基于近似熵及EMD的高铁故障诊断  

Fault Diagnosis of High-speed Rail Based on Approximate Entropy and Empirical Mode Decomposition

作　　者：赵晶晶[1] 杨燕[1] 李天瑞[1] 曾京[2] 魏来[2] 

机构地区：[1]西南交通大学信息科学与技术学院,成都610031 [2]西南交通大学牵引动力国家重点实验室,成都610031

出　　处：《计算机科学》2014年第1期91-94,99,共5页Computer Science

基　　金：国家自然科学基金(61134002;61170111;61175047);西南交通大学牵引动力国家重点实验室自主研究课题(2012TPL_T15);中央高校基本科研业务费专项资金资助项目(SWJTU11ZT08)资助

摘　　要：抗蛇行减振器故障、横向减振器故障、空气弹簧故障是高铁的3种典型故障。针对高铁的3种常见故障的非线性、非平稳特性,本次研究中将近似熵和经验模态分解应用到高铁故障诊断中进行故障特征提取,并使用BP神经网络作为高铁故障诊断模型进行高铁的故障诊断。实验证明,该方法能够准确有效地进行高铁故障诊断。此外,通过对比实验表明,融合近似熵特征和EMD分解后的第一个模态分量的能量特征比单个特征更有利于高铁故障诊断。The faults of anti-yaw damper, lateral damper and air spring are three kinds of common faults of high-speed rail. According to the non-stationary and nonlinear characteristic of three kinds of common faults of high-speed rail, ap- proximate entropy and empirical mode decomposition were introduced to extract the feature of high-speed rail faults, and BP neural network was used as the model for the fault diagnosis of high-speed rail. The experimental results show that the proposed method is effective. In addition, the comparison experiment indicates that the fault diagnosis based on the combination of approximate entropy and empirical mode decomposition obtains better result than the fault diagnosis based on only one feature.

关 键 词：特征提取 近似熵 经验模态分解 神经网络 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]