Deep learning and similarity-based models for predicting turbofan engine remaining useful life: insights from the CMAPSS dataset DOI
F. Isbilen, Oğuz Bektaş, Mehmet Konar

и другие.

The Aeronautical Journal, Год журнала: 2025, Номер unknown, С. 1 - 32

Опубликована: Апрель 14, 2025

Abstract Turbofan engines are having a growing role in modern aircraft maintenance. Due to this increase, estimation of remaining useful life (RUL) these is an important area study the field reliability and maintenance optimisation. In work, we propose hybrid approach that combines deep learning models with similarity-based methods for accurate RUL estimation. For better comparison, evaluate four architectures: dropout long short-term memory (LSTM), bidirectional LSTM, convolutional neural network 1D (CNN 1D), multi-layer LSTM. The FD002 subset NASA’s Commercial Modular Aero-Propulsion System Simulation dataset used case study. Root mean square error (RMSE) absolute (MAE) were performance metrics. main output suggests LSTM model achieves best prediction accuracy RMSE score 26.547 MAE 18.749. It worth noting achieved despite requiring higher computational resources compared Furthermore, all had difficulties smaller test trajectory lengths such as 50–100 due training data imbalance. Overall, findings highlight promise approaches prediction. However, potential advancements hyperparameter optimisation augmentation still hold further improvements.

Язык: Английский

Deep learning and similarity-based models for predicting turbofan engine remaining useful life: insights from the CMAPSS dataset DOI
F. Isbilen, Oğuz Bektaş, Mehmet Konar

и другие.

The Aeronautical Journal, Год журнала: 2025, Номер unknown, С. 1 - 32

Опубликована: Апрель 14, 2025

Abstract Turbofan engines are having a growing role in modern aircraft maintenance. Due to this increase, estimation of remaining useful life (RUL) these is an important area study the field reliability and maintenance optimisation. In work, we propose hybrid approach that combines deep learning models with similarity-based methods for accurate RUL estimation. For better comparison, evaluate four architectures: dropout long short-term memory (LSTM), bidirectional LSTM, convolutional neural network 1D (CNN 1D), multi-layer LSTM. The FD002 subset NASA’s Commercial Modular Aero-Propulsion System Simulation dataset used case study. Root mean square error (RMSE) absolute (MAE) were performance metrics. main output suggests LSTM model achieves best prediction accuracy RMSE score 26.547 MAE 18.749. It worth noting achieved despite requiring higher computational resources compared Furthermore, all had difficulties smaller test trajectory lengths such as 50–100 due training data imbalance. Overall, findings highlight promise approaches prediction. However, potential advancements hyperparameter optimisation augmentation still hold further improvements.

Язык: Английский

Процитировано

0