摘要详情

ID / 提交时间

168 / 2025-06-10 19:01:29

标题

Adaptive Reconstruction of Sensor Fault Data Based on CNN-LSTM Neural Network

关键字

data reconstruction,sensor fault,LSTM,CNN

主题及专题

4、System Monitoring, Fault Diagnosis and Prediction

状态

全文录用

作者

培鑫邢 / 北京信息科技大学

然贾 / 北京信息科技大学

涛陈 / 北京信息科技大学

傲吴 / 北京信息科技大学

摘要

Accurate condition monitoring is essential for ensuring efficient and reliable operation of integrated transmission systems, where sensor data reliability plays a critical role. To address this challenge, we propose a novel deep learning-based fault reconstruction model that exploits structural and temporal correlations among sensors. When sensor failures occur, our model automatically reconstructs faulty measurements while maintaining data accuracy, thereby enhancing system reliability and operational safety. The proposed framework also serves as a robust reference for subsequent fault diagnosis research.Methodologically, we first employ Exponentially Weighted Moving Average (EWMA) for data smoothing, followed by correlation analysis to identify highly interdependent sensor features. We then develop a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) architecture to achieve precise sensor data reconstruction. Extensive case studies on integrated transmission systems with multiple sensor faults validate the model's reliability, accuracy, and effectiveness. Results demonstrate that our approach provides a technically sound and efficient solution for enhancing sensor reliability in complex transmission systems.