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

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.