An intelligent of fault diagnosis and predicting remaining useful life of rolling bearings based on convolutional neural network with bidirectional LSTM

The importance of the quality of life of rotating machinery increases the Bearing fault diagnosis. Deep learning models (DL)-based databases become increasingly smart in the field of fault diagnostics, the latest research has widely used CNNs (convolutional neural networks). This paper proposes a new way to diagnose bearing failures with CNN with Bilinear LSTM. Traditional CNNs are however not easy to detect defects due to the fixed geometry of complex fault diagnosis with different working conditions. Our primary and secondary classifiers at specified layers replace primitive shape convolutions with reconfigurable convolutions, resulting in classification results with stringent feature time-frequency incompatibility and a larger receptive field. To acquire more adaptive knowledge and insight into the proposed approach, we employ the CWRU (Case Western Reserve University) opensource dataset to compare classification accuracy. The bearing dataset has been subjected to comprehensive experiments and evaluations in order to confirm the efficacy of the suggested technique's diagnostic performance in a variety of settings. By comparing multiple perspectives on the same dataset with related tasks, the proposed method's superiority is proved. To limit the effect of noise and avoid temporal oscillations, degraded index sequences are matched with a CNN. Current and previous inspection data are fed into a new CNN-BiLSTM model, which is then used to predict the useful time and compatible power values of bearing RULs. When it comes to output, go with the lifetime percentage. The proposed method has been tested by accelerating bearing operation to failure, and the results show that the method has advantages in predicting RUL more accurately. The results of the experiments suggest that the proposed core distance measurement method is a viable new tool for intelligent rolling bearing diagnosis. The BiLSTM technique is more diagnostic than some generic models, according to experimental results using the 48 K and 12 K CWRU datasets, with overall accuracy of 99.80% and 98.3%, respectively.