Rolling element bearing fault diagnosis for rotating machinery using vibration spectrum imaging and convolutional neural networks

In this paper, we propose a novel method for the classification of bearing faults using a convolutional neural network (CNN) and vibration spectrum imaging (VSI). The normalized amplitudes of the spectral content extracted from segmented temporal vibratory signals using a time-moving segmentation window are transformed into spectral images for training and testing of the CNN classifier. To show the efficiency of the proposed method, vibratory data for healthy and faulted bearings operating at different speeds are collected from an experimental test bench. The classification accuracy, variable load and speed testing, generalization, and robustness by adding noise to the collected data at different levels (SNR) are then evaluated. The obtained experimental classification results show excellent performance in terms of both accuracy and robustness.