A comparative Analysis of 1D Convolutional Neural Networks for Bearing Fault Diagnosis

Abstract

In many industrial applications, bearing monitoring is ranked as a crucial duty for condition-based maintenance. It enables the avoidance of unplanned maintenance while lowering the cost. For this objective, a variety of strategies have been devised to assure accurate and effective monitoring. This study offers a One-Dimensional Convolution Neural Network (1D-CNN) that is used to diagnose and identify early bearing faults based on Variational Mode Decomposition (VMD). Methods such as VMD-based feature extraction for defect detection are followed by multi-scale feature extraction for classification and diagnosis using CNN and pooling layers. Bearing fault detection and diagnosis are evaluated using the Case Western Reserve University experimental dataset to determine its robustness and performance. Demodulation techniques and machine learning algorithms are also contrasted for their ability to identify and diagnose the fault. A VMD filter and 1D-CNN approach for monitoring bearing fault are clearly promising, according to the results obtained from the experiments.