Scholarlink

The track geometry irregularity has a crucial effect on in dynamic characteristics of vehicle-track coupling system, and its long-term degradation prediction is of important value to railway maintenance. In this paper, a data-driven prediction method is proposed, aiming to infer the future track irregularity for several months based on historical measurements. According to the characteristics of the track irregularity, an end-to-end model based on convolutional long short term memory (ConvLSTM) was established. An advanced spatiotemporal loss function is introduced to accurately evaluate the error between predicted and actual results. The track irregularity measurements set of a heavy haul railway segment in China for 9 consecutive months were prepared to participate in the training and testing of the network. The results show that the proposed method demonstrates elegant prediction performance. This work offers a promising path to the evolution prediction of track irregularity, and provides a methodological support for reducing the risk and maintenance cost of railway transportation.

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

Degradation Prediction of Track Geometry Irregularity from Historical Measurements Based on Deep Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Qinglai Zhang,
  • Shengyang Zhu,
  • Jianmng Gao,
  • Wanming Zhai

摘要

The track geometry irregularity has a crucial effect on in dynamic characteristics of vehicle-track coupling system, and its long-term degradation prediction is of important value to railway maintenance. In this paper, a data-driven prediction method is proposed, aiming to infer the future track irregularity for several months based on historical measurements. According to the characteristics of the track irregularity, an end-to-end model based on convolutional long short term memory (ConvLSTM) was established. An advanced spatiotemporal loss function is introduced to accurately evaluate the error between predicted and actual results. The track irregularity measurements set of a heavy haul railway segment in China for 9 consecutive months were prepared to participate in the training and testing of the network. The results show that the proposed method demonstrates elegant prediction performance. This work offers a promising path to the evolution prediction of track irregularity, and provides a methodological support for reducing the risk and maintenance cost of railway transportation.