Scholarlink

<p>Railroad transportation is an integral part of the transportation sector, especially high-speed railroad, which is a key link in the railroad system. In particular, high-speed railways serve as a crucial component of the railway system, and significant achievements have been made in their construction and operation in recent years. However, as the speed of high-speed trains continues to increase, the use of wheel-rail adhesion is also being challenged. Therefore, to achieve stable tracking of the optimal adhesion state while adopting the feedback dynamic adhesion characteristic model, this study utilizes a machine learning algorithm to perform classification training on discrete points of the simulated adhesion characteristic model under various rail conditions. Subsequently, the trained model is integrated into the vehicle simulation model for real-time identification of current rail surface conditions, and the output value from this module is fed into the subsequent optimal adhesion search algorithm. By dynamically adjusting the initial search step size and its change coefficient, in conjunction with torque controller action, precise operation of the vehicle at the optimal adhesion point can be achieved. Finally, validation is conducted using a semi-physical simulation platform based on ModelinTech real-time simulator.</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

Rail surface identification and adhesion control based on dynamic adhesion characteristics

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月18日

查看全文
  • Wen Liu,
  • Hongfeng Qi,
  • Jingchun Huang,
  • Yiyuan Chen,
  • Sheng He,
  • Haoxiang Feng

摘要

Railroad transportation is an integral part of the transportation sector, especially high-speed railroad, which is a key link in the railroad system. In particular, high-speed railways serve as a crucial component of the railway system, and significant achievements have been made in their construction and operation in recent years. However, as the speed of high-speed trains continues to increase, the use of wheel-rail adhesion is also being challenged. Therefore, to achieve stable tracking of the optimal adhesion state while adopting the feedback dynamic adhesion characteristic model, this study utilizes a machine learning algorithm to perform classification training on discrete points of the simulated adhesion characteristic model under various rail conditions. Subsequently, the trained model is integrated into the vehicle simulation model for real-time identification of current rail surface conditions, and the output value from this module is fed into the subsequent optimal adhesion search algorithm. By dynamically adjusting the initial search step size and its change coefficient, in conjunction with torque controller action, precise operation of the vehicle at the optimal adhesion point can be achieved. Finally, validation is conducted using a semi-physical simulation platform based on ModelinTech real-time simulator.