Scholarlink

Health state prediction and fault time prediction are pivotal in fault prognostics. Traditional approaches often treat these tasks separately, overlooking the time-varying dynamics of system operations, which compromises efficiency and accuracy. To address this, we propose a Gaussian Process Regression (GPR) method with an adaptive update strategy tailored for aeroengine prognostics. This approach continuously refines model parameters and predictive distributions using new data. By integrating variable selection and multivariable fusion, it identifies key health-related variables to construct a comprehensive health index, enabling simultaneous health state and fault time prediction. The method employs single-point and multipoint update strategies within a unified GPR framework, effectively incorporating new sample data for dual-task predictions. Experimental results on the C-MAPSS dataset demonstrate that the proposed method outperforms state-of-the-art techniques.

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

Update-Strategy-Based Gaussian Processes Regression for Processes Fault Prediction with Incremental Data

  • 章节

  • 发表日期：

    2025年

查看全文
  • Hongpeng Yin,
  • Han Zhou,
  • Yi Chai,
  • Qiu Tang

摘要

Health state prediction and fault time prediction are pivotal in fault prognostics. Traditional approaches often treat these tasks separately, overlooking the time-varying dynamics of system operations, which compromises efficiency and accuracy. To address this, we propose a Gaussian Process Regression (GPR) method with an adaptive update strategy tailored for aeroengine prognostics. This approach continuously refines model parameters and predictive distributions using new data. By integrating variable selection and multivariable fusion, it identifies key health-related variables to construct a comprehensive health index, enabling simultaneous health state and fault time prediction. The method employs single-point and multipoint update strategies within a unified GPR framework, effectively incorporating new sample data for dual-task predictions. Experimental results on the C-MAPSS dataset demonstrate that the proposed method outperforms state-of-the-art techniques.