Scholarlink

This paper proposes an innovative WR-IMT (Weight-restricted Improved MGU-TCN) network framework for addressing the problem of long sequence remaining useful life (RUL) prediction in aero engines. The framework integrates modules for subsequence segmentation, bidirectional asymmetric MGU (Minimal Gated Unit), and attention mechanisms to effectively enhance predictive accuracy. Experimental results demonstrate the superior predictive performance of WR-IMT on the C-MAPSS (Commercial Modular Aero-Propulsion System Simulation) platform. Additionally, this paper explores the methods to enhance the adaptability of network structures for complex datasets. By introducing residual learning and bidirectional asymmetric network architectures, the network's ability to capture data features is strengthened. The WR-IMT network framework presented in this paper achieves efficient prediction of the long sequence remaining useful life of aero engines by integrating various advanced technologies. This not only helps to enhance the performance monitoring and maintenance management efficiency of aero engines but also provides new theoretical and methodological support for the field of time series analysis.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

WR-IMT: A Time Series Predictive Model for Remaining Useful Life Prediction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Fei Xiao,
  • Haibo Xing,
  • Jun Huang,
  • Jianxun Li

摘要

This paper proposes an innovative WR-IMT (Weight-restricted Improved MGU-TCN) network framework for addressing the problem of long sequence remaining useful life (RUL) prediction in aero engines. The framework integrates modules for subsequence segmentation, bidirectional asymmetric MGU (Minimal Gated Unit), and attention mechanisms to effectively enhance predictive accuracy. Experimental results demonstrate the superior predictive performance of WR-IMT on the C-MAPSS (Commercial Modular Aero-Propulsion System Simulation) platform. Additionally, this paper explores the methods to enhance the adaptability of network structures for complex datasets. By introducing residual learning and bidirectional asymmetric network architectures, the network's ability to capture data features is strengthened. The WR-IMT network framework presented in this paper achieves efficient prediction of the long sequence remaining useful life of aero engines by integrating various advanced technologies. This not only helps to enhance the performance monitoring and maintenance management efficiency of aero engines but also provides new theoretical and methodological support for the field of time series analysis.