论偏最小二乘校正方法的稳定性

研究了多元校正方法——偏最小二乘（ＰＬＳ）在波长色散Ｘ射线荧光光谱分析中的应用，并将该法与传统的经验系数法（ＥＣＭ）进行了比较．其考察对象为一组转炉渣文献数据、一组不锈钢样和一组锌精矿样．结果表明，ＰＬＳ比ＥＣＭ准确、稳定．经分析探讨，认为其主要原因是ＰＬＳ能滤除噪音．文中建立了描述浓度和强度本质的关系式．

THE STABILITY OF THE PARTIAL LEAST SQUARES REGRESSION CALIBRATION

In the paper,a kind of multivariate calibration method,the partial least squares regression(PLS),used for the wavelength dispersive X ray fluorescence(XRF) spectrometry is studied,and compared with the traditional empirical coefficient methods(ECM).These studies relate to the intensities and concentrations of converter slags from literature,a set of stainless steel samples,and a set of zinc concentrate samples.The results show that the PLS is more accurate and more stable than ECM.The major reason is that the PLS calibration can filter noise,and set up the intrinsical relations between the concentrations and their intensities.