基于高速轮轴实测载荷的非参数拟合外推研究

Non-parametric fitting extrapolation based on measured load of high-speed axle

以高速动车组动车轮轴动应力线路实测数据为依据,通过载荷与应变之间的关系,结合包络线的方法,得到实时的轮轴载荷,并编制实测轮轴载荷谱.对实测的高速轮轴载荷谱进行了非参数核密度估计的拟合外推分析,对比了固定带宽核密度和自适应带宽核密度两种方法的极值预测结果以及扩展载荷谱与实测载荷谱的损伤差值,并讨论了外推倍数对两种方法的影响.结果表明:自适应带宽核密度法预测的极值和频次更接近于实测值,损伤误差更小,由有限数据外推高频次数据的精度更高,更适合于高速轮轴载荷谱的外推,并初步验证了拟合外推方法的合理性.

Based on the measured data of dynamic stress on EMU axle in high-speed lines,the real-time axle load is obtained by analysis of the relationship between the load and the strain,combined with the envelope method.The measured axle load spectrum is thus compiled.Then it is analyzed through the fitting and extrapolation method,which the non-parametric kernel density is estimated.This paper compares the prediction results of the extremum of the fixed bandwidth kernel density and the adaptive bandwidth kernel density.Besides,the damage error between the extended load spectrum and the measured load spectrum is compared.The influence of extrapolation multiple on the two methods is then discussed.The results show that the extremum and frequencies predicted by the adaptive bandwidth kernel density method are closer to the measured values,and they have smaller damage error.Using finite data to extrapolate high-frequency data has higher accuracy,which is more suitable for the extrapolation of the high-speed axle load spectrum.Therefore,the fitting extrapolation method is proved to be feasible after preliminary analysis.