基于多方法证据融合的轨道电路故障诊断  

Fault Diagnosis of Track Circuit Based on Multi Method Evidence Fusion

作　　者：于晓英[1] 董煜 董昱[1] YU Xiaoying;DONG Yu;DONG Yu(School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China;College of Electrical Engineering,Lanzhou Institute of Technology,Lanzhou 730050,China)

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070 [2]兰州工业学院电气工程学院,甘肃兰州730050

出　　处：《铁道学报》2021年第2期86-94,共9页Journal of the China Railway Society

基　　金：国家自然科学基金(61763023);甘肃省高等学校创新能力提升项目(2020A-044)。

摘　　要：针对采用单一方法进行轨道电路故障诊断时存在的局限性及诊断精度偏低的缺陷,提出一种基于灰色关联分析、模糊综合评判和反向传播BP神经网络三种方法证据融合的诊断策略。分别应用每种方法对轨道电路故障进行初步故障诊断,每个诊断输出构造三个证据的基本概率赋值BPA;确定权值并重构证据BPA;按照证据理论合成规则融合重构证据并给出轨道电路故障诊断结论。诊断实例结果表明,基于多方法证据融合的轨道电路故障诊断方法能够克服单一故障诊断方法的不足,有效地提高了故障诊断的准确度。Due to the limitation and the low diagnostic accuracy of single-method track fault diagnosis,a fault diagnosis strategy for track circuits based on multi-method evidence fusion was proposed in this paper.Firstly,the preliminary fault diagnosis was carried out by grey relational analysis,fuzzy comprehensive evaluation and BP(Back Propagation)neural network respectively,to construct the basic probability assignments of three pieces of evidence by the output of each diagnosis.Secondly,the weight was determined according to the reliability and trust of the evidence in order to reconstruct the basic probability assignment of the evidence.Finally,evidence theory combination rules were used to fuse the reconstructed evidence to give the fault diagnosis conclusion of the track circuit.The results of diagnosis examples show that the track circuit fault diagnosis method based on multi-method evidence fusion can overcome the shortcomings of single fault diagnosis method,and effectively improves the accuracy of fault diagnosis.

关 键 词：轨道电路 故障诊断 证据理论 BPA 可靠度 信任度 

分 类 号：U284.24[交通运输工程—交通信息工程及控制]