com — the only website dedicated to all things eddy current testing.