基于神经网络的人工放射性气溶胶中氡子体扣除算法  

Radon daughter subtraction algorithm for artificial radioactive aerosol based on neural network

作　　者：陈立[1] 顾民[2] 曾国强[2] 葛良全[2] 杨坤 肖明 CHEN Li GU Min ZENG Guoqiang GE Liangquan YANG Kun XIAO Ming(Radiation Environmental Management and Monitoring Center of Sichuan Province, Chengdu 611139, China Key Laboratory of Earth Science Nuclear Technology of Sichuan Province, Chengdu University of Technology, Chengdu 610059, China CGNPC Jiuyuan （Chengdu） Technology Co. Ltd., Chengdu 610200, China)

机构地区：[1]四川省辐射环境管理监测中心站,成都611139 [2]成都理工大学地学核技术四川省重点实验室,成都610059 [3]中广核久源(成都)科技有限公司,成都610200

出　　处：《核技术》2017年第9期41-45,共5页Nuclear Techniques

基　　金：国家自然科学基金(No.41474159);国家863计划项目(No.2012AA061803);四川省科技厅青年基金项目(No.2015JQ0035)资助~~

摘　　要：介绍和分析了人工放射性气溶胶在线监测仪氡子体扣除算法中比例系数扣除法,现有算法存在分类粗糙、扣除准确度不高以及适应性不强等不足。为进一步提高扣除的准确度,降低检测限,提出了利用聚类分析先对谱线进行分类,然后在每个类中利用神经网络进行计算,最后进行扣除的方法。测试结果证明了聚类分析和神经网络扣除方法均能明显降低人工放射性气溶胶在线监测仪的检测限。Background： The proportion subtraction method used in radon daughters subtraction algorithm for continuous artificial radioactive aerosol monitor has disadvantages such as rough classfication, less accuracy and low adaptability. Purpose： This study aims to improve the accuracy of subtraction to reduce the detection limit. Methods：A novel algorithm is proposed by classifying the spectral lines through clustering analysis and then calculating each clustering using neural network. Experimental verifcation is performed to compare this method with the proportion subtraction method. Results： The results showed that the cluster analysis and neural network subtraction algorithm can reduce more than 20% of the detection limit for the continuous artificial radioactive aerosol monitor. Conclusion：The algorithm proposed in this paper is effective for subtracting radon daughters.

关 键 词：气溶胶 氡子体 聚类分析 神经网络 

分 类 号：TL99[核科学技术—核技术及应用]