Related Experiment Videos
Remaining Life Prediction of Shielding Sleeves Based on Data Augmentation and Hybrid Models
Xin Zhang1, Xuewei Xiang1, Hui Li1
1State Key Laboratory of Power Transmission Equipment Technology, School of Electrical Engineering, Chongqing University, Chongqing 400044, China.
Sensors (Basel, Switzerland)
|June 12, 2026
Summary
Predicting the remaining useful life (RUL) of shielding sleeves is crucial for nuclear safety. This study introduces a novel method using data augmentation and a hybrid model to overcome data limitations and improve RUL prediction accuracy.
Area of Science:
- Nuclear Engineering
- Materials Science
- Predictive Maintenance
Background:
- Remaining useful life (RUL) prediction for shielding sleeves is vital for preventing nuclear safety risks and catastrophic failures.
- Limited full-life-cycle degradation data hinders the development of accurate data-driven predictive models due to structural and service condition complexities.
- Typical failure modes include bulging and wear, necessitating robust prediction methods.
Purpose of the Study:
- To develop an effective remaining useful life (RUL) prediction method for shielding sleeves.
- To address the challenge of insufficient full-life-cycle degradation data.
- To enhance the accuracy and engineering applicability of RUL prediction models.
Main Methods:
- Proposed a degradation data augmentation method using Monte Carlo simulation based on analytical models of typical shielding sleeve failures (bulging and wear).
- Developed a hybrid RUL prediction model employing Stacking ensemble learning, integrating physical degradation models with deep learning techniques.
- Validated the method using multiple degradation datasets covering different failure modes.
Main Results:
- Achieved a minimum root-mean-square error (RMSE) of 0.0058 and a minimum mean absolute error (MAE) of 0.0044.
- The proposed hybrid model demonstrated superior prediction accuracy compared to single predictive models.
- Experimental verification confirmed the method's performance and engineering applicability.
Conclusions:
- The developed data augmentation and hybrid modeling approach effectively addresses the scarcity of full-life-cycle degradation data for shielding sleeves.
- The Stacking ensemble learning model integrates physical insights with data-driven approaches for enhanced RUL prediction.
- The method offers a reliable and accurate solution for predicting the remaining useful life of shielding sleeves, contributing to nuclear safety.