登录    |    注册

您好,欢迎来到中国测试科技资讯平台!

首页> 《中国测试》期刊 >本期导读>低能过滤X射线神经网络解谱方法研究

低能过滤X射线神经网络解谱方法研究

349    2021-02-07

¥0.50

全文售价

作者:滕忠斌, 宋明哲, 倪宁, 魏可新, 刘蕴韬

作者单位:中国原子能科学研究院计量测试部,北京 102413


关键词:低能过滤X射线;脉冲幅度谱解谱;PIPS探测器;效率刻度;人工神经网络


摘要:

利用神经网络算法实现对PIPS探测器测量得到的低能过滤X射线脉冲幅度谱的快速解谱。首先根据PIPS探测器的计算机断层扫描图像,在MCNP5中建立该探测器的蒙特卡罗(MC)模型。并通过实验、MC效率刻度以及能谱展宽,对该探测器模型进行验证。之后计算PIPS探测器对单能光子(5~30 keV)的响应函数,并将其作为单层线性神经网络的训练数据。使用训练后的神经网络和GRV_MC33程序分别对N10~N30和L10~L30辐射质的X射线脉冲幅度谱进行解谱。结果表明:除N25和N30辐射质外,二者解谱结果相符较好。其解谱结果的差异可能来源于探测器响应函数和GRV_MC33程序解谱方法的不确定度。训练好的神经网络可被移植到微型计算机中,帮助校准实验室实现对低能过滤X射线脉冲幅度谱的快速解谱。


Study on unfolding method of low-energy filtered X-ray spectrum using artificial neural network
TENG Zhongbin, SONG Mingzhe, NI Ning, WEI Kexin, LIU Yuntao
Radiation Metrology Department of China Institute of Atomic Energy, Beijing 102413, China
Abstract: The artificial neural network was used to unfold the low-energy filtered X-ray pulse height spectra measured by a PIPS detector efficiently. Based on the computed tomography image of the PIPS detector, the Monte Carlo (MC) model of the PIPS detector was built in MCNP5. By performing the experiments efficiency calibration, MC efficiency calibration and the comparison of the measured and simulated pulse height spectrum, the MC model of the detector was verified. Then the MC model was used to calculate the response functions of the PIPS detector for the mono-energetic photons (5-30 keV). Then response functions were used as the training data of the neural network. The measured X-ray pulse height spectra of the N10-N30 and L10-L30 radiation qualities were unfolded to the true fluence spectra, and it were compared with the unfolded spectra using the GRV_MC33 program. The results show that except for N25 and N30 radiation quality, the results of the two method are in good agreement, which verifies the feasibility of using neural networks to unfold the low-energy X-ray pulse height spectrum. The difference in the calculated spectrum possibly come from that the uncertainties of the response functions and the unfolding method of the GRV_MC33 code. Ultimately, the trained neural network can be transplanted in a minicomputer and help calibration laboratories achieve the fast unfolding of the low-energy filtered X-rays.
Keywords: low-energy filtered X-rays;unfolding of pulse height spectra;PIPS detector;efficiency calibration;artificial neural network
2021, 47(2):26-31  收稿日期: 2020-06-23;收到修改稿日期: 2020-08-02
基金项目:
作者简介: 滕忠斌(1993-),男,山东聊城市人,博士研究生,研究方向为电离辐射计量
参考文献
[1] Radiological protection—X and gamma reference radiation for calibrating dosemeters and doserate meters and for determining their response as a function of photon energy—Part 3: Calibration of area and personal dosemeters and the measurement of their response as a function of energy and angle of incidence: ISO 4037—3[S]. Geneva: ISO, 2019.
[2] Radiological protection—X and gamma reference radiation for calibrating dosemeters and doserate meters and for determining their response as a function of photon energy—Part 4: Calibration of area and personal dosemeters in low energy X reference radiation fields: ISO 4037—4[S]. Geneva: ISO, 2019.
[3] Radiological protection—X and gamma reference radiation for calibrating dosemeters and doserate meters and for determining their response as a function of photon energy—Part 2: Dosimetry for radiation protection over the energy ranges from 8 keV to 1.3 MeV and 4 MeV to 9 MeV: ISO 4037—2[S]. Geneva: ISO, 2019.
[4] STALLMANN F W. LSL-M2: a computer program for least-squares logarithmic adjustment of neutron spectra[R]. State of Tennessee, America: Oak Ridge National Laboratory, 1986.
[5] SANNA R S. Manual for BON: a code for unfolding multisphere spectrometer neutron measurements[R]. Washington, America: US Department of Energy, 1981.
[6] REGINATTO M, GOLDHAGEN P, NEUMANN S. Spectrum unfolding, sensitivity analysis and propagation of uncertainties with the maximum entropy deconvolution code MAXED[J]. Nuclear Instruments and Methods in Physics Research A, 2002, 476: 242-246
[7] FREEMAN D W, EDWARDS D R, BOLON A E. Genetic algorithms- new technique for solving the neutron spectrum unfolding problem[J]. Nuclear Instruments and Methods in Physics Research A, 1999, 425: 549-576
[8] VEGA-CARRILLO H R, HERNÁNDEZ-DÁVILA V M, MANZANARES-ACUÑA E, et al. Neutron spectrometry using artificial neural networks[J]. Radiation Measurements, 2006, 41: 425-431
[9] VEGA-CARRILLO H R, HERNÁNDEZ-DÁVILA V M, MANZANARES-ACUÑA E, et al. Artificial neural networks technology for neutron spectrometry and dosimetry[J]. Radiation Protection Dosimetry, 2007, 126(1/4): 408-412
[10] YAN J, LIU R, LI C, et al. Application of artificial neural networks for unfolding neutron spectra by using a scintillation detector[J]. Science China-Physics Mechanics & Astronomy, 2011, 54(3): 465-469
[11] IDO A S, BONYADI M R, ETAATI G R, et al. Unfolding the neutron spectrum of a NE213 scintillator using artificial neural networks[J]. Applied radiation & Isotopes, 2009, 67(10): 1912-1918
[12] AVDIC S, POZZI S, PROTOPOPESCU V. Detector response unfolding using artificial neural networks[J]. Nuclear Instruments and Methods in Physics Research A, 2006, 565(2): 742-752
[13] EGOROV O B, ADDLEMAN R S, O’HARA M J, et al. Direct measurement of alpha emitters in liquids using passivated ion implanted planar silicon (PIPS) diode detectors[J]. Nuclear Instruments and Methods in Physics Research A, 2005, 537: 600-609
[14] PADILLA ALVAREZ R, VAN ESPEN P, ESTEVEZ ALVAREZ J R. The advantages of using digital signal processing in polarized X-ray fluorescence analysis[J]. X-Ray Spectrom, 2006, 35: 178-183
[15] 汪林, 叶志耀, 董滨江. PIPS探测器的能量响应的MC计算[J]. 辐射防护, 2007(3): 151-155
[16] MCNP - A general Monte Carlo N-particle transport code version 5[R]. California, America: University of California, 2005.