Prediction of remaining useful life of aero-engines based on CNN-LSTM- Attention

Abstract Predicting an aircraft engine's remaining life with accuracy is crucial for maintaining both financial stability and aviation safety. The capacity to analyze spatiotemporal data enhances machine RUL's predictive performance. Nevertheless, the majority of spatiotemporal information processing models now in use lack the ability to extract adaptive features in addition to having complicated topologies. In light of this, the paper suggests a deep learning-based approach to forecast an engine's remaining life. This method has the potential to enhance data feature recognition and, as a result, boost model prediction accuracy. First, we standardize the input features and compute the aircraft engine dataset test set's Remaining Useful Life (RUL) using CMAPSS. After extracting characteristics from the input data using a CNN (Convolutional Neural Network), the extracted data is fed into an LSTM model (Long Short Term Memory) and attention (multi head attention) is added to forecast how long the aircraft engine would last. Finally, ablation study and comparison model experiments were used to evaluate and compare the suggested aircraft engine model. the research results show that the CNN-LSTM-Attention model performs well predictively on the dataset and its Root Mean Square Error (RMSE) of 15.306, 12.913, 12.507, and 13.902 in FD001, FD002, FD003, and FD004, respectively,. In comparison to other models, our proposed model has the best prediction accuracy on the CMPASS dataset, indicating strong reliability and accuracy.