<![CDATA[Accurate Remaining Useful Life (RUL) prediction is critical for the predictive maintenance and operational safety of aircraft turbofan engines. This research develops and evaluates a stacked Long Short-Term Memory (LSTM) network for RUL prediction using the NASA C-MAPSS FD001 dataset as a fundamental case study. A systematic data preprocessing pipeline was employed, including sensor selection, RUL value clipping at 130 cycles, and feature normalization to prepare the data for modeling. The LSTM model was trained with regularization techniques and an EarlyStopping callback to ensure robustness and prevent overfitting. Evaluation results on the unseen test data show the final model achieved a solid and competitive performance with a Root Mean Squared Error (RMSE) of 15.22 and a PHM08 Score of 311.20. These results demonstrate that a well-configured LSTM architecture provides a reliable baseline for engine prognostic tasks, exhibiting strong generalization capabilities on new data.]]>
Copyrights © 2025
