基于Logistic回归的无砟轨道层间位移预警研究

无砟轨道层间位移是运营期间荷载作用下轨道板与砂浆层产生的离缝宽度,也是影响行车安全与养护维修的关键参数。针对层间位移状态的预警问题,以华东地区某线路无砟轨道为研究对象,基于现场实测数据,以环境温度、太阳辐射、风速、日温差、前4小时太阳辐射量均值、前6小时环境温度均值等气象参数为输入,无砟轨道层间位移值为输出,建立基于Logistic回归的无砟轨道层间位移分类预警模型,利用实测数据进行模型验证并与传统的BP神经网络模型和决策数模型作对比。研究结果表明:无砟轨道层间位移预警模型的准确率为95.21%,预测结果优于BP神经网络94.33%和决策数模型95.07%,为无砟轨道结构的病害预警与养护维修提供指导和建议。

Research on early warning of interlayer displacement of ballastless track based on Logistic regression

The ballastless track interlayer displacement is the width of the gap between the track slab and the mortar layer under load during operation,and is also a key parameter that affects driving safety and maintenance.Aiming at the early-warning problem of the interlayer displacement status,the ballastless track of a line in East China was taken as the research object.Based on the field measurement data,the environmental temperature,solar radiation,wind speed,daily temperature difference,mean solar radiation amount in the first 4 hours,and environment in the first 6 hours,meteorological parameters such as temperature mean value are input,and the ballastless track interlayer displacement value was output.A logistic regression-based early warning model for ballastless track interlayer displacement classification was established.The model was verified by using measured data and compared with the traditional BP neural network model and decision number.The research results are as follows.The accuracy rate of the ballastless track interlayer displacement early warning model is 95.21%.The prediction result is better than the BP neural network 94.33%and the decision number model 95.07%.the results can provide guidance and suggestions for the disease warning and maintenance of the ballastless track structure.