Abstract: Aiming at the difference of vibration data distribution of rolling bearing under different equipment and working conditions, and the problem that the traditional deep learning model was difficult to adapt to the inconsistent distribution of data sets, a bearing fault location method based on convolution twin neural network was proposed. Firstly, the twin network was selected as the basic framework to expand the preprocessed equipment data to achieve the purpose of data enhancement. Then, the stochastic pooling ELU-CNN(SE-CNN) model was selected as the feature extractor to extract the in-depth features from the equipment operation data, and the back-propagation algorithm was used to update the model parameters to optimize the feature extraction performance and classification performance of the model. Finally, experiments were carried out on the mixed data set of two data sets, and the fault location results of rolling bearing were obtained by using the trained network model. The results show that the evaluation indexes of fault location are more than 95% using the bearing fault location method based on convolution twin neural network, and the comprehensive data index reaches 0.986 3. Comparing with other advanced methods, the fault location accuracy of the model across equipment and under multiple working conditions is significantly improved by 0.024 6. The model can effectively adapt to the differences of data distribution and has good generalization performance.

Key words: rolling bearing; fault location method; deep learning; convolution twin neural network