Research on end-to-end fault prediction method of rolling bearing based on deep learning

In rolling bearing fault prediction, the selection of time-domain and frequency-domain features is often influenced by subjective factors, and the full utilization of time and space features is challenging. To address these issues, a method based on Convolutional Neural Network (CNN) and Bi-directional Long Short-Term Memory (BiLSTM) is proposed. The method involves two main steps. Firstly, a convolutional neural network is used to extract original vibration signal features. A batch regularization layer is added after the convolutional layer to optimize weight and accelerate model training. The efficiency of feature extraction is improved by extending the first convolutional layer and adjusting the step size. Secondly, a bidirectional long-short-term memory neural network is introduced to enhance the utilization of temporal information and extract temporal features. The model's robustness is strengthened through the incorporation of batch regularization layers and dropout layers, reducing data-to-data dependencies. The proposed method is validated using two sets of rolling bearing test data. The results demonstrate the improved fault prediction accuracy compared to traditional methods, along with better performance under different working conditions.