Friction Evolution Prediction and Wear Stage Diagnosis of Reciprocating Seals in Aircraft Hydraulic Systems

In aircraft hydraulic systems, the friction behavior of reciprocating seals evolves progressively under long-term high-pressure operation, making wear-stage monitoring difficult. To address this issue, this study proposes a multi-sensor data-fusion framework for friction prediction and wear-stage diagnosis. Friction force, oil film thickness, contact stress, and temperature were synchronously acquired using load, ultrasonic, and fiber-optic sensors. A health index was constructed to characterize seal degradation, a difference-learning long short-term memory (LSTM) network was developed to predict friction trends, and the predicted trend features were further incorporated into an extreme gradient boosting (XGBoost) classifier for online wear-stage diagnosis. Experiments were conducted under nine operating conditions formed by three seal configurations and three pressure levels. The LSTM model achieved an average relative error below 2% in friction prediction, while the XGBoost classifier with friction-trend features reached an overall diagnostic accuracy of 96.1% for mild, moderate, and severe wear stages. These results indicate that the proposed framework can effectively track friction evolution and support wear-stage assessment of reciprocating seals.