Scholarlink

Frequently occurred pantograph arc has become the main factor affecting the safe and reliable operation of electric locomotive. Accurate identification of pantoraph arc is of great significance for evaluating current collection quality and guiding line maintenance of pantograph-catenary system. A pantograph arc identification method based on three-dimensional (3-D) input signals and an improved ShuffleNet network was proposed. The input signals are arc current, arc light and arc sound, respectively. The Savitzky-Golay filter was used to filter the noise in the measured arc current, arc sound and arc light signals. A three-channel feature matrix was constructed and used as the input of the pantograph arc identification model. The effective channel attention mechanism was used to improve the ShuffleNet network, and a pantograph arc identification model was established. The identification accuracy of the model is 98.35%. Compared with only using the current signal, when using the three-dimensional input signals, the identification accuracy is improved by 8.87%.

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

Identification of Pantograph Arc via 3-D Input Signals

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zhiyong Wang,
  • Xinyu Mao,
  • Tiecheng Wu,
  • Cong Du,
  • Jian Zhang

摘要

Frequently occurred pantograph arc has become the main factor affecting the safe and reliable operation of electric locomotive. Accurate identification of pantoraph arc is of great significance for evaluating current collection quality and guiding line maintenance of pantograph-catenary system. A pantograph arc identification method based on three-dimensional (3-D) input signals and an improved ShuffleNet network was proposed. The input signals are arc current, arc light and arc sound, respectively. The Savitzky-Golay filter was used to filter the noise in the measured arc current, arc sound and arc light signals. A three-channel feature matrix was constructed and used as the input of the pantograph arc identification model. The effective channel attention mechanism was used to improve the ShuffleNet network, and a pantograph arc identification model was established. The identification accuracy of the model is 98.35%. Compared with only using the current signal, when using the three-dimensional input signals, the identification accuracy is improved by 8.87%.