K-means-WNN法解析EDXRF中Fe、Ta含量算法评价

Estimation of Fe and Ta Contents in EDXRF by K-means—WNN Method

运用能量色散X射线荧光分析方法,结合聚类分析和小波神经网络建立改进算法模型对EDXRF中Fe、Ta元素含量进行更好的定量准确分析。结果表明,改进后的算法模型与未加入聚类分析的小波神经网络算法结果相比,对五种矿物Fe含量的平均相对误差降低了11.89%,对Ta含量的平均相对误差降低了9.94%;对Fe含量的平均均方根误差降低了15.21%,对Ta含量的平均均方根误差降低了11.24%。该模型可以更好地反映元素特征X射线强度计数与含量之间的非线性关系,提高解析精准度。

This paper makes use of analysis method of energy dispersion of X ray fluorescence,and combines with clustering analysis and wavelet neural network to build up an improved algorithm model. This model is able to quantify Fe and Ta in EDXRF accurately.The results show that the developed model lowers the average error of the content of Fe and Ta by 11.89%and 9.94%,and average mean square root by 15.21%and 11.24%in comparison with the model of wavelet neural network without clustering analysis.This model reflects the nonlinear relation of element characteristic X ray count to content of the element,and thus improves the analysis accuracy.