An Improved Bearing Fault Investigation Scheme Using 1D CNN with PCA and SVM

This study presents a novel approach for accurate motor bearing fault detection through a sequential process involving one-dimensional convolutional neural network (1D CNN) for feature extraction, Principal Component Analysis (PCA) for dimensionality reduction, and Support Vector Machine (SVM) for classification. The main objective of this approach is to improve the model performance for complex fault combinations by achieving high precision and recall scores. This study has been carried out using online available datasets of bearings faults provided by Bearing Data Center, Case Western Reserve University (CWRU), USA. In this study, bearing inner race, outer race and rolling ball defects are analysed under variable load operations using the proposed 1D-CNN scheme. This research highlights the potential of the proposed model to advance motor condition monitoring systems, offering a reliable and efficient method for fault detection across diverse operational scenarios.