轻量级神经网络模型在核脉冲参数预测中的应用研究

唐琳; 周爽; 廖先莉; 李波

您当前的位置：

首页 >

文章列表页 >

轻量级神经网络模型在核脉冲参数预测中的应用研究

核物理、交叉学科研究 | 更新时间：2024-11-27

- 轻量级神经网络模型在核脉冲参数预测中的应用研究
- Application of lightweight neural network models for nuclear pulse parameter prediction
- 核技术 2024年文章编号：XXXXXX
- 作者机构：
  
  1.成都大学电子信息与电气工程学院成都610106
  2.成都理工大学核技术与自动化工程学院成都610059
  3.南洋理工大学电气与电子工程学院新加坡639798
- 作者简介：
  
  唐琳，女，1988年出生，2019年于成都理工大学获博士学位，副教授，新加坡南洋理工大学访问学者，研究领域为核辐射探测与电子学
  李波，E-mail：libo@cdu.edu.cn
- 基金信息：
  
  国家自然科学青年基金(12305214);四川省自然科学青年基金项目(2023NSFSC1366);安徽大学农业生态大数据分析与应用国家工程研究中心开放研究基金(AE202209)
- DOI：
  中图分类号： O657.34
- 网络出版日期：2024-11-27，
  
  收稿日期：2024-06-13，
  
  修回日期：2024-07-16，
- 稿件说明：
移动端阅览
唐琳, 周爽, 廖先莉, 等. 轻量级神经网络模型在核脉冲参数预测中的应用研究[J/OL]. 核技术, 2024,XXXXXX

TANG Lin, ZHOU Shuang, LIAO Xianli, et al. Application of lightweight neural network models for nuclear pulse parameter prediction. [J/OL]. NUCLEAR TECHNIQUES, 2024,XXXXXX
唐琳, 周爽, 廖先莉, 等. 轻量级神经网络模型在核脉冲参数预测中的应用研究[J/OL]. 核技术, 2024,XXXXXX DOI：

TANG Lin, ZHOU Shuang, LIAO Xianli, et al. Application of lightweight neural network models for nuclear pulse parameter prediction. [J/OL]. NUCLEAR TECHNIQUES, 2024,XXXXXX DOI：

摘要

在核辐射测量中，由于测量系统本身以及测量环境的干扰，脉冲畸变不可避免。如果无法准确估计这类脉冲的参数，将会降低能谱的分辨性能。本文提出将6种轻量级神经网络模型用于畸变脉冲的参数预测的方法，预测对象包括脉冲幅度参数和畸变时间参数。以预定义的数学模型生成的畸变脉冲为基础数据集，经数字三角成形得到模型训练所需的数据集。模型性能评估结果表明：传统的数字成形法虽然在时间参数预测上具有绝对优势，但在幅度参数预测中却受到脉冲畸变的限制，无法获得准确的幅度预测结果。在对包括UNet在内的6种神经网络模型进行参数预测性能评估时，UNet模型在测试集上获得了最低的相对误差，其中幅度参数的相对误差约为0.57%，时间参数的相对误差为3.51%。在信噪比实验中，探讨噪声对具有出色的抗噪性能参数预测性能的UNet模型与CNN-LSTM模型的影响，进一步证明了UNet模型优秀的抗噪性能。

Abstract

Background

In nuclear radiation measurement

pulse distortion is inevitable due to the interference of the measurement system itself and the measurement environment. If the parameters of such pulses cannot be accurately estimated

the resolution performance of the energy spectrum will be reduced.

Purpose

This study aims to accurately estimate the height of distorted pulses using neural network model.

Methods

Firstly

six lightweight neural network models

i.e.

LeNet5

LSTM

GRU

UNet

CNN-GRU

CNN-LSTM

were applied to parameter prediction of distorted nuclear pulses

including pulse amplitude parameters and distortion time parameters. Then

based on the distorted pulses generated by predefined mathematical models

the dataset required for model training was obtained through digital triangulation shaping. Finally

parameter prediction performances of those neural network models on test set with additional white noise

Gaussian noise and flicker noise were compared with each other

as well as with the traditional digital forming method.

Results

When evaluating the parameter prediction performance of six neural network models

the UNet model achieves the lowest relative error on the test set

with a relative error of approximately 0.57% for amplitude parameters and 3.51% for time parameters. In the signal-to-noise ratio experiment

noise is superimposed on the test set to obtain noise test sets with different signal-to-noise ratios.

Conclusions

The results of this study show that the proposed models can achieve accurate estimation of the parameters of distorted pulses.

关键词

深度学习轻量级神经网络UNet核脉冲参数

Keywords

Deep learningLightweight neural networkUNetNuclear pulse parameters

references

荆薇, 邓玉福, 孟德川, 等. EDXRF法测量饲料添加剂中钠元素含量[J]. 核技术, 2022, 45(4): 040201. DOI: 10.11889/j.0253-3219.2022.hjs.45.040201http://dx.doi.org/10.11889/j.0253-3219.2022.hjs.45.040201.

JING Wei, DENG Yufu, MENG Dechuan, et al.Quantitative detection of sodium content in feed additive sodium chloride based on EDXRF method[J]. Nuclear Techniques, 2022, 45(4): 040201. DOI: 10.11889/j.0253-3219.2022.hjs.45.040201http://dx.doi.org/10.11889/j.0253-3219.2022.hjs.45.040201.

张磊. γ能谱符合相加效应修正方法及影响因素[J]. 核技术, 2022, 45(6): 060401. DOI: 10.11889/j.0253-3219.2022.hjs.45.060401http://dx.doi.org/10.11889/j.0253-3219.2022.hjs.45.060401.

ZHANG Lei. Coincidence summing correction methods and influence factors of γ spectrum[J]. Nuclear Techniques, 2022, 45(6): 060401. DOI: 10.11889/j.0253-3219.2022.hjs.45.060401http://dx.doi.org/10.11889/j.0253-3219.2022.hjs.45.060401.

李雷鸣, 江书益, 陈珍平, 等. 核反应堆结构材料网格精细活化分析研究[J]. 核技术, 2022, 45(8): 080601. DOI: 10.11889/j.0253-3219.2022.hjs.45.080601http://dx.doi.org/10.11889/j.0253-3219.2022.hjs.45.080601.

LI Leiming, JIANG Shuyi, CHEN Zhenping, et al. Mesh-based activation analysis for structural materials in nuclear reactor[J]. Nuclear Techniques, 2022, 45(8): 080601. DOI: 10.11889/j.0253-3219.2022.hjs.45.080601http://dx.doi.org/10.11889/j.0253-3219.2022.hjs.45.080601.

Nader M N, Fleming D E B. Assessment of alternative methods for analyzing X-ray fluorescence spectra[J]. Applied Radiation and Isotopes, 2019, 146: 133–138. DOI: 10.1016/j.apradiso.2019.01.033http://dx.doi.org/10.1016/j.apradiso.2019.01.033.

Radeka V, Karlovac N. Least-square-error amplitude measurement of pulse signals in presence of noise[J]. Nuclear Instruments and Methods, 1967, 52(1): 86–92. DOI: 10.1016/0029-554X(67)90561-7http://dx.doi.org/10.1016/0029-554X(67)90561-7.

Jordanov V T. Pile-up rejection using pulse-shape discrimination[C]. 2018 IEEE Nuclear Science Symposium and Medical Imaging Conference Proceedings (NSS/MIC). November10-17, 2018. Sydney, Australia. IEEE, 2018. DOI: 10.1109/nssmic.2018. 8824551http://dx.doi.org/10.1109/nssmic.2018.8824551.

Statham P J. Pile-up correction for improved accuracy and speed of X-ray analysis[J]. Microchimica Acta, 2006, 155(1): 289–294. DOI: 10.1007/s00604-006-0558-1http://dx.doi.org/10.1007/s00604-006-0558-1.

钱云琛, 周建斌, 刘权卫, 等. 基于三角函数的核信号类高斯成形算法研究[J]. 核技术, 2024, 47(6): 060401. DOI: 10.11889/j.0253-3219.2024.hjs.47.060401http://dx.doi.org/10.11889/j.0253-3219.2024.hjs.47.060401.

QIAN Yunchen, ZHOU Jianbin, LIU Quanwei, et al. Trigonometric functions-based gaussian-like shaping algorithms for nuclear signals[J]. Nuclear Techniques, 2024, 47(6): 060401. DOI: 10.11889/j.0253-3219.2024.hjs.47.060401http://dx.doi.org/10.11889/j.0253-3219.2024.hjs.47.060401.

Tang L, Yu J, Zhou J B, et al. A new method for removing false peaks to obtain a precise X-ray spectrum[J]. Applied Radiation and Isotopes, 2018, 135: 171–176. DOI: 10.1016/j.apradiso.2018.01.033http://dx.doi.org/10.1016/j.apradiso.2018.01.033.

Tang L, Zhou J B, Fang F, et al. Counting-loss correction for X-ray spectra using the pulse-repairing method[J]. Journal of Synchrotron Radiation, 2018, 25(Pt 6): 1760–1767. DOI: 10.1107/S160057751801411Xhttp://dx.doi.org/10.1107/S160057751801411X.

Micieli D, Minniti T, Evans L M, et al. Accelerating Neutron Tomography experiments through Artificial Neural Network based reconstruction[J]. Scientific Reports, 2019, 9(1): 2450. DOI: 10.1038/s41598-019-38903-1http://dx.doi.org/10.1038/s41598-019-38903-1.

Akhavanallaf A, Shiri I, Arabi H, et al. Whole-body voxel-based internal dosimetry using deep learning[J]. European Journal of Nuclear Medicine and Molecular Imaging, 2021, 48(3): 670–682. DOI: 10.1007/s00259-020-05013-4http://dx.doi.org/10.1007/s00259-020-05013-4.

唐琳, 李勇, 唐羽锋, 等. 基于深度学习的LSTM模型在X荧光光谱中的应用[J]. 核技术, 2023, 46(7): 070502. DOI: 10.11889/j.0253-3219.2023.hjs.46.070502http://dx.doi.org/10.11889/j.0253-3219.2023.hjs.46.070502.

TANG Lin, LI Yong, TANG Yufeng, et al. Application of an LSTM model based on deep learning through X-ray fluorescence spectroscopy[J]. Nuclear Techniques, 2023, 46(7): 070502. DOI: 10.11889/j.0253-3219.2023.hjs.46.070502http://dx.doi.org/10.11889/j.0253-3219.2023.hjs.46.070502.

Regadío A, Esteban L, Sánchez-Prieto S. Unfolding using deep learning and its application on pulse height analysis and pile-up management[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2021, 1005: 165403. DOI: 10.1016/j.nima.2021.165403http://dx.doi.org/10.1016/j.nima.2021.165403.

Fu C, Di Fulvio A, Clarke S D, et al. Artificial neural network algorithms for pulse shape discrimination and recovery of piled-up pulses in organic scintillators[J]. Annals of Nuclear Energy, 2018, 120: 410–421. DOI: 10.1016/j.anucene.2018.05.054http://dx.doi.org/10.1016/j.anucene.2018.05.054.

Ma X K, Huang H Q, Huang B R, et al. X-ray spectra correction based on deep learning CNN-LSTM model[J]. Measurement, 2022, 199: 111510. DOI: 10.1016/j.measurement.2022.111510http://dx.doi.org/10.1016/j.measurement.2022.111510.

Tang L, Shi K B, Shen H T, et al. Application of transformer model in peak correction of X-ray fluorescence spectra[J]. IEEE Transactions on Nuclear Science, 2023, 70(11): 2479–2489. DOI: 10.1109/TNS.2023.3320807http://dx.doi.org/10.1109/TNS.2023.3320807.

Tang L, Ma X K, Shi K B, et al. A method for correcting characteristic X-ray net peak count from drifted shadow peak[J]. Nuclear Science and Techniques, 2023, 34(11): 175. DOI: 10.1007/s41365-023-01333-whttp://dx.doi.org/10.1007/s41365-023-01333-w.

Chen Y H, Chen X M, Zhang X D, et al. Study of n-γ discrimination in low energy range (above 40 keVee) by charge comparison method with a BC501A liquid scintillation detector[J]. Chinese Physics C, 2014, 38(3): 036001. DOI: 10.1088/1674-1137/38/3/036001http://dx.doi.org/10.1088/1674-1137/38/3/036001.

Nakhostin M, Walker P M. Application of digital zero-crossing technique for neutron–gamma discrimination in liquid organic scintillation detectors[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2010, 621(1–3): 498–501. DOI: 10.1016/j.nima.2010.06.252http://dx.doi.org/10.1016/j.nima.2010.06.252.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

轻量级神经网络模型在脉冲参数预测中的应用研究

基于改进UNet模型的截断脉冲高度估计器

嵌入位置编码的多头注意力机制在堆积脉冲幅度估计中的应用

基于深度学习的LSTM模型在X荧光光谱中的应用