神经网络融合多源信息的列车测速方法研究  

Research on Train Speed Measurement Method Based on Multiple-source Information Fusion Using Neural Network

作　　者：陆海亭 孙春洋 王亮军 付保明 陈林山[1] Lu Haiting;Sun Chunyang;Wang Liangjun;Fu Baoming;Chen Linshan(Nanjing Vocational Institute of Transport Technology,Nanjing 211188,China;Suzhou Rail Transit Group Co.,Ltd.,Suzhou 215006,China)

机构地区：[1]南京交通职业技术学院,南京211188 [2]苏州市轨道交通集团有限公司,江苏苏州215006

出　　处：《铁路通信信号工程技术》2024年第5期90-95,共6页Railway Signalling & Communication Engineering

基　　金：国家重点研发计划项目(2020YFB1600700);南京交通职业技术学院重点科研项目(JZ2204)。

摘　　要：地铁列车自主测速定位的准确性与可靠性是保障其行车安全和效率的先决条件。在目前常用地铁列车测速方法的基础上,研究传感器的选型以及相应多源数据融合算法,提出运用人工神经网络融合光电式传感器、多普勒雷达传感器以及加速度计的地铁列车智能测速方法。该方法充分利用人工神经网络的自学习、自适应、非线性的能力,选用光电式传感器、多普勒雷达、加速度计这三种传感器的实测数据作为其输入,选取RBF人工神经网络智能融合快速寻优,自适应地调整它们的实测数据权重,从而得到地铁列车的实时速度值,以期达到进一步提高列车测速的精确性与可靠性的目的。The accuracy and reliability of autonomous speed measurement and positioning for subway trains are prerequisites for ensuring their driving safety and efficiency.This paper builds on the commonly used speed measurement methods for subway trains,and studies the selection of sensors and corresponding algorithms for multi-source data fusion.On this basis,it proposes an intelligent speed measurement method for subway trains using artificial neural networks and combining photoelectric sensors,Doppler radars and accelerometers.The proposed method fully utilizes the self-learning,adaptive,and non-linear capabilities of artificial neural networks.It uses the measured data from three types of sensors,i.e.photoelectric sensors,Doppler radars and accelerometers as its input.It utilizes RBF artificial neural networks for intelligent fusion and rapid optimization.Thus,the weights of the measured data from these sensors are adaptively adjusted to obtain the real-time speed values of subway trains,to achieve the goal of further improving the accuracy and reliability of train speed measurement.

关 键 词：测速方法 多源信息融合 RBF神经网络 地铁列车 

分 类 号：U284.481[交通运输工程—交通信息工程及控制]