X-射线荧光光谱分析中偏最小二乘回归技术与基本参数法的研究  

Study of Partial Least Squares Regression and Fundamental Parameters Method in X-Ray Fluorescence Analysis

作　　者：罗立强[1] 马光祖[1] 吉昂[2] 

机构地区：[1]中国地质科学院岩矿测试技术研究所,北京100037 [2]中国科学院上海硅酸盐研究所

出　　处：《分析化学》1992年第9期1074-1077,共4页Chinese Journal of Analytical Chemistry

摘　　要：将偏最小二乘回归技术(PLS)与X-射线荧光分析中的基本参数法(FP)相结合,编制PLSFP软件,运用熔样技术,可以定量测定地质岩石样品中的十三个主,次量元素,并使准确度提高,模型预测能力增强。模型选择和最佳维数的确定方法是成功的关键。PLSFP软件采用结构化编程,由标准FORTRAN 77语言写成,可在UNIX,XENIX和DOS操作系统下运行。The partial least squares regression (PLS) combined with fundamental parameters methcd(FP) is applied to the determination of 13 elements in the geological rock samples by X-ray fluorescence analysis(XRFA). The combinaticn of PLS With FP can make accuracy better, and only several standard samples(5～7 samples) are needed This is because the PLS eliminates. the colinearity problem and by using only the sighificant number of latent vatiables a noise filtering effect is obtained which resuits in improvd predictive ability of PLS, and the FP corrects matrix effect with a few standard samples. The key points of PLS-FP method are how to build up PLS model and select the optimal number of dimensicns for PLS model. The PLS-FP model here is based on the theoretical intensities calcuated by the fundamental parameter equations, the concentrations and the intensities of the standard samples. The selection of the optimal numbe of dimensicns for the PLS mcdel is accomplished by using the cross-validaticn (CV) method leaving cut one sample at a time. The predietion residual error sum of squares (PRESS) is calculated in the same manner each time a few factors are added to the PLS mcdel. The PRESS-dimension graphs are given in this paper and analytical results of 13 elements in 5 geological rock standard samples by the PLS-FP method are listed. The analytical accuracy is good. The PLS-Fp software is written in FORTRAN 77 and runs in the UNIX, XENIX and DOS operation systems.

关 键 词：基本参数法 X射线 荧光谱 PLS技术 

分 类 号：O657.34[理学—分析化学]