Abstract
Predictive maintenance helps reduce operational costs and improve machine reliability by anticipating failures. However, existing solutions are often too expensive or complex for small rotating machinery such as fans or low-power motors. This work presents a low-cost, IoT-based monitoring system using an ESP32 microcontroller combined with MEMS sensors (an accelerometer and a microphone). The system continuously collects vibration and acoustic signals, which are then processed using RMS and FFT techniques. Machine learning algorithms, such as anomaly detection or basic classification, are used to identify deviations from normal operation. A working prototype was tested under various fault conditions, including imbalance and wear. The system successfully identified abnormal states through signal deviations in both time and frequency domains, with over ~73% detection accuracy. The proposed solution is cost-effective, simple to implement, and well-suited for educational or industrial environments. It demonstrates the potential of embedded systems and basic signal analysis for scalable predictive maintenance applications.