Endpoint AI ─ Making Equipment Smarter and Sustainable

Nuvoton’s AI MCU solution adds intelligence to edge devices, transforming customer equipment into smarter, sustainable solutions which create new value and enrich the quality of human life. Our offerings include a line-up of AI MCUs, a robust embedded development environment with seamless technical support for data acquisition, data analysis, AI model development, and final deployment on MCUs. AI anomaly detection is an example of time-series sensing–based AI that shows value creation, actual development, and implementation method of AI on MCUs 

• Features
  

  -	Integration of AI anomaly detection technology with IoT solutions
  -	Detects minute vibration changes
  -	Detects filter clogging using AI
  -	Realtime “On-device” learning

• Specifications
  

  -	Variational Autoencoder (VAE)-based AI
  -	Uses Cortex-M4 MCU (e.g., M467, KM1M4x)
  -	Acceleration Sampling Frequency: 952 Hz
  -	Inference Time: 0.7 ms
  -	On-device learning time: 30 s (Data collection: 25 s; Training: 5 s)

• System Block Diagram