摘要
为了提高钢轨胶接绝缘接头绝缘缝预测的精度,提出基于自回归移动平均(ARMA)模型的Grubbs-ARMA预测模型。该模型在ARMA模型基础上增加改进的预处理算法,保留原始数据的每一次波动规律,具体包括使用格拉布斯准则法对每个目标时刻附近的监测数据集合识别可疑值、剔除异常值和求均值,然后对得到的序列做平稳性分析,根据分析结果确定模型阶数,建立模型,利用京沈高铁沈阳段的闭塞区段某监测点采集的20组数据,分别构造Grubbs-ARMA模型和ARMA模型,根据前15组数据预测后5组数据。结果显示:Grubbs-ARMA预测模型的残差不仅均低于0. 15,且均低于ARMA预测模型的残差对应组别的残差。证实Grubbs-ARMA预测模型不仅适用于钢轨绝缘缝预测,且比ARMA预测模型有更高的预测精度。
To improve the prediction accuracy of the insulation cracks of rail glued insulated joints, aGrubbs-ARMA prediction model based on auto regressive moving average （ARMA） model is proposed.The model adds an improved preprocessing algorithm to the ARMA model, preserving every fluctuation ofthe original data, including the use of the Grubbs＇ criteria to identify suspicious values, eliminatingoutliers and finding the average of the monitoring data set near each target time. Stability analysis of theobtained sequence is conducted to determine the order of the model and then establish the model.Grubbs-ARMA model and ARMA model are constructed by 20 sets of data collected at a monitoring pointin the blocked section near Shenyang on Beijing-Shenyang high speed railway. The first 15 sets of dataare used to predict the last 5 sets of data. The results show that the residuals of the Grubbs-ARMAprediction model are not only below 0. 15 but also significantly below the corresponding residuals of theARMA prediction model. It is verified that the Grubbs-ARMA prediction model is not only suitable forthe prediction of rail insulation cracks, but also has higher prediction accuracy than the ARMA predictionmodel.
作者
李存荣
王博文
LI Cun-rong;WANG Bo-wen(School of Mechanical and Electronic Engineering,Wuhan University of Technology,Wuhan 430070,China)
出处
《铁道标准设计》
北大核心
2018年第11期49-53,共5页
Railway Standard Design
基金
国家自然科学基金项目(71171154)
湖北省科技支撑计划项目(2015BAA063)
关键词
胶接绝缘接头
高速铁路
时序分析
绝缘缝
Glued insulated joint
High-speed railway
Time series analysis
Insulation crack
