Real-Time Vibration Detection and Signal Analysis of Machines
Abstract
This work is a condition-based maintenance low-cost real-time vibration monitoring system of industrial machines based on the ESP32 microcontroller and MPU6050 triaxial accelerometer. The axis-wise and overall RMS vibration amplitude is calculated in the system by time domain processing, which offsets them based on a specified threshold that is installed in MATLAB and which is also embedded in the profile parameters per the requirements of vibration severity. The FFT-based real-time spectral analysis is used to detect the vibration frequency components that are predominant in mechanical faults. The machine condition is automatically determined to be safe, warning or dangerous based on quantitative factors such as mean RMS vibration and danger ratio based on time. Multimachine experimental results record values of RMS vibration between 0.02 g and 0.78 g and a predominant low-frequency vibration signal between 0.3 and 3 Hz. The proposed system was implemented with classification accuracy of 91.67%, analysis latency of less than 100ms. The entire self-sacrificing system has a cost of less than 15 USD, making it fit the limited resource-based industrial settings.