轨道不平顺作用下铁路列车车体振动状态的PCA-SVM预测分析  

PCA-SVM Forecast of Car-body Vibration States of Railway Locomotives and Vehicles under the Action of Track Irregularities

作　　者：徐磊[1] 陈宪麦[1] 

机构地区：[1]中南大学土木工程学院,湖南长沙410075

出　　处：《铁道学报》2014年第7期16-23,共8页Journal of the China Railway Society

基　　金：国家自然科学基金(51008315)

摘　　要：为快速预测铁路机车车辆在不平顺轨道上的的振动状态,根据车体振动加速度的3个评价指标(绝对峰值、标准差、绝对平均值),提出基于PCA-SVM方法的车体振动状态分类预测模型。首先对不同评价指标和车体振动状态下的轨道不平顺样本进行聚类,提取轨道不平顺样本中的特征统计参数,并进行PCA参数降维和信息优化,最后以不同状态下各种评价指标的车体振动主要特征值为训练样本,构建SVM多类分类器。对轨道检查车多次实地检测的数据采用PCA-SVM分类器计算的分析结果表明:绝对均值、方根均值、方根幅值等主要轨道不平顺统计参数控制车体的整体振动状态,其他特征参数起调节车体振动的作用;采用扭曲和水平不平顺作为预测模型的输入,可使车体振动状态的测算准确率达到90%以上。In order to achieve fast prediction of car-body vibration states of railway locomotives and vehicles, this paper put forward the PCA-SVM based dassified car-body vibration prediction model according to three e-valuation indexes （the absolute peak,standard deviation,and absolute mean value）of the car-body vibration acceleration.Firstly,we clustered the track irregularity samples under different evaluation indexes and differ-ent carbody vibration states respectively,extracted the characteristic statistical parameters,and then the process of PCA dimensionality reduction and message optimization was conducted.At last,we constructed the SVM multi-class classifiers by training the principal components samples.Analyzing the data measured by track inspection cars in multiple field tests with the PCA-SVM classifier,we get the following results：The absolute mean value,root-mean-square value and root amplitude,etc.are the main statistical track irregularity parameters for control of overall car-body vibration states,and the other characteristic parameters play the role of regulating car-body vibrations;putting the track twist irregularity and cross-level irregularity as the input of the prediction model,we can obtain more than 90% of the accuracy in predicting car-body vibration states.

关 键 词：轨道不平顺 车体振动 特征参数 主成分分析 支持向量机 分类预测 

分 类 号：U270.1[机械工程—车辆工程]