Scholarlink

In this paper, a novel hybrid fault classification approach that integrates feature engineering and deep neural network training is proposed for micro-thrusters of drag-free spacecraft. Specifically, a feature selection method including Pearson correlation coefficient and F-score is employed on the twelve-channel residual signals collected from model-based fault injection simulations. Subsequently, continuous wavelet transform and channel fusion processing are performed on the selected single-channel residual signal to generate color time-frequency images. In view of this, these images are fed into the ResNet18 model that is constructed with appropriate parameters to complete the model training. Simulation results show that the proposed method achieves high accuracy fault classification of micro-thruster of drag-free spacecraft.

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

Micro-thruster Fault Classification of Drag-Free Spacecraft Using Deep Neural Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zhibo Liang,
  • Xiaodong Shao,
  • Yongxia Shi,
  • Qinglei Hu,
  • Yonghe Zhang,
  • Pengcheng Wang

摘要

In this paper, a novel hybrid fault classification approach that integrates feature engineering and deep neural network training is proposed for micro-thrusters of drag-free spacecraft. Specifically, a feature selection method including Pearson correlation coefficient and F-score is employed on the twelve-channel residual signals collected from model-based fault injection simulations. Subsequently, continuous wavelet transform and channel fusion processing are performed on the selected single-channel residual signal to generate color time-frequency images. In view of this, these images are fed into the ResNet18 model that is constructed with appropriate parameters to complete the model training. Simulation results show that the proposed method achieves high accuracy fault classification of micro-thruster of drag-free spacecraft.