US 12,339,219 B2
Adaptive characteristic spectral line screening method and system based on atomic emission spectrum
Jia Liu, Beijing (CN); Feipeng Cui, Beijing (CN); Xiaopeng Li, Beijing (CN); Ling Liu, Beijing (CN); Hongwei Shi, Beijing (CN); Feifei Guo, Beijing (CN); Ying Zhao, Beijing (CN); Xuejing Shen, Beijing (CN); and Haizhou Wang, Beijing (CN)
Assigned to NCS TESTING TECHNOLOGY CO., LTD., Beijing (CN)
Filed by NCS Testing Technology CO., LTD., Beijing (CN)
Filed on Oct. 10, 2022, as Appl. No. 17/962,602.
Prior Publication US 2024/0068936 A1, Feb. 29, 2024
Int. Cl. G01N 21/31 (2006.01); G06F 18/2132 (2023.01); G06N 3/126 (2023.01)
CPC G01N 21/3103 (2013.01) [G06F 18/21322 (2023.01); G06N 3/126 (2013.01); G06F 18/21326 (2023.01); G06F 2218/04 (2023.01)] 6 Claims
OG exemplary drawing
 
1. An adaptive characteristic spectral line screening method based on atomic emission spectrum, comprising following steps:
step 1: determining a spectral dataset based on original spectral signals of a sample to be analyzed;
step 2: performing a plurality of optimization rounds of characteristic screening on the spectral dataset by using a set characteristic screening optimization method, obtaining an initialized spectral dataset of each round of the plurality of optimization rounds of the characteristic screening and initialized characteristic population genes corresponding to the initialized spectral dataset;
step 3: obtaining an optimal characteristic population gene of the each round by a set analysis method, a fitness function, and an iteration of a genetic algorithm based on the initialized spectral dataset and the initialized characteristic population genes;
step 4: obtaining an optimized characteristic spectral information set corresponding to an optimal characteristic population gene set composed of the optimal characteristic population gene of the each round of the plurality of optimization rounds when the plurality of optimization rounds reach set optimization rounds; and
step 5: performing combination statistics and discriminant analyses on the optimized characteristic spectral information set to complete an adaptive characteristic spectral line screening;
wherein the step 3 comprises:
selecting a corresponding analysis method as the set analysis method according to analysis requirements, the set analysis method determining a model parameter and an evaluation indicator of the set analysis method based on the initialized spectral dataset and the initialized characteristic population genes;
the fitness function obtaining a population fitness based on the initialized characteristic population genes and the model parameter and the evaluation indicator of the set analysis method; and
the genetic algorithm iterating based on the population fitness and the initialized characteristic population genes to obtain a targeted characteristic population gene until an iterative algebra of the genetic algorithm reaches a set maximum value or the population fitness reaches a set fitness threshold value, and taking a last obtained targeted characteristic population gene as the optimal characteristic population gene;
wherein the step 5 comprises:
performing a probability analysis and a frequency analysis on the optimized characteristic spectral information set to obtain statistical information of each characteristic spectral line; and
completing the adaptive characteristic spectral line screening of the optimized characteristic spectral information set by the discriminant analyses, when an evaluation value corresponding to the statistical information of the each characteristic line being greater than a set screening threshold value;
wherein the adaptive characteristic spectral line screening method further comprises:
applying the adaptive characteristic spectral line screening to classify materials based on the atomic emission spectrum.