X射线荧光光谱测定地质样品中27种组分分析结果不确定度的评估

建立了一种X射线荧光光谱分析方法分析结果不确定度的评估模式。以大量有可靠值的硅酸盐类地质样品的偏振能量色散X射线荧光光谱的分析结果为研究对象，对27个组分分别采用浓度分段，分别评估不同浓度段内各个组分分析方法精密度引入的相对不确定度(n=6，一个浓度段内各水平相对标准偏差的平均值)和准确度引入的相对不确定度(一个浓度段内各个水平的相对偏差的相对标准偏差)，依据误差传递理论，合成精密度和准确度分别引入的相对不确定度得到总的相对不确定度。这种基本模式可以解决一定浓度范围内常规大量类似基体未知地质样品不确定度评估难的问题，不必对每个未知样品的分析结果都进行不确定度的评估，可以从这种模式统计的结果中给出不确定度的值。这种评估模式是一种基本的具有实际应用价值的统计模式，可以为后续的不确定度评估的函数模式的建立提供有力的基础。但是该不确定度评估模式涉及的有可靠值的样品数量较多，难以简单应用到其他基体样品分析结果的不确定度评估。争取以本研究为基础尽快建立一种合理的数理统计函数模式以适用于不同类型样品X射线荧光光谱分析结果不确定度的评估。

The Uncertainty Evaluation of Analytical Results of 27 Elements in Geological Samples by X-Ray Fluorescence Spectrometry

Evaluating uncertainty of analytical results with 165 geological samples by polarized dispersive X-ray fluorescence spectrometry (P-EDXRF) has been reported according to the internationally accepted guidelines. One hundred sixty five pressed pellets of similar matrix geological samples with reliable values were analyzed by P-EDXRF. These samples were divided into several different concentration sections in the concentration ranges of every component. The relative uncertainties caused by precision and accuracy of 27 components were evaluated respectively. For one element in one concentration, the relative uncertainty caused by precision can be calculated according to the average value of relative standard deviation with different concentration level in one concentration section, n=6 stands for the 6 results of one concentration level. The relative uncertainty caused by accuracy in one concentration section can be evaluated by the relative standard deviation of relative deviation with different concentration level in one concentration section. According to the error propagation theory, combining the precision uncertainty and the accuracy uncertainty into a global uncertainty, this global uncertainty acted as method uncertainty. This model of evaluating uncertainty can solve a series of difficult questions in the process of evaluating uncertainty, such as uncertainties caused by complex matrix of geological samples, calibration procedure, standard samples, unknown samples, matrix correction, overlap correction, sample preparation, instrument condition and mathematics model. The uncertainty of analytical results in this method can act as the uncertainty of the results of the similar matrix unknown sample in one concentration section. This evaluation model is a basic statistical method owning the practical application value, which can provide a strong base for the building of model of the following uncertainty evaluation function. However, this model used a lot of samples which cannot simply be applied to other types of samples with different matrix samples. The number of samples is too large to adapt to other types samples. We will strive for using this study as a basis to establish a reasonable basis of mathematical statistics function mode to be applied to different types of samples.