News Releases 2017Meiden starts selling small AI edge computer capable of detecting equipment anomalies and conducting predictive maintenance

Meidensha Corporation (Meiden) has started selling a small-size edge computer that collects sensing data from various automated equipment, such as those for manufacturing semiconductors, and detects anomalies in the equipment based on what has been machine-learned by artificial intelligence. This new product will allow users to conduct predictive maintenance.

Transmitting a vast amount of sensing data gathered at a manufacturing factory to “the cloud” can be delayed due to limitations in the network bandwidth. These delays compromise real-time monitoring. However, handling sensing data with the small AI edge computer at the endpoint allows real-time monitoring and prevents a slowdown in transmission speed. Furthermore, AI learning made it possible to reduce false alarms regarding equipment anomalies and ensure highly accurate anomaly detection.

<AI learning procedure>
1. AI edge computer is connected to target equipment to collect data
2. AI edge computer transmits collected sensing data to AI machine-learning server
3. AI learning server conducts machine-learning offline to produce AI parameters
4. Produced parameters are written into AI edge computer
5. AI edge computer autonomously judges if sensing data are normal or not

1. The AI edge computer can autonomously make judgments.
2. Meiden is developing a model compliant with SEMI Equipment Communications Standards, a communications protocol for semiconductor equipment, and aims to complete it in January 2018.
3. Its compact size (200 mm high, 140 mm wide and 47 mm deep) enables easy installation.

<Fields of application>
Automated production lines, such as equipment for production of semiconductors.

Contact at Meiden on this press release:

General Affairs & Public Relations Division
Tel: +81-3-6420-8100