AI-Powered Technique for Bearing Condition Monitoring and Fault Diagnosis in Mixer and Grinders

Bearings and other rotating components play a crucial role in industrial machinery, but they are susceptible to defects caused by factors such as wear, improper installation, and manufacturing errors. It may result in downtime, production losses, safety hazards, and costly repairs. Vibration-based condition monitoring is a commonly employed technique that yields valuable insights into the health status of equipment. Previously, single labelled fault diagnosis was carried through Neural Networks and other machine learning techniques using handcrafted features like Fast Fourier transforms, Wavelet transforms, Short-time Fourier transforms, and statistical moments. Mixed faults (the occurrence of two or more faults simultaneously) have received less research attention in comparison to single faults. The present study implements an AI-powered technique for bearing condition monitoring and fault diagnosis in mixer and grinders with employing any preprocessing techniques. Raw vibration time domain data has been acquired through virtual instrumentation (VI) in LabVIEW. After storing the signal seven different appropriate statistical features are extracted from signal. Then these obtained features will be supplied to classification methods in machine learning. Classification algorithms used in the project are Logistic Regression, KNN, Random Forest, SVM, Kernel SVM, Decision Tree. Classification methods will classify these signals into various faults and hence we can calculate the efficiency of feature-classifier. I have studied results and plotted the graphs for results.