基于神经网络的中子多重性测量方法可行性研究

冯元威; 郑玉来; 李永; 刘超; 张连军; 黄喆; 郭文慧

doi:10.11884/HPLPB202638.250245

基于神经网络的中子多重性测量方法可行性研究

doi: 10.11884/HPLPB202638.250245

冯元威^{1, 2,},
郑玉来^{1, 2, ,},
李永^{1, 2},
刘超^{1, 2},
张连军^{1, 2},
黄喆^{1, 2},
郭文慧^{1, 2}

1.
中国原子能科学研究院核技术综合研究所，北京 102413
2.
IAEA核安保协作中心，北京 102413

详细信息

作者简介:
冯元威，fyw_caini@foxmail.com

通讯作者:
郑玉来，mczyl@sina.com。

中图分类号: O571.53
计量
- 文章访问数: 9
- HTML全文浏览量: 6
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2025-07-31
- 修回日期: 2025-12-29
- 录用日期: 2025-12-25
- 网络出版日期: 2026-01-21

Feasibility study on neutron multiplicity counting method based on neural network

Feng Yuanwei^{1, 2
,},
Zheng Yulai^{1, 2
, ,},
Li Yong^{1, 2},
Liu Chao^{1, 2},
Zhang Lianjun^{1, 2},
Huang Zhe^{1, 2},
Guo Wenhui^{1, 2}

1.
Department Of Nuclear Technology And Application, China Institute of Atomic Energy, Beijing 102413, China
2.
IAEA Collaborating Center for Nuclear Security Technologies, China Institute of Atomic Energy, Beijing 102413, China

摘要

摘要: 中子多重性测量技术作为无损检测领域的核心手段，在裂变材料（²³⁵U）质量测定中发挥关键作用，但其存在测量周期冗长、非理想条件下存在测量偏差等技术瓶颈。借助Geant4与MATLAB软件构建主动井型符合计数器（AWCC）仿真模型实现主动中子多重性测量全流程的高精度模拟。在此基础上，基于PyTorch框架构建反向传播神经网络（BPNN）、卷积神经网络（CNN）、长短期记忆网络（LSTM）三种神经网络对中子多重性分布数据进行分析研究。结果表明，相较于传统主动中子多重性测量方法，三种神经网络模型在测量精度与效率方面均展现出显著优势，能够有效降低测量误差、缩短测量时间。研究结果不仅证实了基于神经网络的中子多重性测量方法的可行性，为中子多重性探测向高效化、智能化方向发展提供了新的解决方案。
- 中子多重性 /
- 蒙特卡罗方法 /
- BPNN /
- CNN /
- LSTM /
- Geant4.
Abstract: Background
Neutron multiplicity measurement technology, as a core method in the field of non-destructive testing, plays a critical role in determining the mass of fissionable material (²³⁵U). However, it suffers from technical bottlenecks such as prolonged measurement cycles and measurement deviations under non-ideal conditions.
Purpose
This paper aims to explore feasible pathways for integrating neutron multiplicity measurement methods with neural network technology. The goal is to provide new research perspectives for advancing neutron multiplicity measurement technology toward greater efficiency and intelligence.
Methods
Leveraging Geant4 and MATLAB software, an Active Well Coincidence Counter (AWCC) simulation model is constructed to achieve high-precision simulation of the entire active neutron multiplicity measurement process. Building upon this, three neural networks—Backpropagation Neural Network (BPNN), Convolutional Neural Network (CNN), and Long Short-Term Memory network (LSTM)—are developed using the PyTorch framework to analyze and investigate neutron multiplicity distribution data.
Results
Compared with traditional calculation methods based on the active neutron multiplicity equation, neural network models represented by CNN and LSTM demonstrate significant advantages in measurement accuracy and efficiency. Specifically, in terms of relative error metrics, neural network models can reduce errors to lower levels; in the time dimension of measurement, they substantially shorten data processing cycles, effectively overcoming the timeliness constraints inherent to traditional approaches.
Conclusions
This achievement fully validates the theoretical feasibility and technical superiority of the neural network-based neutron multiplicity measurement approach, providing a novel solution for advancing neutron multiplicity detection toward greater efficiency and intelligence. Subsequent work will enhance the adaptability and noise resistance of neural network models for complex data by increasing simulation scenario complexity and introducing diversified factors such as noise interference and geometric variations. Meanwhile, building upon simulation studies, physical experimental validation will be conducted using AWCC instrumentation to drive the transition of neural network-based neutron multiplicity measurement technology from simulation to engineering application.
- neutron multiplicity counting /
- Monte Carlo Method /
- BPNN /
- CNN /
- LSTM /
- Geant4.

HTML全文

图 1 BPNN结构图^[16]

Figure 1. BPNN structure diagram^[16]

下载: 全尺寸图片幻灯片

图 2 CNN结构图(以LeNet5为例)^[16]

Figure 2. CNN structure diagram (taking LeNet5 as an example)^[16]

下载: 全尺寸图片幻灯片

图 3 LSTM结构图^[16]

Figure 3. LSTM structure diagram^[16]

下载: 全尺寸图片幻灯片

图 4 AWCC结构模型^[21]

Figure 4. AWCC structural model^[21]

下载: 全尺寸图片幻灯片

图 5 Geant4中构建的AWCC模型

Figure 5. AWCC model built in Geant4

下载: 全尺寸图片幻灯片

图 6 主动中子多重性测量计算结果

Figure 6. Calculation results of active neutron multiplicity measurement

下载: 全尺寸图片幻灯片

图 7 神经网络计算²³⁵U质量误差

Figure 7. The ²³⁵U mass error calculated by neural network

下载: 全尺寸图片幻灯片

表 1 AWCC计数器特征

Table 1. AWCC counter features

pressure of ³He tube/MPa	³He tube size (height×diameter)/cm	³He tube quantity	sample cavity size (height×diameter)/cm	counter size (height×diameter)/cm	AmLi source holder (height×diameter)/cm
0.4	50.8×2.54	42	20.6×22.9	50.8×49.3	5.7×3.2

下载: 导出CSV

表 2 部分模拟计算结果

Table 2. Part of simulation calculation results

total mass/g	²³⁵U enrichment/%	double coincidence rate D/s⁻¹	triple coincidence rate T/s⁻¹	prediction mass/g	quality deviation/%
3079.35	100(²³⁵U 3079.35 g)	1048.1	844.9	3330.11	8.14
3079.35	95(²³⁵U 2925.38 g)	963.7	733.3	3089.95	5.61
3079.35	90(²³⁵U 2771.41 g)	887.1	629.2	2816.62	1.63
3079.35	85(²³⁵U 2617.45 g)	812.9	550.5	2682.63	2.49
3079.35	80(²³⁵U 2463.48 g)	741.7	478.2	2449.62	0.56
3079.35	70(²³⁵U 2115.54 g)	625.3	351.1	2086.87	1.36
3079.35	60(²³⁵U 1847.61 g)	521.5	256.7	1752.97	5.12

下载: 导出CSV

表 3 各方法计算计算²³⁵U质量误差结果

Table 3. The ²³⁵U mass error result calculated by each method

method	measurement time/s	mean of relative error/%	median of relative error/%	max of relative error/%
active neutron multiplicity analysis	1 000	3.32	2.25	10.91
LSTM	1 000	0.36	0.37	0.77
LSTM	100	0.83	0.70	3.31
LSTM	10	2.50	2.02	11.48
LSTM	1	6.44	5.21	38.27
1D-CNN	1 000	0.72	0.58	1.89
1D-CNN	100	1.78	1.83	6.16
1D-CNN	10	2.41	2.03	11.08
1D-CNN	1	6.49	5.25	38.40
2D-CNN	1 000	0.44	0.42	1.05
2D-CNN	100	1.75	1.74	6.21
2D-CNN	10	2.47	1.92	13.89
BPNN	1 000	1.81	1.72	4.90
BPNN	100	2.24	1.98	9.23
BPNN	10	3.28	2.78	16.49
BPNN	1	7.58	6.53	40.53

下载: 导出CSV

参考文献(24)

[1]	Shin T H, Di Fulvio A, Clarke S D, et al. Prompt fission neutron anisotropy in low-multiplying subcritical plutonium metal assemblies[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2019, 915: 110-115. doi: 10.1016/j.nima.2018.09.085
[2]	Nakae L F, Chapline G F, Glenn A M, et al. The use of fast neutron detection for materials accountability[J]. International Journal of Modern Physics: Conference Series, 2014, 27: 1460140. doi: 10.1142/S2010194514601409
[3]	易凌帆. 中子多重性测量分析方法仿真研究[D]. 衡阳: 南华大学, 2016 Yi Lingfan. Simulation study of neutron multiplicity measurement and analysis methods[D]. Hengyang: University of South China, 2016
[4]	Di Fulvio A, Shin T H, Jordan T, et al. Passive assay of plutonium metal plates using a fast-neutron multiplicity counter[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2017, 855: 92-101.
[5]	Di Fulvio A, Shin T H, Basley A, et al. Fast-neutron multiplicity counter for active measurements of uranium oxide certified material[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2018, 907: 248-257. doi: 10.1016/j.nima.2018.05.049
[6]	Hou Suxia, Luo Jijun. Improvement of plutonium sample property measurement based on fast neutron multiplicity counting[J]. Annals of Nuclear Energy, 2021, 156: 108219. doi: 10.1016/j.anucene.2021.108219
[7]	Heger G, Dubi C, Ocherashvili A, et al. Identifying neutron shielding in neutron multiplicity counting[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2018, 901: 40-45. doi: 10.1016/j.nima.2018.05.058
[8]	Langner D G, Stewart J E, Pickrell M M, et al. Application guide to neutron multiplicity counting[R]. LA-13422-M, 1998.
[9]	刘晓波, 胡倩, 肖建国. 铀材料的中子多重性测量分析方法[J]. 核动力工程, 2009, 30(1): 74-77 Liu Xiaobo, Hu Qian, Xiao Jianguo. Neutron multiplicity counting and analysis for uranium material[J]. Nuclear Power Engineering, 2009, 30(1): 74-77
[10]	Ming Xingchen, Zhang Hongfei, Xu Ruirui, et al. Nuclear mass based on the multi-task learning neural network method[J]. Nuclear Science and Techniques, 2022, 33: 48. doi: 10.1007/s41365-022-01031-z
[11]	韦子豪, 王端, 王东东, 等. 神经网络-遗传复合算法在压水堆堆芯换料设计中的应用[J]. 原子能科学技术, 2020, 54(5): 825-834 doi: 10.7538/yzk.2019.youxian.0788 Wei Zihao, Wang Duan, Wang Dongdong, et al. Application of neural network-genetic composite algorithm in core refueling design for PWR[J]. Atomic Energy Science and Technology, 2020, 54(5): 825-834 doi: 10.7538/yzk.2019.youxian.0788
[12]	Li Yongyi, Zhang Fan, Su Jun. Improvement of the Bayesian neural network to study the photoneutron yield cross sections[J]. Nuclear Science and Techniques, 2022, 33: 135. doi: 10.1007/s41365-022-01131-w
[13]	Shang Tianshuai, Li Jian, Niu Zhongming. Prediction of nuclear charge density distribution with feedback neural network[J]. Nuclear Science and Techniques, 2022, 33: 153. doi: 10.1007/s41365-022-01140-9
[14]	Wang Zi’ao, Pei Junchen, Liu Yue, et al. Bayesian evaluation of incomplete fission yields[J]. Physical Review Letters, 2019, 123: 122501. doi: 10.1103/PhysRevLett.123.122501
[15]	黎素芬, 张全虎, 弟宇鸣, 等. 有源中子多重性测量计算机模拟研究[C]//第十六届全国核电子学与核探测技术学术年会论文集(下册). 2012: 654-661 Li Sufen, Zhang Quanhu, Di Yuming, et al. Computer simulation research of active neutron multiplicity counting and analysis[C]//Proceedings of the 16th National Academic Annual Conference on Nuclear Electronics and Nuclear Detection Technology. 2012: 654-661
[16]	Zhang A, Lipton Z C, Li Mu, et al. Dive into deep learning[M]. Cambridge: Cambridge University Press, 2023.
[17]	Gu Jiuxiang, Wang Zhenhua, Kuen J, et al. Recent advances in convolutional neural networks[J]. Pattern Recognition, 2018, 77: 354-377. doi: 10.1016/j.patcog.2017.10.013
[18]	杨丽, 吴雨茜, 王俊丽, 等. 循环神经网络研究综述[J]. 计算机应用, 2018, 38(s2): 1-6,26 doi: 10.11896/jsjkx.220600007 Yang Li, Wu Yuxi, Wang Junli, et al. Research on recurrent neural network[J]. Journal of Computer Applications, 2018, 38(s2): 1-6,26 doi: 10.11896/jsjkx.220600007
[19]	Agostinelli S, Allison J, Amako K, et al. GEANT4—a simulation toolkit[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2003, 506(3): 250-303.
[20]	Menlove H O. Description and operation manual for the active well coincidence counter[R]. LA-7823-M, 1979.
[21]	Canberra JCC-71 product description, Canberra, measurement solutions for nuclear safety and security[EB/OL]. 2010. http://www.canberra.com/products/715.asp.
[22]	李永明, 王亮, 陈想林, 等. ²⁵²Cf自发裂变中子发射率符合测量的回归分析[J]. 物理学报, 2018, 67: 242901 doi: 10.7498/aps.67.20181073 Li Yongming, Wang Liang, Chen Xianglin, et al. Regression analysis of coincidence measurements for determinating the neutron emission rate of ²⁵²Cf spontaneous fission[J]. Acta Physica Sinica, 2018, 67: 242901 doi: 10.7498/aps.67.20181073
[23]	Tagziria H, Looman M. The ideal neutron energy spectrum of ²⁴¹AmLi(α, n)¹⁰B sources[J]. Applied Radiation and Isotopes, 2012, 70(10): 2395-2402. doi: 10.1016/j.apradiso.2012.07.008
[24]	朱剑钰, 李瑞, 黄孟, 等. 用时序探测事件模拟提升中子多重性计算效率[J]. 强激光与粒子束, 2018, 30: 026003 doi: 10.11884/HPLPB201830.170256 Zhu Jianyu, Li Rui, Huang Meng, et al. Improving calculation efficiency of neutron multiplicity counting by sequential detection events simulation[J]. High Power Laser and Particle Beams, 2018, 30: 026003 doi: 10.11884/HPLPB201830.170256