Fault Diagnosis of Rolling Bearing Based on DBN With Dynamic Addition Algorithm

Aiming at the uncertainty of the depth belief network (DBN) layer, the fault diagnosis accuracy is not high. A DBN diagnosis method based on dynamic addition algorithm is proposed. Firstly, the paper firstly determines the number of hidden layers by dynamic adding algorithm, and then sets the number of hidden layer neurons in the model according to the principle of layer-by-layer decreasing; Taking the rolling bearing as the research object, the classification error curve of the training sample and the test sample is analyzed to show the diagnostic accuracy of the DBN method based on the dynamic addition algorithm on the rolling bearing fault, and the diagnostic performance of the DBN model with different depths is compared. The method has obvious advantages in fault diagnosis of rolling bearings.