以对铁道车辆轴箱振动非高斯特征与分布为对象开展研究.基于列车线路轴箱实测加速度信号,提取由轨道冲击引起的轴箱振动特征非高斯信号.使用多个概率密度函数(Probability Density Function,PDF)模型对实测信号进行拟合,并与实测特征信号的经验分布进行对比,评估各模型对轴箱特征非高斯信号的拟合精度.基于W-H非线性变换模型,建立一种非高斯信号模拟方法.利用模拟信号分析非高斯特征对各模型拟合精度的影响.结果表明:列车在行驶过程中具有非高斯特征,当列车经过轨道焊接接头、道岔与波磨路段时,由于轮轨冲击,非高斯特征明显增大,车轮多边形对信号非高斯特征几乎没有影响;基于W-H模型的非线性变换法,可以在保证模拟信号功率谱与指定功率谱基本一致的情况下,进行不同非高斯特征的信号模拟;高斯混合模型能够对铁道车辆非高斯信号较为准确地拟合;随着模拟非高斯信号峭度与偏度的增大,各模型与经验分布的相对误差也会增大,其中高斯混合模型拟合精度相对较高.
Non-Gaussian Characteristics and Distribution of Axlebox Vibration in Railway Vehicles
The non-Gaussian characteristics and distribution of axle box vibration of railway vehicles are studied.Based on the measurement signal of the axle box of the train,the characteristic non-Gaussian signal of the axle box vibration caused by the track impact is extracted.Multiple probability density function models are used to fit the non-Gaussian signals,and the fitting accuracy of each model to the non-Gaussian signals is evaluated by comparing with the empirical distribution of track characteristic signals.Based on W-H nonlinear transformation model,a non-Gaussian signal simulation method is established.The influence of non-Gaussian characteristics on the fitting accuracy of each model is analyzed by using the analog signal.The results show that the train operation process has non-Gaussian characteristics.When the train passes through the welded joints of the track,the turnout and the wave-worn section,the non-Gaussian characteristics increase obviously due to the wheel-rail impact,and the wheel polygon has little effect on the non-Gaussian characteristics of the signal.The nonlinear transformation method based on W-H model can simulate the signal with different non-Gaussian characteristics under the condition that the power spectrum of the reproduced signal is basically consistent with the specified power spectrum.Gaussian mixture model can accurately fit the non-Gaussian signal of railway rolling stock.As the kurtosis and skewness of simulated non-Gaussian signals increase,the relative errors of each model and empirical distribution also increase,and the fitting accuracy of Gaussian mixture model is relatively high.
vibration and wavenon-Gaussian stochastic vibrationGaussian mixed modelprobability density functionrail featuresnonlinear transformation
万方数据
维普
