Related Experiment Video
Updated: Jun 10, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
975
A method for predicting remaining useful life using enhanced Savitzky-Golay filter and improved deep learning
Xiangyang Li1, Lijun Wang2, Chengguang Wang1
1School of Management and Economics, North China University of Water Resources and Electric Power, Zhengzhou, 450045, China.
Scientific Reports
|October 14, 2024
Summary
This study introduces a deep learning approach for predicting equipment health and remaining useful life (RUL). The novel framework enhances accuracy in fault prediction, outperforming traditional methods.
Area of Science:
- Engineering
- Computer Science
- Data Science
Background:
- Operational integrity of large-scale equipment relies on effective fault prediction and health management.
- Prognostics and Health Management (PHM) struggles with accurate Remaining Useful Life (RUL) prediction from multivariate sensor data.
- Traditional PHM methods often require extensive prior knowledge for feature engineering.
Purpose of the Study:
- To present a novel multi-channel, multi-scale deep learning approach for enhanced fault prediction and RUL estimation.
- To address limitations of traditional methods by leveraging deep learning for complex operational datasets.
- To improve the accuracy and robustness of PHM systems.
Main Methods:
- An improved Savitzky–Golay filter (ISG) was employed for efficient data preprocessing of large, dynamic sensor volumes.
- A hybrid deep learning framework integrating Convolutional Neural Networks (CNNs) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal dependency modeling was developed.
- Fusion of CNN and LSTM outputs was utilized to enhance integrated prediction capabilities.
Main Results:
- Experimental validation on the C-MAPSS dataset demonstrated the framework's promising performance, especially under dynamic operational conditions.
- Comparative analyses confirmed the superiority of the proposed deep learning approach over classical algorithms for single fault type prediction.
- The study identified optimal parameters and evaluated filtering effectiveness through various fusion methods and CNN depths.
Conclusions:
- The developed multi-channel, multi-scale deep learning framework offers a robust and accurate solution for equipment fault prediction and RUL estimation.
- While not optimized for multi-fault prediction, the approach significantly outperforms traditional methods in single fault scenarios.
- This deep learning strategy advances Prognostics and Health Management (PHM) capabilities for industrial applications.
Keywords:
Deep learningNeural networkPrognostics and health management (PHM)Remaining useful life (RUL) predictions
