Fault Prediction of Landing Gear Actuator Based on Attention-GRU Model

The landing gear system operates in a complex environment with heavy impact loads, leading to a high failure rate and extensive maintenance tasks. This paper focuses on a specific civilian aircraft's landing gear retractable system. A hydraulic simulation model is built to simulate normal operations, and by adjusting model parameters, two typical faults caused by component degradation are simulated, providing data support for subsequent fault prediction. This paper proposes a fault prediction method for typical failure modes of landing gear retractable actuators using the Attention mechanism and Gated Recurrent Unit (GRU). By assigning corresponding weights to different features of fault data through the Attention mechanism, the GRU model adequately learns crucial data and reduces the loss of long sequence data. This enhances fault prediction accuracy compared to GRU, Support Vector Regression (SVR), Long Short-term Memory (LSTM), Recurrent Neural Network (RNN), and Attention-LSTM methods. Mean Absolute Error and Mean Squared Error are chosen as evaluation metrics. Experimental results demonstrate the higher prediction accuracy of the proposed method.