Related Experiment Video
Updated: Nov 29, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.5K
Remaining Useful Life Prognosis for Turbofan Engine Using Explainable Deep Neural Networks with Dimensionality
Chang Woo Hong1, Changmin Lee1, Kwangsuk Lee1
1School of Electrical & Electronic Engineering, Yonsei University, 50 Yonsei-Ro Seodamun-Gu, Seoul 03722, Korea.
Sensors (Basel, Switzerland)
|November 24, 2020
Summary
This study enhances turbofan engine health management by accurately predicting remaining useful life using deep learning. Techniques like dimensionality reduction and SHAP address data complexity and model interpretability for improved system prognosis.
Area of Science:
- Aerospace Engineering
- Artificial Intelligence
- Mechanical Engineering
Background:
- Effective health management of turbofan engines is critical for operational safety and efficiency.
- Predicting remaining useful life (RUL) is a key aspect of engine prognostics.
- Deep learning models offer potential for accurate RUL prediction but face challenges like high dimensionality and lack of interpretability.
Purpose of the Study:
- To develop and validate a deep learning model for accurate turbofan engine RUL prognosis.
- To address the challenges of high dimensionality (curse of dimensionality) and the black-box nature of deep neural networks in engine prognosis.
- To enhance the interpretability of deep learning models for identifying critical engine components.
Main Methods:
- A deep learning architecture combining one-dimensional convolutional neural networks (1D-CNN), long short-term memory (LSTM), and bidirectional LSTM was employed.
- Dimensionality reduction techniques were applied to manage the curse of dimensionality and prevent overfitting.
- Shapley Additive Explanation (SHAP) was utilized to analyze and visualize the deep learning model's predictions, identifying key sensor contributions.
Main Results:
- The proposed deep learning model, incorporating dimensionality reduction, demonstrated high accuracy and efficiency in RUL prognosis.
- Dimensionality reduction effectively reduced model complexity and prevented overfitting while preserving predictive accuracy.
- SHAP analysis successfully enhanced the explainability of the deep learning model, enabling the identification of problematic components and critical sensors.
Conclusions:
- The integrated deep learning approach with dimensionality reduction and SHAP provides a robust and interpretable solution for turbofan engine RUL prognosis.
- This methodology significantly improves upon conventional deep learning models by addressing practical implementation challenges.
- The study highlights the importance of explainability techniques in prognostics for enhanced system health management and decision-making.