A Remaining Useful Life Prognosis of Turbofan Engine Using Temporal and Spatial Feature Fusion

Area of Science:

• Aerospace Engineering
• Mechanical Engineering
• Artificial Intelligence

Background:

• Predictive maintenance and remanufacturing of turbofan engines rely on accurate Remaining Useful Life (RUL) prognosis.
• Traditional shallow machine learning methods often struggle with single feature extraction, leading to suboptimal RUL prediction accuracy.
• High failure rates and maintenance costs necessitate improved RUL prediction strategies.

Purpose of the Study:

• To propose a novel hybrid deep learning model for enhanced RUL prognosis in turbofan engines.
• To address the limitations of single feature extraction in traditional RUL prediction methods.
• To improve the accuracy and reliability of RUL predictions for effective maintenance scheduling.

Main Methods:

• A hybrid model integrating one-dimensional Convolutional Neural Networks with Full Convolutional Layers (1-FCLCNN) and Long Short-Term Memory (LSTM) was developed.
• LSTM was employed to extract temporal features, while 1-FCLCNN extracted spatial features from turbofan engine datasets (FD001 and FD003).
• Feature fusion from both LSTM and 1-FCLCNN was used as input for subsequent Convolutional Neural Networks (CNN) to predict RUL.

Main Results:

• The proposed hybrid model demonstrated superior RUL prediction accuracy compared to existing popular models.
• Evaluation metrics confirmed the effectiveness and enhanced performance of the combined 1-FCLCNN and LSTM approach.
• The model successfully captured both temporal and spatial dependencies within the turbofan engine data for accurate prognosis.

Conclusions:

• The proposed 1-FCLCNN and LSTM hybrid model offers a significant advancement in turbofan engine RUL prediction.
• This approach overcomes the limitations of single feature extraction, providing more accurate and reliable prognostics.
• The findings highlight the model's superiority and effectiveness for practical applications in predictive maintenance and remanufacturing.