An Intelligent Sensing System for Real-Time Bearing Fault Monitoring Based on Vibration Signals

To address the challenge of timely fault identification in industrial rotating machinery, this paper presents the design and implementation of an intelligent monitoring system based on vibration signal analysis. The system utilizes a low-power STM32L433 microcontroller integrated with a triaxial vibration sensor to efficiently collect critical vibration data during bearing operation. A lightweight Support Vector Machine (SVM) model is embedded into the microcontroller to enable on-device classification of various bearing fault types. With the integration of a LoRa wireless communication module, the system can transmit processed results over long distances to a host computer, thereby enabling real-time visualization of node status and remote fault warnings. Owing to the compact architecture and high computational efficiency of the SVM model, combined with the ultra-lowpower characteristics of the microcontroller platform, a single system node can operate stably for over 3 months on two AA batteries. Experimental evaluations demonstrate outstanding performance in bearing fault classification, with a balanced accuracy of 99.46 %, an F1-score of 99.46 %, a precision of 99.47 %, and a ROC AUC of 99.76 %. With a model size of only 19.5 KB, the system proves to be both feasible and practical for deployment in edge-level industrial scenarios.