Bearing Fault Diagnosis Using One-Dimensional Convolutional Neural Network

Abstract

In this paper, a fault diagnosis strategy using one-dimensional convolutional neural network (CNN) is developed for rolling bearing. Firstly, each basic unit in the CNN model to be proposed is introduced in detail, and the optimization algorithm required for the CNN is described to show the working principle, which provides a theoretical basis for the one-dimensional CNN model. Next, a series of preprocessing such as overlap sampling and unique thermal coding are performed on the rolling bearing dataset from Case Western Reserve University, and a batch normalization algorithm is proposed to improve the training efficiency and performance of the CNN model. Finally, the designed one-dimensional CNN model is trained, the adaptive ability of the model with variable load is tested, and good results are obtained.