核能譜測(cè)量中重疊譜峰解析的算法研究
發(fā)布時(shí)間：2018-08-25 11:09
【摘要】：在核科學(xué)技術(shù)給人類的生活帶來便捷服務(wù)和清潔能源的同時(shí),人們逐漸開始關(guān)注核輻射給環(huán)境和身體帶來的影響。通常輻射環(huán)境中存在的放射性物質(zhì)都會(huì)釋放出γ射線,通過對(duì)γ射線的測(cè)量,可以了解放射性物質(zhì)中的核素種類,判斷核素的含量和活度等。然而現(xiàn)實(shí)測(cè)量中,環(huán)境或其他干擾射線的影響會(huì)導(dǎo)致譜信號(hào)出現(xiàn)頻繁的重疊現(xiàn)象。常用的γ射線探測(cè)器中,NaI(Tl)探測(cè)儀由于其探測(cè)效率高、維護(hù)方便、價(jià)格適中等優(yōu)點(diǎn)被廣泛使用,但其對(duì)能量相近的重疊峰分辨能力不強(qiáng),這使得重疊峰的分解成為譜分析中的難題。因此基于此背景,本文根據(jù)γ能譜的統(tǒng)計(jì)分布規(guī)律,在MATLAB平臺(tái)上運(yùn)用期望最大值法、遺傳算法和粒子群算法完成了對(duì)模擬重疊峰的分解。論文主要工作及成果如下:1、首先討論了能譜及其數(shù)學(xué)模型,然后根據(jù)能譜的統(tǒng)計(jì)漲落特性,在MATLAB平臺(tái)上模擬出了原始的重疊峰譜線,并作為后續(xù)算法的研究對(duì)象,為誤差分析依據(jù)。2、針對(duì)期望最大值法在解決重疊峰分解問題時(shí)計(jì)算時(shí)間過長(zhǎng)的缺陷,提出一種快速算法,并且有效的利用該算法完成了重疊峰的分解任務(wù)。3、簡(jiǎn)單闡述了遺傳算法的優(yōu)越性并對(duì)重疊峰進(jìn)行分解:將待求的解集空間和解集空間中的解看作遺傳算法中的染色體和基因,結(jié)合遺傳算法工具箱,通過一系列的選擇及遺傳操作后,在全局模式找出最符合原始重疊峰的參數(shù)組合。4、尋找粒子群算法與重疊峰分解之間的聯(lián)系,完成初始參數(shù)的討論、適應(yīng)度函數(shù)的選擇、粒子評(píng)價(jià)、粒子的位置更新以及個(gè)體極值與全局極值的更新等工作,最終得到良好的分解效果;其次運(yùn)用該算法完成232Th和226Ra核素的實(shí)際重疊峰的分解任務(wù)。5、運(yùn)用期望最大值算法、遺傳算法及粒子群算法均實(shí)現(xiàn)了雙峰重疊峰及三峰重疊峰的分解工作。雙峰重疊峰分解中:當(dāng)初始參數(shù)均未知時(shí)三種方法可以分解的最小的峰位間距分別為17KeV、13KeV和5KeV;當(dāng)初始峰位已知時(shí)最大期望值法可以完成8KeV道值窗的重疊峰分解,且遺傳算法的權(quán)重和標(biāo)準(zhǔn)偏差的誤差均有所降低。在三峰重疊峰分解中:利用峰位與偏差的關(guān)聯(lián)性可以提高最大期望值法和遺傳算法的參數(shù)精度;對(duì)于粒子群算法,即使在初始參數(shù)均未知時(shí),也可以完成185KeV、195KeV及203KeV三峰位的分解,且分解結(jié)果較好。理論上本文研究的三種算法均可以實(shí)現(xiàn)能量相近的多峰重疊峰的分解工作,且解效果較好,對(duì)于實(shí)際的分辨率低的重疊峰分解問題具有一定的參考價(jià)值。
[Abstract]:While nuclear science and technology bring convenient service and clean energy to human life, people begin to pay more and more attention to the influence of nuclear radiation on environment and body. Usually the radioactive material in the radiation environment releases 緯-ray. Through the measurement of 緯-ray the types of nuclides in the radioactive material can be understood and the content and activity of the radionuclides can be judged. However, the influence of environmental or other interference rays will lead to frequent overlapping of spectral signals in real measurement. The commonly used 緯 -ray detector, Nai (Tl) detector, is widely used because of its high detection efficiency, convenient maintenance and moderate price, but its resolution ability to overlapping peaks with similar energy is not strong. This makes the decomposition of overlapping peaks a difficult problem in spectral analysis. Therefore, based on the statistical distribution of 緯 energy spectrum, the expected maximum value method, genetic algorithm and particle swarm optimization algorithm are used to decompose the overlapped peaks on MATLAB platform. The main work and results are as follows: firstly, the energy spectrum and its mathematical model are discussed. Then, according to the statistical fluctuation characteristics of energy spectrum, the original overlapping peak lines are simulated on the MATLAB platform, which is regarded as the research object of the subsequent algorithms. In order to solve the problem of overlapping peak decomposition, a fast algorithm is proposed to solve the problem of overlapping peak decomposition. And effectively use the algorithm to complete the overlapping peak decomposition task. 3, the advantages of genetic algorithm and decomposition of overlapping peaks: the solution set space and solution set space solution space as the chromosomes and genes in genetic algorithm. Combined with genetic algorithm toolbox, after a series of selection and genetic operation, the parameter combination. 4 is found out in the global mode, and the relation between particle swarm optimization algorithm and overlapping peak decomposition is found, and the discussion of initial parameters is completed. The selection of fitness function, particle evaluation, the update of particle position and the update of individual extremum and global extremum, etc., finally get a good decomposition effect. Secondly, the algorithm is used to decompose the actual overlapping peaks of 232Th and 226Ra. The expected maximum value algorithm, genetic algorithm and particle swarm optimization algorithm are used to decompose the overlapping peaks of two peaks and three peaks. When the initial parameters are unknown, the minimum peak spacing of the three methods is 17KeV 13KeV and 5KeV, respectively. When the initial peak position is known, the maximum expectation value method can complete the overlapping peak decomposition of the 8KeV channel window. The weight and the error of standard deviation of genetic algorithm are reduced. In the three-peak overlapping peak decomposition, the maximum expectation value method and genetic algorithm can be improved by using the correlation between peak position and deviation. For particle swarm optimization, even when the initial parameters are unknown, the decomposition of the three peaks of 185KeV and 203KeV can be completed. The results of decomposition are good. Theoretically, the three algorithms studied in this paper can realize the decomposition of multi-peak overlapped peaks with similar energy, and the results are good, which has a certain reference value for the practical overlapping peak decomposition problem with low resolution.
【學(xué)位授予單位】：成都理工大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TL81;TP18
本文編號(hào)：2202697
[Abstract]:While nuclear science and technology bring convenient service and clean energy to human life, people begin to pay more and more attention to the influence of nuclear radiation on environment and body. Usually the radioactive material in the radiation environment releases 緯-ray. Through the measurement of 緯-ray the types of nuclides in the radioactive material can be understood and the content and activity of the radionuclides can be judged. However, the influence of environmental or other interference rays will lead to frequent overlapping of spectral signals in real measurement. The commonly used 緯 -ray detector, Nai (Tl) detector, is widely used because of its high detection efficiency, convenient maintenance and moderate price, but its resolution ability to overlapping peaks with similar energy is not strong. This makes the decomposition of overlapping peaks a difficult problem in spectral analysis. Therefore, based on the statistical distribution of 緯 energy spectrum, the expected maximum value method, genetic algorithm and particle swarm optimization algorithm are used to decompose the overlapped peaks on MATLAB platform. The main work and results are as follows: firstly, the energy spectrum and its mathematical model are discussed. Then, according to the statistical fluctuation characteristics of energy spectrum, the original overlapping peak lines are simulated on the MATLAB platform, which is regarded as the research object of the subsequent algorithms. In order to solve the problem of overlapping peak decomposition, a fast algorithm is proposed to solve the problem of overlapping peak decomposition. And effectively use the algorithm to complete the overlapping peak decomposition task. 3, the advantages of genetic algorithm and decomposition of overlapping peaks: the solution set space and solution set space solution space as the chromosomes and genes in genetic algorithm. Combined with genetic algorithm toolbox, after a series of selection and genetic operation, the parameter combination. 4 is found out in the global mode, and the relation between particle swarm optimization algorithm and overlapping peak decomposition is found, and the discussion of initial parameters is completed. The selection of fitness function, particle evaluation, the update of particle position and the update of individual extremum and global extremum, etc., finally get a good decomposition effect. Secondly, the algorithm is used to decompose the actual overlapping peaks of 232Th and 226Ra. The expected maximum value algorithm, genetic algorithm and particle swarm optimization algorithm are used to decompose the overlapping peaks of two peaks and three peaks. When the initial parameters are unknown, the minimum peak spacing of the three methods is 17KeV 13KeV and 5KeV, respectively. When the initial peak position is known, the maximum expectation value method can complete the overlapping peak decomposition of the 8KeV channel window. The weight and the error of standard deviation of genetic algorithm are reduced. In the three-peak overlapping peak decomposition, the maximum expectation value method and genetic algorithm can be improved by using the correlation between peak position and deviation. For particle swarm optimization, even when the initial parameters are unknown, the decomposition of the three peaks of 185KeV and 203KeV can be completed. The results of decomposition are good. Theoretically, the three algorithms studied in this paper can realize the decomposition of multi-peak overlapped peaks with similar energy, and the results are good, which has a certain reference value for the practical overlapping peak decomposition problem with low resolution.
【學(xué)位授予單位】：成都理工大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TL81;TP18
【參考文獻(xiàn)】
相關(guān)期刊論文 前10條
1 王麗;馮燕;;基于粒子群優(yōu)化的圖像稀疏分解算法研究[J];計(jì)算機(jī)仿真;2015年11期
2 黃洪全;丁衛(wèi)撐;龔迪琛;方方;;基于統(tǒng)計(jì)遺傳算法的X射線熒光重疊峰分解[J];光譜學(xué)與光譜分析;2015年08期
3 曾立暉;王南萍;田貴;;NaI(Tl)伽瑪能譜全譜數(shù)據(jù)高斯分解軟件的實(shí)現(xiàn)與應(yīng)用[J];核電子學(xué)與探測(cè)技術(shù);2011年12期
4 沈晴;徐筠;康慧珍;卜建平;郭文剛;;通過數(shù)學(xué)方法進(jìn)行重疊峰分解的國(guó)內(nèi)外研究現(xiàn)狀綜述[J];價(jià)值工程;2011年04期
5 黃洪全;何子述;方方;龔迪琛;丁衛(wèi)撐;;多重譜峰的分解方法[J];原子能科學(xué)技術(shù);2010年09期
6 黃洪全;方方;龔迪琛;丁衛(wèi)撐;;GMM模型在核能譜平滑濾波中的應(yīng)用[J];核技術(shù);2010年05期
7 秦小輝;;Matlab遺傳算法工具箱在物流網(wǎng)絡(luò)設(shè)計(jì)中的應(yīng)用[J];鐵道運(yùn)輸與經(jīng)濟(jì);2009年01期
8 杜祥琬;;讓核技術(shù)為國(guó)家可持續(xù)發(fā)展再創(chuàng)輝煌[J];中國(guó)工程科學(xué);2008年01期
9 裴少英;王南萍;;航空γ能譜全譜數(shù)據(jù)的高斯分解方法[J];科技導(dǎo)報(bào);2006年11期
10 屈國(guó)普,凌球,郭蘭英,趙立宏,陳堅(jiān)禎;NaI(Tl)閃爍譜儀譜漂移原因分析[J];南華大學(xué)學(xué)報(bào)(自然科學(xué)版);2005年01期
相關(guān)博士學(xué)位論文 前1條
1 劉永剛;γ能譜譜數(shù)據(jù)分解方法研究[D];中國(guó)地質(zhì)大學(xué)（北京）;2011年
相關(guān)碩士學(xué)位論文 前5條
1 魏晉軍;粒子群優(yōu)化算法的改進(jìn)及應(yīng)用[D];太原理工大學(xué);2015年
2 居鳳霞;粒子群優(yōu)化算法的改進(jìn)及應(yīng)用[D];華南理工大學(xué);2014年
3 張宏?yáng)|;EM算法及其應(yīng)用[D];山東大學(xué);2014年
4 曹道友;基于改進(jìn)遺傳算法的應(yīng)用研究[D];安徽大學(xué);2010年
5 岳佳;基于EM算法的模型聚類的研究及應(yīng)用[D];江南大學(xué);2007年
，本文編號(hào)：2202697
本文鏈接：http://sikaile.net/kejilunwen/zidonghuakongzhilunwen/2202697.html
最近更新
教材專著
