Fault diagnosis of electric transmission system based on graph-enhanced deep feature fusion network model using efficient decision mapping

Abstract As the core component of the electric drive system, the operating status of bearings directly affects the normal operation of the electric drive system, thereby indirectly affecting the safety and reliability of the vehicle. Due to the special working environment and complex structure of vehicle electric drive systems, traditional vibration-based fault diagnosis methods have many limitations in practical operation and maintenance, making it difficult to meet the needs of accurate diagnosis. This study takes the current signal of the driving motor as the research object and innovatively proposes a graph enhanced deep feature fusion network (GE-DFFN), aiming to accurately identify and effectively solve the problem of weak and difficult to capture defect features in the current signal. Firstly, the model combines graph attention network layer, deep feature enhancement layer, and singular decision mapping layer to enhance its ability to understand and identify complex fault features. In addition, the model adopts a graph attention node update strategy to improve training efficiency. Based on the effectiveness score and attention allocation weight of the decision mapping layer, the node update strategy can eliminate invalid features and enhance weak fault features. The feasibility and superiority of the proposed model method were comprehensively verified using current data from the rolling bearing experimental platform and actual measurement data from the electric drive system of electric vehicles. Through in-depth analysis and verification of these real and diverse data, the excellent performance and practical value of the proposed GE-DFFN in the field of electrical transmission system fault diagnosis have been fully demonstrated.