Development of Rolling Bearing Health Diagnosis and Prediction System Using MEMS Accelerometer Vibration Sensing Module

This work implements an in-house fabricated, cost-effective MEMS piezoelectric accelerometer module to acquire the vibration data for fault diagnosis of a rolling bearing in an electric motor. Choosing the right combination of data pre-processing method (Ensemble Empirical Mode Decomposition, EEMD) and machine learning model (Back Propagation Artificial Neural Network, BP-ANN) offers 99.61% accuracy in detecting the location and state of various bearing faults under varying motor speeds of 500, 1000, and 1500rpm, with fewer pre-processing steps. The developed module features excellent compatibility with industrial 4.0 intelligent manufacturing applications.