Design and Implementation of a custom PCBBased Edge AI System Using STM32 for Real-Time State Detection and Condition Monitoring

Abstract

This thesis presents an edge-computing system for real-time industrial machine monitoring, combining an STM32F103RB microcontroller with NanoEdge AI Studio to Monitor Machine state on-device without cloud dependency. Vibration data from an MPU6050 sensor is captured via DMA, analyzed by embedded AI, and displayed locally on an OLED while being relayed to a web dashboard (via CAN bus and ESP32). A custom 6-layer PCB and modular firmware architecture ensure low-latency processing and scalability. The solution demonstrates how lightweight embedded AI can enable cost-effective predictive maintenance.