基于轨道不平顺的机器学习方法建模和预测
Modeling and Prediction of Machine Learning Method Based on Track Irregularity

DOI: 10.12677/CSA.2021.1110247, PP. 2417-2427

Keywords: 轨道不平顺，机器学习，Prophet模型，神经网络，时间序列预测
Track Irregularity, Machine Learning, Prophet Model, Neural Network, Time Series Prediction

Full-Text   Cite this paper   Add to My Lib

Abstract:

针对日常运营中火车轨道在列车荷载冲击作用下导致的轨道不平顺问题，使用机器学习方法Prophet算法和基于卷积神经网络的时间卷积网络，对轨道质量指数(TQI)数据进行建模分析和预测。对北京–上海某区段的TQI数据进行分析，并与传统模型ARIMA和三次指数平滑模型进行比较，发现其精确度更高，拟合效果更好。说明了所使用方法处理轨道不平顺数据的有效性。
According to the problem of the track irregularity caused by the impact of train load in daily operation, the machine learning method Prophet algorithm and the temporal convolutional network based on convolutional neural network are used to analyze and predict the track quality index (TQI) data. Analyzing the TQI data of a certain section from Beijing to Shanghai and comparing it with the traditional model ARIMA and Exponential Smoothing models, it is found that the accuracy of Prophet and TCN model is higher and the fitting effect is better. It illustrates the effectiveness of the method that we used to deal with track irregularity data.

References

[1]	常惠, 饶志强, 赵玉林. 轨道几何不平顺的预测研究进展[C]//中国计算机用户协会网络应用分会2019年第二十三届网络新技术与应用年会, 2019.
[2]	许玉德, 李海峰, 周宇. 铁路轨道高低不平顺的预测方法[J]. 同济大学学报, 自然科学版, 2003, 31(3): 291-295.
[3]	Kawaguchi, A. and Miwa, M. (2005) Actual Data Analysis of Align-ment Irregularity Growth and Its Prediction Model. Quarterly Report of RTRI, 46, 262-268. https://doi.org/10.2219/rtriqr.46.262
[4]	Lee, J.S., Hwang, S.H., Choi, I.Y., et al. (2018) Prediction of Track De-terioration Using Maintenance Data and Machine Learning Schemes. Journal of Transportation Engineering Part A: Systems, 144, Article ID: 04018045. https://doi.org/10.1061/JTEPBS.0000173
[5]	常燕龙. ARMA-BP组合模型在某高速铁路轨道不平顺预测中的应用[J]. 科学技术创新, 2021(12): 21-22.
[6]	李志国, 钟将, 钟璐蔓. 复杂事件管理的多元时序数据处理技术研究[J]. 计算机科学, 2019, 46(6): 61-69.
[7]	Gan, Z., Li, C., Zhou, J., et al. (2021) Temporal Convolutional Networks Interval Prediction Model for Wind Speed Forecasting. Electric Power Systems Research, 191, 106865. https://doi.org/10.1016/j.epsr.2020.106865
[8]	孔震, 张华鲁, 岳圣凯, 袁明磊, 路通. 基于时域卷积网络的多尺度双线性天气预测模型[J]. 图学学报, 2020, 41(5): 80-86.
[9]	郭继昌, 吴洁, 郭春乐, 等. 基于残差连接卷积神经网络的图像超分辨率重构[J]. 吉林大学学报(工学版), 2019, 49(5): 1726-1734.
[10]	沈时宇, 陈明. Prophet时序模型在短期水质溶氧预测中的应用[J]. 渔业现代化, 2020(3): 29-35.
[11]	赖慧慧. 基于时间序列Prophet模型的乘用车消费税预测[J]. 税收经济研究, 2020(1): 34-39.
[12]	罗微. 高低和TQI的轨道不平顺预测模型研究[D]: [硕士学位论文]. 成都: 西南交通大学, 2013.
[13]	魏世斌, 杨凤春, 翁绍德. 轨道质量指数的研究和应用[J]. 中国铁道科学, 1996(2): 23-33.
[14]	Hochenbaum, J., Vallis, O.S. and Kejariwal, A. (2017) Automatic Anomaly De-tection in the Cloud via Statistical Learning. arXiv:1704.07706 [cs.LG]

Full-Text