Fault Diagnosis of Rolling Bearing based on CNN-GRU

Bearings play a vital role in various rotating equipment, including motors, engines, and wind turbines. They support and enable the rotation of additional loads and are exposed to factors such as vibration, impact, and high-speed rotation during operation. Therefore, the proper functioning of bearings is essential for equipment reliability and operational efficiency. Bearing faults can lead to equipment shutdowns, noise, vibrations, and ultimately result in equipment damage or accidents. Hence, the timely and accurate diagnosis of bearing faults holds great significance. Convolutional Neural Network (CNN) models have robust feature extraction capabilities, while Gated Recurrent Unit (GRU) models demonstrate excellent classification abilities with high algorithmic efficiency. To combine the advantages of both, this study presents a fault diagnosis network model called CNN-GRU based on CNN and GRU. Ten different types of fault data under the same operating conditions are selected from a bearing fault dataset for validation. A comparison is made between the proposed CNNGRU model and traditional CNN and GRU models. The results demonstrate that this method successfully addresses the issue of traditional methods relying heavily on manually set features and shows remarkable improvements in terms of accuracy and algorithmic efficiency.