Scholarlink

<p>This paper presents the development of a&#xa0;prototype system for in-situ wear monitoring of helical gears using vision-based condition monitoring (VCM). The research addresses the challenge of monitoring gearbox wear, particularly pitting, an inherent failure mechanism in high-power applications such as wind turbines. A&#xa0;VCM system was integrated into an FZG gear test rig to automatically capture images of gear surfaces during operation, using an oil splash mitigation strategy to ensure high image quality. These images were processed using an image processing pipeline, incorporating a&#xa0;deep learning Feature Pyramid Network for automated annotation of wear defects, trained based on an iterative approach to reduce annotation effort during the training phase. Hereafter, this paper explores metrics related to pitting evolution, including pit area, height, width, and location on the gear surface, as well as insights these metrics give on pitting evolution. This framework’s ability to track the spatiotemporal development of gear wear provides valuable insights for condition monitoring and predictive maintenance in industrial applications</p>

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

A vision-assisted condition monitoring framework for wear monitoring of gearboxes

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年5月20日

查看全文
  • Djordy Van Maele,
  • Jean Carlos Poletto,
  • Roeland De Geest,
  • Wenzhi Liao,
  • Ney Francisco Ferreira,
  • Patric Daniel Neis,
  • Dieter Fauconnier,
  • Patrick De Baets

摘要

This paper presents the development of a prototype system for in-situ wear monitoring of helical gears using vision-based condition monitoring (VCM). The research addresses the challenge of monitoring gearbox wear, particularly pitting, an inherent failure mechanism in high-power applications such as wind turbines. A VCM system was integrated into an FZG gear test rig to automatically capture images of gear surfaces during operation, using an oil splash mitigation strategy to ensure high image quality. These images were processed using an image processing pipeline, incorporating a deep learning Feature Pyramid Network for automated annotation of wear defects, trained based on an iterative approach to reduce annotation effort during the training phase. Hereafter, this paper explores metrics related to pitting evolution, including pit area, height, width, and location on the gear surface, as well as insights these metrics give on pitting evolution. This framework’s ability to track the spatiotemporal development of gear wear provides valuable insights for condition monitoring and predictive maintenance in industrial applications