Fault diagnosis of low-speed heavy load super large rolling bearing based on deep learning

The conventional eigenvalue alarm mode has a high rate of false alarm and missed alarm for the low-speed heavy load super large rolling bearing. Besides, the traditional signal processing method such as envelope spectral analysis is difficult to extract its fault characteristic frequencies, resulting in a high rate of false diagnosis and missed diagnosis. In order to solve the above problems, an intelligent diagnosis method for the low-speed heavy load super large rolling bearing based on deep learning is proposed. The proposed method mainly utilizes the strong robustness of deep learning algorithm to the quality of original vibration data in the field of fault diagnosis. Firstly, an effective signal acquisition scheme is designed to solve the problem that the signal characteristics of low-speed heavy load super large rolling element bearing are difficult to be acquired. Then, the collected data are randomly divided into training sets, verification sets and test sets by using data enhancement technology. Subsequently, input the divided training set samples into 1-dimensional convolution neural network (1DCNN) deep learning model for learning and training to construct the 1DCNN learning model and set network structure parameters. Meanwhile, the optimal training model is obtained by validating the updating effect of model parameters through validation set. Finally, the test data is input into the trained model to realize intelligent diagnosis. Effectiveness of the proposed method is verified by the vibration data of a wind power main bearing.