Fault diagnosis model of variable working condition bearing based on deep learning

This paper presents a fault diagnosis model based on rolling bearing fault diagnosis, which can improve the operating efficiency of the motor and is suitable for variable working conditions and noisy environment. Combined with convolutional neural network bidirectional long short-term memory neural network (CCN BiLSTM), soft attention mechanism and exit technique. Firstly, one-dimensional vibration signal is transformed into two-dimensional fault image by continuous wavelet transform (WT). The original vibration signals collected by rolling bearing under different working conditions were used for pre-training. Second, a attention mechanism is introduced in BiLSTM to extract the more critical deep fault features, and Dropout is used during training to suppress overfitting. The optimized model is then used to input the processed data into the improved model for training. Finally, the Softmax classifier is used for fault diagnosis. Experimental results show that the proposed method has stronger robustness and noise resistance than other methods.