AI Visual Detection of Defects and Corrosion on Leads in Electronic Components

Join the CALCE/SMTA Counterfeit Parts and Materials Symposium 2022 for this and other informative presentations

Eyal Weiss

Cybord

Abstract: A large-scale evaluation of the quality of electronic components at the time of the electronic board assembly is presented. Counterfeit components are often recycled or old components. Therefore, the quality of components and the soldering leads is a good indicator of the component’s authenticity. The quality of the components is evaluated based on their visual appearance by quantifying their visual defects and the corrosion evidence as they appear on the component and its soldering leads. A machine learning method to detect body defects and evidence of corrosion on soldering leads is presented. Over 11 million components images were inspected by the AI algorithm. We have shown that 290 components out of a million had body visual defects that cannot be seen by conventional AOI. In addition, over 1,100 out of a million had visible corrosion evidence on their soldering leads. Corrosion on the soldering not only affects the production yield but is the most common cause of random statistical failures in the field resulting in products failure. The presented method allows inspection of all the components used in production thus reducing the risk of failures in the field caused by poor quality electronic components originating from counterfeit, bad storage, or handling.

Bio: Eyal Weiss is CTO and Founder at Cybord. Cybord protects companies from infiltration of counterfeit or malicious hardware-cyber components into their products by using big-data, AI and in-line image processing on all the assembled components. Estimates are that 5 to 10 percent of components are counterfeit components. Cybord’s technology enables companies to test 100% of the components and gain an unparalleled level of quality, security, and reliability, reducing the return merchandise authorization (RMA) of products by more than 30%.