Research on the Condition Assessment Method of Mechanical Seal Based on Fusing Multi‐Graph Neural Networks From Decision Layer

Mechanical seals play a crucial role in mechanical equipment by effectively preventing liquid or gas leakage, ensuring the normal operation of the equipment, and avoiding energy waste and environmental pollution. Especially in pumps, compressors, and other devices, mechanical seals ensure sealing performance while extending the equipment’s lifespan and improving work efficiency. Therefore, research on the condition assessment of mechanical seals is both necessary and important. In order to achieve high accuracy in the assessment model, a comprehensive evaluation model that fuses multilevel information is proposed. Firstly, several types of sensors are used to monitor the operational status of the mechanical seal comprehensively and accurately, capturing different signal features to provide richer multidimensional data. Secondly, multiple methods are used to process and convert the collected data into graph data, ensuring the diversity of the training data through different channels and graph construction techniques. Thirdly, in order to future improve the assessment performance, multi‐GNNs models are fused by using different combined methods. Finally, the effectiveness of the assessment method is validated by using the test data of mechanical seal.