基于机器学习算法的四角读出型塑料闪烁体探测器缪子定位与散射成像的模拟研究  

Simulation Study of Muon Positioning and Scattering Imaging of Plastic Scintillator Detector Based on Machine Learning

作　　者：何列 罗思远 张海峰 肖万成[1,2] 王晓冬 HE Lie;LUO Siyuan;ZHANG Haifeng;XIAO Wancheng;WANG Xiaodong(School of Nuclear Science and Technology,University of South China,Hengyang 421001,China;Key Laboratory of Advanced Nuclear Energy Technology Design and Safety,Ministry of Education,Hengyang 421001,China)

机构地区：[1]南华大学核科学技术学院,衡阳421200 [2]先进核能技术设计与安全教育部重点实验室,衡阳421200

出　　处：《核电子学与探测技术》2024年第2期253-261,共9页Nuclear Electronics & Detection Technology

基　　金：国家自然科学基金(No.12275120);湖湘青年英才(No.2022RC1202);湖南省自然科学基金(No.2021JJ20006);中国科学技术部(No.2020YFE0202001)。

摘　　要：宇宙线缪子是存在于自然界中一种穿透力极强的天然射线。目前主流的缪子探测器包括气体探测器和塑料闪烁体条探测器,虽然它们具有高精度,但成本较高且结构复杂。本研究基于四角读出的塑料闪烁体探测器,利用长短时记忆网络(LSTM)算法和密度聚类的DBSCAN算法的定位。使用Geant4仿真软件构建大面积(800 mm×800 mm)的四角读出塑料闪烁体缪子成像系统,对模拟数据实现缪子位置重建与成像结果。首先,采用LSTM算法作为探测器上的缪子定位方法,通过与实验中已获得的400 mm×400 mm探测器重建入射位置图像的结果比较以验证本研究模拟方法的可靠性,结果表明,模拟得到的位置分辨可达到厘米级。其次,利用DBSCAN算法对钨块构建的“U”“S”“C”三种形状模型的PoCA散射成像结果进行聚类优化,实现对钨块成像点与噪点的区分,使得成像结果更有区分度,为缪子探测器系统的构建提供新的思路与方向。Muon cosmic rays are highly penetrating background radiation present in the natural environment.Currently,the mainstream muon detectors include gas detectors and plastic scintillator strip detectors,known for their high precision but associated with high costs and complex structures.This study is based on the existing four-corner readout plastic scintillator detector in the laboratory.Through the application of the Long Short-Term Memory(LSTM)algorithm and Density-Based Spatial Clustering of Applications with Noise(DBSCAN)algorithm,and utilizing the Geant4 simulation software,we simulated the construction of a large-area(800 mm×800 mm)four-corner readout plastic scintillator muon imaging system for position reconstruction and imaging results.Firstly,we employed the LSTM algorithm as a method for muon localization on the detector.Through comparing the reconstructed incident position image results with those obtained from a 400 mm×400 mm detector in experiments,the reliability of the simulation methodology in this study was demonstrated.The position reconstruction results indicate that the simulated position resolution can reach the centimeter level.Secondly,we utilized the DBSCAN algorithm to perform clustering optimization on the PoCA scatter imaging results of“U”“S”and“C”shaped models constructed with tungsten blocks.This step achieved a clear distinction between imaging points and noise points related to the tungsten block,enhancing the precision of the imaging results.This provides new insights and directions for the construction of muon detector systems.

关 键 词：蒙特卡罗模拟 GEANT4 缪子成像 机器学习 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]