TOKYO, June 29, 2026 — OKI has developed an AI-based inspection technology that reduces the time required for manual checks after automated optical inspection, or AOI, by about 80% for large, high-density printed circuit boards used in products such as AI servers, the company said on Monday.

The technology will be introduced to production lines serving customers of OKI's Marugoto EMS electronics manufacturing service from July 1, 2026. It is designed to reduce false detections of solder defects generated during AOI after components are mounted on printed circuit boards, while also improving inspection accuracy.

The system incorporates OKI-developed algorithms built from manufacturing know-how for large, high-density boards into AOI judgment programs. It uses AI to inspect four categories: solder wetting, misalignment, missing components and lifted parts.

OKI also built in a proprietary program that treats some component-specific variations as acceptable when they do not affect function, including differences in serial or lot numbers and faint printing. The system also uses a "good-product-only learning" method, in which AI is trained only on non-defective data and judges products based on the degree of deviation from that baseline.

The company developed the technology for high-mix, variable-volume manufacturing environments handling tens of thousands of material types and thousands of equipment types. In recent years, semiconductors mounted on printed circuit boards have become larger, finer and more layered as AI adoption has expanded, making solder-defect judgments more difficult, particularly in AI servers that use many large parts and narrow-pitch components.

• Prev
• Next