Scholarlink

Planetary gearboxes play a vital role in various industries due to their compact size, high efficiency, and reliability, but are prone to failures caused by factors such as changing loads and harsh operating environments. Traditional fault diagnosis techniques are typically data-intensive, demanding substantial amounts of annotated data, which can be a significant hurdle in scenarios where such data are sparse. To overcome this challenging, this research proposes a few-shot learning model, specifically dynamic graph attention network (DGAT), which effectively leverages limited labeled data. By dynamically adjusting focus based on contextual relationships among data points, the dynamic attention mechanism within DGAT significantly enhances the model's precision in distinguishing different fault categories. This approach not only improves the diagnostic process using small amounts of labeled data, thereby ensuring higher equipment reliability and safety while minimizing data requirements. The efficiency and effectiveness of the proposed DGAT-based approach are validated through experimental testing on a planetary gearbox setup.

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

Few-Shot Graph Neural Networks Framework Incorporating DGAT for Planetary Gearbox Diagnosis

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Jia Gao,
  • Peng Chen,
  • Yaqiang Jin,
  • Chaojun Xu

摘要

Planetary gearboxes play a vital role in various industries due to their compact size, high efficiency, and reliability, but are prone to failures caused by factors such as changing loads and harsh operating environments. Traditional fault diagnosis techniques are typically data-intensive, demanding substantial amounts of annotated data, which can be a significant hurdle in scenarios where such data are sparse. To overcome this challenging, this research proposes a few-shot learning model, specifically dynamic graph attention network (DGAT), which effectively leverages limited labeled data. By dynamically adjusting focus based on contextual relationships among data points, the dynamic attention mechanism within DGAT significantly enhances the model's precision in distinguishing different fault categories. This approach not only improves the diagnostic process using small amounts of labeled data, thereby ensuring higher equipment reliability and safety while minimizing data requirements. The efficiency and effectiveness of the proposed DGAT-based approach are validated through experimental testing on a planetary gearbox setup.