Keeping counterfeit parts out of the supply chain

The Aviation Supply Chain Integrity Coalition released recommendations for preventing the introduction of unapproved parts into the aviation supply chain. Members of the coalition include senior representatives from aircraft and engine original equipment manufacturers (OEMs), airlines, and a maintenance, repair, and overhaul (MRO) firm. Thirteen recommended actions offer short-, medium-, and long-term solutions to vendor accreditation, document traceability and verification, and non-serialized parts traceability.

However, the recommendations don’t go far enough, according to Roei Ganzarski, who identified shortcomings in the report.

Ganzarski has held leadership roles at magniX, Boeing, and Eviation, and created the software company Alitheon to eliminate fraudulent, counterfeit, and gray market parts from the supply chain using optical artificial intelligence (AI).

He asserts the recommendations continue industry reliance on documentation that won’t prevent fraudulent parts from entering the marketplace. Ganzarski believes improving parts traceability should be an immediate goal.

Reliance on being able to identify and authenticate an aircraft part based solely on the paperwork that goes with it isn’t enough. “We want to get the aviation industry to a point where the part is its identifier, not the paperwork,” Ganzarski says.

One may think a digital record would be better than a paper record and not as easily faked, but that’s not necessarily so. “A proxy – a hologram sticker, a tag, engraving – can be a counterfeit,” Ganzarski explains. “If you want to protect data, then you want to start using blockchain, but at best, you’re protecting the data,” he continues. Fake data can be input into a blockchain or digital twin. Even if the data is authentic, what connects it to the part – a QR code or a hologram sticker – may not be. “If someone fakes or removes it, you have data that’s correct but connected with a fraudulent part,” he says.

Ganzarski thinks the solution is a manufacturing fingerprint. “Fingerprints have good attributes: they’re unique, persistent, and inherent. More importantly, I can’t give it to anyone. I can give a badge to someone who pretends to be me.”

Because parts are manufactured within tolerances, slight surface irregularities happen in the manufacturing process. As long as a part’s within tolerance, it performs the same.

“The likelihood of having two parts with the same manufacturing attributes is about one in 3.5 trillion,” Ganzarski explains. “People have been trying to use the fingerprint aspect of manufacturing as an identifier for a long time without success. We’ve developed our FeaturePrint system with an off-the-shelf industrial camera or mobile phone that, along with our algorithms, can identify and serialize parts. We only have to take one two-dimensional image to find the 5,000 parameters needed to identify parts.”

AI processes the image within seconds. “It works even with only a percentage of the part, scratches, etc., and we’ll do re-registrations in real time.”

“I’m just hoping our industry doesn’t wait for some catastrophic event from a fraudulent part and then say, ‘Oh, we should have done something,’” Ganzarski adds. – Eric