Mouse PCB Inspection
Verify correct assembly, detect defects, and read text on PCBs in one operation
Graphical programming environment for deep learning-based industrial image analysis
A printed circuit board (PCB) is a complex assemblage of components, solder connections, substrate, and printed text. The large number of components and connections means a large number of possible defects, but their complexity and component density make visual inspection difficult.
Subcomponent inspection to confirm the correct components are in the necessary positions, detect possible solder problems, scratches, component misalignment, and other defects, and read and confirm the text characters on the board might take three separate stations. Often there is insufficient space, the expense is too high, or the production delays are too great, so some manufacturers feel forced to accept a high error rate and pull completed assemblies when functional testing reveals an error, or even live with a higher rate of returns.
The multiple aspects that must be inspected, given the complexity of the boards, is almost impossible to do adequately with conventional machine vision.
Cognex Deep Learning quickly and reliably solves PCB assembly verification. It trains on image sets of both good and bad PCBs. Three different deep learning tools inspect these boards at a single station, in a unified operation that does not impose production delays.
The assembly verification tool checks whether all components are present in the correct positions. The defect detection tool marks any solder problems, damage to components, chips in the board, or other flaws. The OCR tool reads all of the text characters on the board and components, and outputs a text string for the read characters.
These deep learning tools can quickly be retrained as different PCBs are inspected, or designs change, without the need for any programming.