MACHINE LEARNING BASED FAULTY BEARING DIAGNOSIS IN CNC MACHINE

Abstract

The prediction of faulty bearing in rotating machineries like CNC machine, induction motor, wind turbine etc. is very important. Bearings are essential parts of such machines and mechanical systems to reduce friction between moving parts and to support the weight of rotating machineries. The noise produced by the machine can make it difficult to detect a fault or diagnose a problem. This is because the noise can mask or obscure the signal that would indicate a fault. To overcome this challenge, researchers may need to use advanced signal processing techniques to separate the signal of interest from the background noise. In this proposed work the vibration signal responses of CNC machine bearing was studied during faulty and normal bearing conditions. Early faulty bearing diagnosis was made using Support Vector Machines (SVM) to identify whether a bearing is faulty or not, what type of fault it has (inner race, outer race, or rolling element fault). This model is effective when there is a clear boundary between the classes by finding a hyper plane that separates the data into different classes. To decompose the signal Fourier transform is used to analyze signals in the frequency domain. It decomposes a signal into its constituent frequencies. Once the model is trained and tested, we can visualize the accuracy, precision, recall, and F1 score using confusion matrix to show how many normal and faulty behavior instances were correctly or incorrectly classified by the model.