High-Density Flip Chip Solder Bump Bridging Detection Using YOLOv7

Solder bump bridging (SBB) induced during chip attach process could cause electrical failure and device to malfunction. These SBB defects went undetected due to complex circuitry, test detectability and coverage. This paper focuses on AI development using YOLOv7 to improve SBB detectability in manufacturing assembly. We’ve implemented an image quality and alignment check algorithm, augmented the images, tuned the hyperparameters and trained an SBB defect detection model. Our model can generalize across multiple datasets achieving a recall score above 0.97. The experimental results show that our proposed method is effective in detecting solder bumps bridging defects.