Research on Fault Diagnosis of New Energy Vehicle Drive Motor Based on Graph Neural Networks (GNN) with Attention Mechanisms

The rapid development of new energy vehicles has made the drive motor a core component whose reliability directly affects safety and performance. Traditional fault diagnosis methods often struggle to handle the complex nonlinear relationships and high-dimensional signals generated by drive motors. These challenges lead to difficulties in early fault detection, misclassification, and reduced accuracy under variable operating conditions. To overcome these problems, this research introduces a hybrid approach combining Graph Neural Networks (GNN) with Attention mechanisms. The GNN effectively captures spatial and structural relationships between motor signals, while Attention selectively emphasizes the most informative features for accurate classification. The dataset used consists of vibration, current, and temperature signals collected from drive motor systems under different fault types and operating loads. Preprocessing techniques such as normalization, noise filtering, and feature extraction are applied to ensure clean inputs for training. The proposed GNNAttention model maps raw sensor data into graph structures, applies message passing, and assigns dynamic weights to highlight critical fault indicators. Experimental results shows that the method achieves higher accuracy of 97.72% and robustness compared to traditional neural networks. This demonstrates its effectiveness in improving intelligent fault diagnosis for next-generation new energy vehicle drive motors.