基于波形特征提取和FA-Grid SVM的MVB故障诊断

列车通信网络的故障诊断一直是列车健康管理的难点,文章针对列车MVB(多功能车辆总线)网络,提出了一种基于波形特征提取和联合萤火虫网格寻优支持向量机(FA-Grid Support Vector Machines, FA-Grid SVM)相结合的故障诊断方法。通过提取MVB总线物理波形的时域特征,作为支持向量机的样本,构建MVB故障数据集;基于SVM较优参数点基本集中于同一区域这一现象,提出FA-Grid两步寻优的参数优化模型。试验结果表明,与传统网格寻优和遗传算法(GA)相比,提出的FA-Grid寻优模型时间复杂度低,分类效率高,能够准确地对MVB故障进行诊断。

MVB Fault Diagnosis Based on Waveform Feature Extraction and FA-Grid SVM

The fault diagnosis of train communication network has always been achallenge in train health management. A fault diagnosis method based on waveform feature extraction and FA-Grid SVM for multi-function vehicle bus(MVB) was proposed. The time-domain features were extracted from physical waveform of the MVB bus and used as inputs of SVM which construct MVB fault dataset. Due to the concentration of optimal parameters of SVM, a two-step parameter optimization method based on FA-Grid was provided. Experimental results show that compared with traditional grid optimization and genetic algorithm(GA), the proposed FAGrid optimization model has lower complexity and higher efficiency and could accurately diagnose MVB faults.