摘要
为降低铁路空车调配的扣车率,在详细分析空重车流组织过程的基础上,提出在编组站对空车调配进行源头分类的对策。首先通过货运部门扣车检查项点筛选出厂修时长等评估指标,创建基于扣车标准的评价指标体系,其次利用Pearson相关性系数分析了评估指标的非线性特征,最后使用BP神经网络方法研究了空车扣车分类。结果表明:单隐层BP神经网络空车调配识别率达到0.8750;双隐层BP神经网络空车调配识别率达到0.8906。可见在编组站源头进行空车调配分类对减少货运、车辆、调度等部门的工作量,提升编组站空车调配区域内的车流组织效率都有较为积极的影响。
In order to reduce the defective rate of empty railway wagons,based on the detailed analysis of the organisation process of empty and load railway wagons flow,this paper puts forward the countermeasures for the source classification of the empty railway wagons in the marshalling station.Firstly,an evaluation index system based on the criteria of railway wagons detention has been established by screening the evaluation indicators of railway wagons detention in freight department,such as the time frame of overhaul.Secondly,the Pearson correlation coefficient has been utilised to analyse the nonlinear characteristics of the evaluation index.Finally,the classification of empty railway wagons detention has been studied by using BP neural network method.The results indicate that the recognition rate of single hidden layer BP neural network has reached 0.8750.The double hidden layer BP neural network has a recognition rate of 0.8906.It can be seen that the classification of empty railway wagons at the source of marshalling stations has a positive impact on reducing the workload of freight,depot and dispatching departments and improving the traffic flow organisation efficiency in the area of empty railway wagons distribution at the marshalling station.
作者
孔德扬
高磊
张力
陈栋
Kong Deyang;Gao Lei;Zhang Li;Chen Dong(School of Traffic and Transportation,Xi'an Traffic Enginering Institute,Xi'an 710000,China;Xi'an Xi Railway Station,China Railway Xi'an Bureau Group Co.,Ltd.,Xi'an 710000,China;JiangSu Port Group Logistics Co.,Ltd.,NanJing 210000,China)
出处
《甘肃科学学报》
2021年第6期56-61,68,共7页
Journal of Gansu Sciences
基金
西安交通工程学院2019年度中青年基金项目(19KY-11)。
关键词
扣车率
空车调配分类
Pearson相关性系数
BP神经网络
Defective rate
Classification of the empty railway wagons
Pearson correlation coefficient
BP neural network
