LIBS分析模型对铝合金定量分析精度的影响  

Influence of LIBS Analysis Model on Quantitative Analysis Precision of Aluminum Alloy

作　　者：李明亮 戴宇佳 秦爽 宋超[2] 高勋[1] 林景全[1] LI Ming-liang;DAI Yu-jia;QIN Shuang;SONG Chao;GAO Xun;LIN Jing-quan(School of Science,Changchun University of Science and Technology,Changchun 130022,China;School of Chemistry and Environmental Engineering,Changchun University of Science and Technology,Changchun 130022,China)

机构地区：[1]长春理工大学理学院,吉林长春130022 [2]长春理工大学化学与环境工程学院,吉林长春130022

出　　处：《光谱学与光谱分析》2022年第2期587-591,共5页Spectroscopy and Spectral Analysis

基　　金：国家自然科学基金项目(61575030);吉林省科技厅项目(20200301042RQ)资助。

摘　　要：为了提高铝合金定量分析的精度,将激光诱导击穿光谱技术与多变量线性回归、中值高斯核支持向量机回归及标准化偏最小二乘回归等方法相结合,建立铝合金中Cu元素定量分析模型。对采集的LIBS光谱进行三阶极小值去背景和小波阈值降噪处理,从而提高LIBS光谱的信背比。将处理后数据筛选最佳训练集、预测集并用多变量线性回归、中值高斯核支持向量机回归法和标准化偏最小二乘拟合回归等建立定标模型。选用CuⅠ324.80 nm,CuⅠ327.43 nm两条特征谱线以及323~329 nm范围内的LIBS光谱数据进行定量分析,对比分析三种LIBS定量分析模型的拟合系数(R^(2))、定标均方根误差(RMSEC)、预测均方根误差(RMSEP)和平均相对误差(ARE)等。结果表明,相对于多变量线性回归和中值高斯核支持向量机回归法两种LIBS定量分析模型,对于铝合金中的Cu元素定量分析,标准化PLSR模型的精度和准确度都有明显的提高,并且LIBS定标曲线的R^(2),RMSEC,RMSEP和ARE分别为0.997,0.014 Wt%,0.129 Wt%和3.053%。研究结果表明在提高定标模型精确度与泛化性方面,标准化PLSR方法更具有优势,能够有效地减小参数波动和自吸收效应对铝合金定量分析的影响。In order to improve the accuracy of quantitative analysis of aluminum alloy,a quantitative analysis model of Cu element in aluminum alloy was established by combining laser-induced breakdown spectroscopy with multivariate linear regression,median Gaussian kernel support vector machine regression and standardized partial least squares regression.Third order minimum background removal and wavelet threshold denoising were performed on the collected LIBS spectra to improve the SNR of LIBS spectra.The optimal training set and prediction set were selected from the processed data.The calibration model was established using multi variable linear regression method,medium Gaussian kernel support vector machine regression method and normalized partial least squares fitting regression method.Two characteristic lines of CuⅠ324.80 nm and CuⅠ327.43 nm and Libs spectral data in the range of 323~329 nm were used for quantitative analysis.The fitting coefficient(R^(2)),root mean square error(RMSEC),root mean square error of prediction(RMSEP)and average relative error(ARE)of the three Libs quantitative analysis models were compared and analyzed.The results show that compared with the multivariable linear regression method and medium Gaussian kernel support vector machine regression method,the precision and accuracy of the standardized PLSR model are significantly improved for the quantitative analysis of Cu element in aluminum alloy,and the R^(2),RMSEC,RMSEP and ARE of the Libs calibration curves are 0.997,0.014 Wt%,0.129 Wt%and 3.053%,respectively.The results show that the standardized PLSR method has more advantages in improving the accuracy and generalization of the calibration model,and can effectively reduce the influence of parameter fluctuation and self-absorption effect on the quantitative analysis of aluminum alloy.

关 键 词：激光诱导击穿光谱 标准化偏最小二乘回归 中值高斯核支持向量机回归 多变量回归 铝合金 

分 类 号：O433.4[机械工程—光学工程]