Wireless Predictive Maintenance for BLDC Fans using STM32

Abstract

This paper demonstrates the feasibility of combining STM32 with Edge AI - an integrated AI processing on the device, enabling real-time decisions without internet dependency - to achieve accurate and low-latency predictive maintenance, where quick detection and response are critical to prevent costly machine failures, especially in industrial settings. Predictive maintenance is a preventive approach that ensures the smooth functioning of a machine by avoiding breakdowns. This is done by detecting abnormalities in the normal day-to-day functioning of the machine. This enables minimizing downtime and maintenance costs as well as improves the production/operation efficiency of the machine. Edge AI is a unique tool that provides real- time anomaly detection with immediate responses instead of relying on cloud-based alternatives, which can introduce delays. By analyzing data from the GY-521 gyroscope-accelerometer module, the system identifies different fan behaviours and categorizes them into three different states: "Normal condition", "Maintenance required soon", and "Critical fault". The ESP32 hosts a web server that displays the fan’s condition through a user-friendly interface, allowing remote monitoring.