基于LSTM预测误差的轨道交通弓网接触力异常识别算法
杨劲松邵奇刘金朝陶凯郭剑峰彭楠
Identification Algorithm of Rail Transit Pantograph-catenary Contact Force Abnormality Based on LSTM Prediction Error
YANG JinsongSHAO QiLIU JinzhaoTAO KaiGUO JianfengPENG Nan
-
作者信息:中国铁道科学研究院集团有限公司基础设施检测研究所，100081，北京
-
Affiliation:Infrastructure Inspection Research Institute, China Academy of Railway Sciences Group Co., Ltd., 100081, Beijing, China
-
关键词:
-
Key words:
-
DOI:10.16037/j.1007-869x.2024.08.013
-
中图分类号/CLCN:U225.1
-
栏目/Col:研究报告
摘要:
［目的］接触力是轨道交通弓网综合检测的重要内容，也是弓网系统性能的重要评价因素，但在检测过程中常受到外部环境的影响而出现异常检测数据。目前针对弓网接触力异常检测数据的剔除主要依赖于人工，影响数据分析效率，因此需深入研究弓网接触力异常识别算法。［方法］梳理了弓网接触力常见的异常形式，分析了不同异常形式下接触力检测数据的特征。提出了一种基于LSTM(长短期记忆网络)预测误差的弓网接触力异常识别算法，通过使用正常的接触力数据训练LSTM模型，使该模型能够对接触力变化趋势进行预测。为实现正常区段与异常点的精确划分，使用基于置信区间的异常检测数据识别方法。为降低长距离异常数据对LSTM模型预测效果的影响，提出了一种基于预测值置换的异常数据处理方式。通过高速综合检测列车测得的真实检测数据，分别对三种常见异常形式的弓网接触力识别效果进行验证。［结果及结论］提出的算法能够较好地实现对弓网接触力异常的识别。
Abstracts:
［Objective］ Contact force is a crucial aspect of comprehensive detection in the rail transit PC (pantograph-catenary) system and serves as an important evaluation factor of PC system performance. However, during the detection process, the external environment factors often lead to abnormal detection data. Currently, the elimination of abnormal PC contact force detection data mainly relies on manual methods, which affect data analysis efficiency. Therefore, there is a need to conduct in-depth research on an identification algorithm for PC contact force abnormality in PC system. ［Method］ The common abnormal forms of PC contact force are categorized, and the characteristics of contact force detection data under different abnormal conditions are analyzed. An identification algorithm for PC contact force abnormality based on LSTM (long short-term memory) prediction error is proposed. Using normal contact force data to train an LSTM model enables it to predict the trend of contact force variations. To achieve precise differentiation between normal segments and abnormal points, an abnormal data detection method based on the confidence interval is used. To mitigate the impact of long-distance abnormal data on LSTM model prediction performance, a prediction value replacement method for handling abnormal data is proposed. The effectiveness of identifying three common abnormal forms of PC contact force is verified using real detection data obtained from high-speed comprehensive inspection trains. ［Result & Conclusion］ Research results show that the proposed algorithm can effectively identify abnormalities in the PC contact force.
- 上一篇: 高铁物流运输模式及其可行性
- 下一篇: 碳纤维地铁列车的商用将开启轨道车辆用材的新时代
过刊浏览
资料下载
友情链接