基于粗糙集理论和FCM的轨道电路故障诊断模型

由于轨道电路数据多且其维数高,这往往会导致所选特征之间存在冗余和相容性的问题。基于主分量启发式算法,引入相容度概念,并提出一种改进的主分量启发式属性约简算法,提取初始数据的主要特征属性来降低样本的维数。同时将模糊认知图概念引入到轨道电路故障诊断中,利用最小二乘法完成模糊认知图节点间权值的选择,最后根据权值建立轨道电路诊断模型并对预处理的样本进行训练和分类。实验结果表明,与单独的FCM分类器相比较,加入属性约简算法后, FCM分类器可提高分类性能,与采用人工确定权值的FCM方法对比,最小二乘法提高了FCM分类的精度。

Track Circuit Fault Diagnosis Model Based on Principal Component Heuristic Algorithm

The data of the track circuit is large in quantity and high in dimension, which often leads to the problem of redundancy and compatibility between selected features. An improved principal component heuristic attribute reduction algorithm is proposed by introducing the concept of compatibility based on the principal component heuristic algorithm to extract the main feature attributes in the original data and reduce the dimensionality of the sample. At the same time, the concept of the fuzzy cognitive map is introduced into the fault diagnosis of track circuits. The least squares technique is used to complete the selection of weights between fuzzy cognitive graph nodes. Finally, the track circuit diagnostic model is built according to the weights and the pre-processed samples are trained and classified. The experimental results show that compared with the single FCM classifier, the FCM classifier with attribute reduction algorithm improves the classification performance. Compared with the FCM method with manual weight determination, the least square technique improves the accuracy of FCM classification.