Publication

Abstract : This study describes a real-time vibration and temperature monitoring system that uses an analog temperature sensor, an MPU6050 MEMS accelerometer, and an STM32 microprocessor. To track machine operating conditions, the suggested system continuously collects heat readings and tri-axial vibration data. The sensor data is processed to obtain time-domain statistical features for condition assessment, including mean, root mean square (RMS), standard deviation, and peak-to-peak values. To differentiate between problematic and normal operational states, a machine-learning-based categorization approach is utilized. TinyML is used to install the learned model on the STM32 microcontroller, allowing for entirely edge-based real-time inference independent of the cloud. Experimental results reveal that the system reliably detects vibration disturbances and anomalous temperature rise, enabling prompt problem indication through LED alerts and serial monitoring. The suggested technology offers a low-cost, low-latency, and efficient approach for embedded condition monitoring and predictive maintenance applications.