Smart Bearing Diagnosis System using MEMS Accelerometer and Neural Network Analysis

This study introduces ball bearing fault diagnosis method that integrates the power spectrum analysis and artificial neural network (ANN). In this research work a test workbench is indigenously developed with embedded PSoC controller and LabVIEW virtual instrumentation. It employing a high-sensitivity MEMS accelerometer to measure vibrations on bearings testing, it collects the data automatically based on the vibration. The amplitude and power spectrum were generated from the data, which is inputted into a trained ANN classifier in real-time fault diagnosis. It explores the faults in ball bearings through the investigation of good working, and defective bearings that have different types of faults such as inner race defect, outer race defect, and ball defect. The ANN model was trained on 200 bearing samples and tested and it had a 96.5% classification rate, precision. The experimental findings indicates that the combined ANN-based system is much better than the traditional manual inspection by 65% accuracy and power spectrum analysis alone provides 88% accuracy, and diagnostic time is reduced to 5 seconds per bearing instead of 30 seconds. This setup provides a promising solution which is automated predictive maintenance in manufacturing industries.