摘要
采用普鲁克分析(PA)算法和分段直接校正(PDS)算法,解决化学计量学多元校正中的模型传递问题。选择红外谱图严重混叠的4种大气有机毒物——丙酮、苯、三氯甲烷和甲醇作为分析对象,文中的光谱数据来自2部分:美国环境保护署(EPA)数据库和实验室的实测遥感傅里叶变换红外(RS-FTIR)光谱数据。研究PA算法、PDS算法中主因子数及传递样品数对传递结果的影响,分别计算2种算法对该4组分体系的预测均方根误差(RMSEP)并进行比较。预测结果表明:2种算法均取得了较好的模型传递效果,其中PDS算法丙酮的RMSEP为0.145,PA算法丙酮的RMSEP为0.122,PA算法优于PDS算法。
To solve the problem of calibration transfer in multivariate calibration of chemometrics, Procrustes analysis (PA) algorithm and piecewise direct standardization (PDS) algorithm are applied. Acetone, benzene, chloroform and methanol, four air toxic organic compounds with the seriously overlapping IR spectrogram, are analyzed. The spectra data are selected from American EPA database and remote sensing Fourier transform infrared (RS-FTIR) spectrum data measured in the laboratory. The influence of the principal factor number and the transfer sample number on transferring results is studied, and the root mean square error of prediction (RMSEP) of four component systems in PA and PDS algorithm is calculated and compared. The predicted result shows that both of the two methods gain good calibration transfer effect and PA is better than PDS. Taking acetone for an example, the RMSEP is 0. 145 in PDS while the RMSEP is 0. 122 in PA.
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2008年第6期788-792,共5页
Journal of Nanjing University of Science and Technology
基金
国家自然科学基金(60571055)
江苏省教育厅项目(08KJB610007)
南通大学博士科研启动基金(03080112)
关键词
遥感傅里叶变换
红外光谱
分段直接校正
普鲁克分析
多元校正
模型传递
多组分分析
remote sensing Fourier transform
infrared spectrum
piecewise direct standardization
Procrustes analysis
multivariate calibration
calibration transfer
multicomponent analysis