Using visual algorithms to check components for authenticity.
Counterfeit electronics is a multibillion-dollar industry worldwide. The challenge is finding them, and this is where Israeli startup Cybord is working to gain a foothold.
The company has developed an AI-driven solution that checks for counterfeit parts during product assembly. “It’s a huge task to check electronic components, said Cybord CEO Zeev Efrat. “It’s not capacitors only, or resistors only, or chips only. It’s everything.”
Artificial intelligence (AI) has advanced to the point where it can identify these fake parts, said Efrat. Cybord’s visual software verifies the authenticity of each component through its measurements, date code, lot code, and batch. It also detects lead age, reprogramming, and external damage.
“We are providing traceability, based on a real image of every component,” said Efrat. “That means means if you have a problem, we are able to trace exactly which components are affecting your product.”
From a financial perspective, this is a big deal. An estimated $7.5 billion is lost each year due to phony parts. This problem is alarming to manufacturers and product designers, as well the military. But in the past, the recall process was inefficient.
Cybord contends that the best way to simplify and speed up this process is to ensure that all components are in good condition and functional, which can significantly lower the odds of failure.
“Today, a lot of scrap is because the bad components are sometimes never found, or sometimes found only at the end of the manufacturing process. We are enabling surgical traceability, and we are enabling high productivity during the manufacturing process,” Efrat said.
As with any good AI algorithm, the challenge is finding sufficient good data. Cybord claims to have compiled data for billions of electronic parts in its database, analyzing about 250 million components a month. And that’s still only “the tip of the industry,” said Efrat, who notes that the system is 98.5% accurate.
Cybord already is working with paying customers. So far, it has garnered $5 million in cash from two rounds of funding.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
How does a component/piece part image prove traceability? It seems to me that traceability is conferred via signed trace documentation from the Original Component Manufacturer or the OCM’s Authorized Distributor to the buyer.
Traceability is not made by just the image of a single component.
The Image is always related to the reel it came from, to the board it is placed on.
The documentation used today is not related to a single component but to a reel; therefore, the practice is that when you have a problem, the best is to retrieve all components on the reel.
This is done only if you have a traceability system already installed on your SMT line; if not, it will be the whole batch.
Cybord traceability is based on the image and the related board; therefore, we can identify the board affected by the single component.
Another point you take as an assumption is that documentation is correct in facts; it is not always accurate in many cases because of manual entry, wrong barcode reading, counterfeit, and manufacturing constraints of splicing; you only know the spread of boards, not the unique board.
Cybord Traceability checks the authentication of each component, and if the documentation, date code, and lot code are not as expected, it will flag that; the process is fully automated with no human intervention, so zero trust.
Thanks for the reply. At the end of your response, you indicate: ” Cybord Traceability checks the authentication of each component, and if the documentation, date code, and lot code are not as expected, it will flag that; the process is fully automated with no human intervention, so zero trust.” Practically speaking, parts from some Original Component Manufacturers are not marked the same as the stated Certificates of Conformance or Traceability, either due to small part sizes or OCM part marking practices. How does Cybord Traceability checks account for those marking differences?
Hi Mark,
Cybord’s advanced deep network that learns from diversity
Cybord developed a unique visual Deep-AI algorithm that analyzes each component based on a vast database of billions of electronic components images. This process provides accurate analysis utilizing a high-speed and efficient method.
Check our website https://cybord.ai/technology/
and I would appreciate explaining deeply more on our technology and solution with a brief call
[email protected]
Asaf Jivilik
VP of Marketing