激光诱导击穿光谱技术对高铁车轮硬度检测模型优化研究

Optimization of Hardness Testing Model of High-Speed Iron Wheel by Laser-Induced Breakdown Spectroscopy

我国铁路跨度长、运营时间长、运行环境变化较大,故对于车轮的磨损较大,为保障高速铁路的安全运行,高速列车车轮表面硬度就成为了一项重要参考指标。激光诱导击穿光谱(LIBS)实验平台对8块不同硬度的HS7高铁车轮用钢样品进行击穿获取LIBS光谱数据,发现基体元素(Fe)和合金元素(Cr,Mo,W)的谱线强度、离子与原子线的强度比值(Ⅱ/Ⅰ)以及合金元素谱线强度与基体元素谱线强度的强度比值(A/M),分别与样品硬度有着不同程度的相关关系。利用此相关关系分别建立了以谱线强度和谱线强度结合谱线强度比值为变量的偏最小二乘法(PLS)定量分析模型,在建立模型前采用标准正态变量变换(SNV)、Savitzky-Golay卷积二阶导和高斯滤波(Gaussian filter)三种预处理方法来减小实验误差。结果表明,以谱线强度为变量的模型中采用SNV预处理后建立的PLS模型效果最佳,校正集的确定系数为0.98,均方根误差为1.30,预测集的确定系数为0.90,均方根误差为2.43;以谱线强度结合谱线强度比值为变量的模型中采用原始数据建立的PLS模型效果最佳,校正集的确定系数为0.99,均方根误差为0.79,预测集的确定系数为0.94,均方根误差为2.44,且通过对比发现以谱线强度结合谱线强度比值为变量的模型其预测精确度及其稳定性相比于以谱线强度为变量的模型均有所提升。该结果表明,利用谱线强度和离子与原子线的强度比、合金元素谱线强度与基体元素谱线强度的强度比相结合的结果作为模型变量,能显著提升PLS模型对于金属材料表面硬度预测的能力,可以构建一种相关性更强的定量分析模型。研究表明,采用激光诱导击穿光谱技术结合偏最小二乘法定量分析高铁车轮硬度具有一定可行性,可将该技术应用于现场诊断、估算高速列车车轮表面硬度,为维持高速列车安全运行提供一定的保障。

China’s railway has a long span,long operation time and great changes in operation environment,so the wear of wheels is large.In order to ensure the safe operation of high-speed railways,the surface hardness of high-speed train wheels has become an important parameter.The laser-induced breakdown spectroscopy(LIBS)experimental platform was used to conduct the breakdown of eight HS7 high-speed rail wheel steel samples with a different hardness to obtain the LIBS spectral data.It was found that the spectral intensity of matrix elements(Fe)and alloy elements(Cr,Mo,W),the intensity ratio of ion line to atomic line(Ⅱ/Ⅰ),and the spectral intensity ratio of alloy elements to matrix elements(A/M)had different degrees of correlation with the hardness of the samples.Partial least squares(PLS)quantitative analysis model with spectral line intensity and spectral line intensity combined with spectral line intensity ratio as variables was established.Before the establishment of the model,three preprocessing methods,standard normal variable transformation(SNV),Savitzky-Golay convolution second derivative and Gaussian filter(Gaussian filter),were used to reduce the experimental error.The results show that the PLS model established by SNV pretreatment is the best in the model with spectral line intensity as a variable.The determination coefficient of the calibration set is 0.98,the root mean square error is 1.30,the determination coefficient of the prediction set is 0.90,and the root means square error is 2.43.The PLS model established with the original data has the best effect in the model with the ratio of spectral line intensity to spectral line intensity as the variable.The determination coefficient of the calibration set is 0.99,the root mean square error is 0.79,the determination coefficient of the prediction set is 0.94,and the root means square error is 2.44.Through comparison,it is found that the prediction accuracy and stability of the model with the ratio of spectral line intensity to spectral line intensity as the variable are improved compared with the model with the spectral line intensity as the variable.The results show the combined results of spectral line intensity and the intensity ratio of ions to atomic lines.Moreover,the spectral line intensity ratio of alloy elements to matrix elements is used as model variables,which can significantly improve the solution of the PLS model for the prediction of surface hardness of metal materials and construct a quantitative analysis model with stronger correlation.Studies have shown that it is feasible to quantitatively analyze the hardness of high-speed railway wheels by using laser-induced breakdown spectroscopy combined with the partial least squares method.This technology can be applied to the field diagnosis and estimation of the surface hardness of high-speed train wheels,guaranteeing the safe operation of high-speed trains.