现代应用物理2025，Vol.16Issue(1)：125-132,8.DOI:10.12061/j.issn.2095-6223.202407032
一种基于人工神经网络改进的脉冲X射线能谱测量解谱算法
An Iterative Unfolding Algorithm for Pulsed X-ray Spectra Based on Artificial Neural Networks
摘要
Abstract
As the power density and repetition rate of laser devices increase,traditional iterative unfolding spectra algorithms such as the least squares method and the expectation maximization(EM)algorithm have issues of slow unfolding speed and need for manual input of the initial iterative energy spectra,making it difficult to measure X-ray energy spectra in laser produced plasma in real time.In this paper,an iterative unfolding spectra algorithm based on artificial neural network model is proposed.The particle-in-cell(PIC)method is used to simulate the interaction between laser and plasma,and the super hot electrons and characteristic temperature are obtained.Then,the super hot electrons are used as the source term and the Monte Carlo(MC)method is used to obtain the pulsed X-ray spectra and corresponding energy spectrum temperature,so as to establish the relationship between laser experimental parameters and X-ray energy spectra.Based on the PIC and MC methods,the electron energy spectra and corresponding characteristic temperature of pulsed X-ray spectra in the laser power density range of 1018～1020 W·cm-2 are obtained,which are used as the data set of the neural network,and the training of the neural network model is completed on the Matlab platform.Furthermore,the proposed algorithm is compared with the Gravel algorithm and EM algorithm,its key indicators such as accuracy and speed of unfolding spectra are verified and tested.The results show that the proposed algorithm has improved in terms of mean square error(MSE),average absolute relative deviation(AARD),quality spectrum(QS),running speed,and accuracy of characteristic temperature of X-ray spectra.At present,the algorithm has been successfully applied to laser devices.关键词
超短超强激光/脉冲X射线/能谱/神经网络/迭代解谱算法Key words
ultra-short and ultra-intense laser/pulsed X-ray/energy spectrum/neural network/iterative unfolding algorithm分类
核科学引用本文复制引用
刘一宁,宋鸿鹄,邱睿,衣宏昌,李君利..一种基于人工神经网络改进的脉冲X射线能谱测量解谱算法[J].现代应用物理,2025,16(1):125-132,8.基金项目
国家自然科学基金资助项目(U2167209) （U2167209）
国家重点研发计划资助项目(2021YFF0603600) （2021YFF0603600）
