基于粗糙集-C4.5的轨道电路故障诊断方法研究

Research on Railway Track Circuit Fault Diagnosis Based on Rough set and C4.5 Decision Tree Algorithm

针对轨道电路故障信息存在大量重复样本和冗余属性,提出一种基于粗糙集和C4.5决策树算法相融合的轨道电路故障诊断方法。轨道电路故障特征数据多为连续量,需要根据模糊集理论对故障样本进行模糊化,形成离散决策表。利用粗糙集处理不完备决策表的能力,去除离散决策表的冗余属性得到约简表,结合决策树C4.5算法对约简决策表进行快速训练提取诊断规则,产生的诊断规则清晰、可解释性强,能够直接运用于轨道电路故障诊断中。最后利用模拟数据仿真验证该方法的有效性,与ID3算法和BP神经网络法进行对比,仿真测试表明该方法具有更高的诊断效率和准确率,对实现轨道电路快速鲁棒故障诊断具有一定意义。

For the issue of many duplicate samples and redundancy attributes of railway track circuit fault information, this paper presents the track circuit fault diagnosis method of integrating a rough set and C4.5 decision tree algorithm. Because track circuit fault data is mostly continuous, the fault samples need to be fuzzified to discrete decision table according to the fuzzy set theory. The rough set＇s ability of handling incomplete decision table is used to reduce the redundant attributes for resulting in a reduction table. And C4.5 algorithm is used to train reduction table rapidly for extracting clear diagnostic rules which can be directly applied to track circuit fault diagnosis. The simulation results show that the diagnostic efficiency and accuracy of this method is higher than that of ID3 and BP neural network, which is of great significance to fast robust fault diagnosis of track circuits.