基于最大期望算法的列车完整性检测方法

Expectation Maximization Algorithm Based Train Integrity Detection Method

面向中国中西部低密度线路条件,基于GNSS的列车完整性检测系统是一种经济有效的列车完整性解决手段,也可作为制动管风压检测的辅助冗余检测方式,以进一步保障列车运行安全。针对列车运行过程中基于GNSS列车完整性检测系统问题建立完整性检测模型,模型包含列车完整状态及断裂状态下的系统时间动态转换过程。基于系统参数先验统计特性未知情况下的模型参数估计,提出基于最大期望算法的离线训练学习方法,得到列车完整性检测系统模型后验参数的极大似然估计。在列车完整性检测过程中,针对当前时刻的观测值,进行列车完整性所有状态的概率推理估计,采用高斯和滤波的方法实现状态的概率推理计算,选取最大概率下的列车完整性状态,实现列车完整性检测。结合现场测试及仿真数据,针对本文提出的列车完整性检测方法进行测试验证,结果证明了该方法的有效性。

Global Navigation Satellite System（GNSS）-based Train Integrity Monitoring System（TIMS）is a costeffective solution for train integrity detection for the middle and low density railway lines in western China.At the same time,it can be used as an auxiliary redundancy detection method for air pressure-based brake pipe detection,which allows for the increased level of safety of train operation.This paper investigated the problem of train integrity detection which determines whether train consists remain intact during train operation.A new train integrity detection model was built,where both train complete state and parted state were considered.Based on the model parameter estimation under the unknown a priori statistical characteristics of the system parameters,an off-line training learning method based on expectation maximization algorithm was presented to find out the posterior system parameter of train integrity detection to reach the maximum likelihood estimation.During the train integrity detection process,in respect of observed value for the current time,probabilistic inference estimation on all-state train integrity was carried out.The Gaussian Sum Filtering method was introduced to realize the probabilistic inference calculation to select train integrity state under maximum probabilityto achieve train integrity detection.Field tests were carried out,and a simulation was presented to evaluate the proposed method based on the statistical data.The results showed the effectiveness of the proposed train integrity detection method.