Research On Information Restoration Technology Of γ Spectra

In order to effectively respond to nuclear war and diversified military missions involving nuclear safety,there is an urgent need to develop high-performance equipment technology ofγspectrum measurement and analysis.However,limited by the inherent performance of hardware such as detectors and electronics,the current military and civilianγspectrometers are not yet fully capable of qualitative identification and quantitative analysis of radionuclides in complex radiation fields.In recent years,γspectrum measurement and analysis technology is developing continuously towards digitalization and algorithmization,and spectrum deconvolution methods have been investgated and applied.All these make it possible to use complex methods and algorithms to restore the original information from the measured spectra,thereby improving the conprehensive performance ofγspectrometers.Focusing on the difficult problem of complexγspectrum analysis,this work treats the digitalγspectrometer from a new perspective of information system.Using simulation and experiment methods,the mechanism of information conversion and processing in theγspectrometer system and the information restoration methods are intensively studied,and a set of technical solution forγspectrum information restoration is proposed.This work provides key technical support to overcome the hardware constraints onγspectrometers and improve their performance.The main research contents and conclusions are summarized as follows:（1）Informationized modeling and simulation ofγspectrometer system.To systematically study the influence of various technical factors in the measurement process on spectral information,theγspectrometer is modeled as an information processing system of three parts,including a pulse generator of standard source spectrum,an ideal spectrometer,and a noise module.With Monte Carlo simulation method,the system response matrices of Na I（Tl）and La Br₃（Ce）spectrometers are established,and the influence of the measurement environment and geometry on system response is discussed.Based on MATLAB/Simulink platform,an informationized simulation system ofγspectrometer with adjustable parameters is built,making it practical to simulate the complete information flow from source items toγspectra.The Na I（Tl）and La Br₃（Ce）spectra simulated with this tool are consistent with the measured spectra,and the spectrum counts obey the statistical characteristics of the Poisson distribution,and the spectrum drift is simulated reasonably.（2）Full-spectrum stabilization based on software with nonlinear least squares algorithm.To reduce the negative effect of spectral drift on spectrum information restoration,an optimized software-based spectrum stabilization method is proposed.Theγspectra are recorded in consecutive short intervals.The drifts in different time intervals relative to the first reference spectrum are estimated with a full spectrum nonlinear optimization technique,and the drifted spectra are corrected by a spectrum reallocation method.Based on the holistic characteristics of spectra,this method needs neither reference peaks nor pre-calibration,and is applicable to both simple and complex spectra.The test results with simulated and measured spectra show that the stabilization accuracy can be significantly improved using this approach,with peak position drift less than 1 channel and energy resolution change at 662 ke V less than0.05%.（3）Study on deconvolution methods ofγspectra.To restore the originalγspectrum information,12 deconvolution methods have been systematically studied and compared.The results show that the nonnegative least squares（NNLS）and the weighted nonnegative least squares（WNNLS）methods are faster and more accurate.Based on WNNLS,the influence of factors,such as statistical fluctuations,energy resolution and system response matrix error,on deconvolution is studied,and the measured spectra of Na I（Tl）and La Br₃（Ce）spectrometers are deconvoluted.Two new types of system response matrices are proposed,namely the nuclide-based full-energy peak response matrix and the nuclide-based mixed response matrix.They can identify nuclides and estimate their activity by one step of calculation.Processing results with simulated and measured spectra show that the nuclide-based mixed response matrix is more practical,with better performance than that of classical library least squares method.In addition,theγspectrum background estimation methods based on multilayer perceptron（MLP）and radial basis function（RBF）neural network are explored.The RBF network can result in serious overfitting,while the MLP network estimates accurately,worthy of further study.（4）Design and verification of theγspectrum information restoration scheme.To further enhance practicability,an overall scheme ofγspectrum information restoration is proposed.A software package forγspectrum information restoration is designed and developed using C++/Qt.Based on this software and the system on a programmable chip（SOPC）platform for nuclear radiation measurement,a Na I（Tl）spectrum measurement and information restoration experimental system is set up.The performance of spectrum stabilization and deconvolution is verified by experiments.The errors of the peak position restored by the full-spectrum response matrix method and the characteristic peak response matrix method are both less than 2 channels,and the error of nuclide activity restored by the nuclide-based mixed response matrix method is less than 9%.This work proposes a new idea of modeling theγspectrometer as an information system,and creates an informationized simulation system of digitalγspectrometer.Using optimization algorithms,the software-based full-spectrum stabilization method is improved.The proposed nuclide-based mixed response matrix solves the difficult problem of the Compton continuum interference on spectrum deconvolution.The developed software package forγspectrum information restoration is verified by experiments to have excellent performance.This work provides an information restoration technical solution for improving the comprehensive performance ofγspectrometers.