Open AccessApplication of small sample BP neural network in quantitative analysis of EDXRF
Author(s) -
Fēi Li,
Zhuoyao Tang,
Liangquan Ge,
Nanxing Wu,
Cheng Feng,
K. Sun
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1423/1/012054
Subject(s) - randomness , computer science , artificial neural network , quantitative analysis (chemistry) , sample (material) , matlab , algorithm , data mining , biological system , artificial intelligence , pattern recognition (psychology) , statistics , mathematics , chemistry , chromatography , biology , operating system
Quantitative analysis of EDXRF is affected by matrix effect, randomness of nuclear radiation, interaction of elements, statistical fluctuation of radiation detection process, etc. Its algorithm needs to consider many factors, and the established functional relationship is often a complex non-linear function. Therefore, effective quantitative analysis method has always been the key research direction of spectral resolution technology. In this paper, we use MATLAB software to study the effect of BP neural network in quantitative analysis of EDXRF by establishing the non-linear relationship between the counting rate of each element and the content of a single element in a small sample. The experimental results show that the small sample neural network can establish a stable structure and be applied to quantitative analysis of EDXRF, but the number of samples restricts the accuracy of prediction results, most of which can only guarantee 20% to 30% relative error.