基于决策树模型的轨道服役状态预警研究

Study on Early Warning of Track Service Conditions Based on Decision Tree Model

随着高速铁路服役时间的增加,各类轨道病害不断发生,尤其是高温天气引起无砟轨道结构服役状态持续劣化,出现严重结构伤损问题。以工程应用为导向,建立可靠的轨道结构服役状态预警模型,提出判别服役状态的预警方法,是实现铁路运输安全生产的重要保障措施之一。本文通过在某线路搭建服役环境和轨道板温度在线监测系统,将温度梯度出现概率等于0.3%时的取值作为轨道板温度梯度预警限值,基于人工智能算法决策树模型进行轨道结构服役状态预警研究。研究结果表明:(1)采用决策树模型可以有效预测轨道结构服役状态,实现轨道板温度梯度的质量等级划分;(2)轨道服役状态预警结果准确性与样本数量密切相关,丰富监测样本数据库,将会更全面精准地预测轨道结构异常状态,以便保障轨道养护维修的及时性。

With the increasing service time of high-speed railways,various track defects continuously emerge,particularly severe structural damage caused by elevated temperatures,which would affect the service conditions of ballastless track.Establishing a reliable early warning model for track service conditions and developing a method for identifying such conditions are among the essential measures for ensuring the operation safety of railway transport.For the purpose of this study,an online monitoring system was set up along a railway for on-line monitoring of the track environment and slab temperature,and the value at the probability of 0.3%for temperature gradient occurrence was taken as the threshold for slab temperature gradient warnings.Then an AI algorithm decision tree was employed to construct the early warning model for track structure service conditions.The findings indicate that:(1)The decision tree model can effectively predict track structure service conditions,enabling the determination of slab temperature gradient quality levels.(2)The accuracy of early warning for track service conditions is strongly correlated with the quantity of samples in the database,and enhancing the richness of monitoring sample database will enable a more comprehensive and precise prediction of abnormal track structure conditions,ensuring timely track maintenance and repair.