Fujitsu Develops AI for Identifying Product Abnormalities in Manufacturing

Fujitsu has introduced an innovative AI solution designed to detect abnormalities in product appearances, significantly enhancing the early identification of manufacturing issues. This proactive approach helps prevent waste of materials, minimizing both direct costs and environmental impacts. Furthermore, it also mitigates the potential reputational damage and costs arising from product returns or recalls after defects are discovered post-shipment.

The technology utilizes an AI model trained on images of products exhibiting various abnormalities, eliminating the need for real-life defective product images during training. Fujitsu conducted tests at its Nagano Plant, which specializes in electronic equipment manufacturing, and observed a remarkable 25% decrease in man-hours required for inspecting printed circuit boards.

Highly effective, the AI can identify issues such as frayed threads or incorrect wiring patterns with exceptional accuracy, even in products that are designed to have variations, like different colored carpets or uniquely wired electronic components. Fujitsu’s AI achieved an impressive Area Under the Receiver Operating Characteristics (AUROC) score exceeding 98% when evaluating products with deviations from their typical appearance.

Going forward, Fujitsu intends to leverage this AI advancement within its COLMINA brand, which focuses on driving digital transformation in the manufacturing sector.

For those interested in discussing this innovation further, industry leaders will be participating in events such as the 5G Expo, IoT Tech Expo, Blockchain Expo, AI & Big Data Expo, and Cyber Security & Cloud Expo World Series, with upcoming gatherings in Silicon Valley, London, and Amsterdam.