A Novel prediction of remaining useful life time of rolling bearings using Convolutional Neural network with Bidirectional Long Short Term Memory

Detection of bearing problems increases the importance of the service life of rotating machinery. Convolutional neural networks (CNNs) are often used in current research, and databases built on deep learning (DL) models have improved capabilities in the field of defect diagnosis. We use the publicly available Case Western Reserve University (CWRU) dataset to compare the classification accuracy and gain more adaptive knowledge and insights about the proposed approach. Extensive tests and evaluations are performed on the dataset to verify the diagnostic effectiveness of the recommended method in different situations. To demonstrate the superiority of the proposed method, we compare multiple views of the same dataset with similar tasks. CNN supports degraded index sequences to reduce noise and stop temporal oscillations. A new CNN-BiLSTM model is used to capture current and historical inspection data and predict the RUL's service life and supported power levels. Regarding production, we follow health rates. The proposed method was evaluated by accelerating the bearing motion to failure, and the results demonstrated its advantages in terms of more accurate RUL prediction. According to the experimental results, the proposed center distance measurement method is a new and valuable means for intelligent bearing diagnosis. Experimental results using 48 K and 12 K CWRU datasets show that the overall accuracy of the BiLSTM method is 99.80% and 98.3%, respectively, which is better in diagnosis than some popular models.