A Novel Method of Bearing Fault Diagnosis for Train Bogie Transmission System Based on MpResCNN-BiLSTM Model With Attention Mechanism

Safety is a crucial prerequisite for the operation and development of rail transit systems. With the rapid growth of the rail industry, train speeds are increasing, and operational densities are becoming higher, thereby placing greater demands on the safety and reliability of train operations. As a core component of the train’s power system, the health of the bogie transmission system bearing is directly linked to the safety of train operations and transportation efficiency. To address these challenges, this paper proposes a fault diagnosis method for rolling bearings using Generative AI techniques, specifically the MpResCNN-BiLSTM model enhanced with an attention mechanism. First, the method integrates multi-scale cascaded midpoint residual blocks with a bidirectional long short-term memory network (BiLSTM) to extract multi-resolution features from the raw data. Second, the model leverages the exponential linear unit (ELU) activation function and attention mechanisms to improve expressiveness, generalization, noise resistance, and feature selection capabilities. Experimental results demonstrate that the proposed model achieves 100% accuracy in noiseless and low-noise environments, and 99.57% accuracy under high-noise conditions (signal-to-noise ratio = 1). Compared with other traditional models, the accuracy of this model is significantly improved. These results effectively validate the superiority and feasibility of the method for bearing fault diagnosis. This paper presents a novel and effective deep learning method for bearing fault diagnosis, with considerable potential for widespread application and further innovation in intelligent transportation technologies.