Research Of Neutron Signal Processing For²⁵²Cf-source-drive Nuclear Material Based On Compressive Sensing

Nowadays, the world economic and political situation is complicated,under thetimes theme of peace and development, there still exist many problems that troublehuman society, nuclear proliferation is one of them. How to prevent and control thespread of nuclear weapons (nuclear disarmament or nuclear arms control verification),has been a serious issue that the international communityconcern.252Cf neutron sourceof random signal measurement technology isa representative of the active verificationtechniques, as its advantages of anti-interference ability and small reliance onenvironment, researchers have concerned on how to improve it to do further work.This paper makes National Natural Science Foundation " Applied basic research ofactive multi-channel random neutron pulse real-time spectrum analysis for nuclear armscontrol verification " as the basis, for the random neutron signal has the characteristic of"sparsity", Compressive sensing theory and252Cf neutron source drive signal analysisare combined in an innovative way, the research of random neutron signalreconstruction, spectrum identify and de-noising algorithms under the framework ofcompressive sensing is studied here, including the following contents:①Neutron pulse signal reconstruction and spectrum analysis based oncompressed sensing. According to the principle of neutron-induced fission and actualmeasurement result of252Cf source driven high speed nuclear signal measuring system,the neutron pulse signals of the three channels are all "0、1" structure with obvioussparse feature. In this paper a Gaussian random measurement matrix is to constructed toimprove the sparsity, via judging the singular value of the least-squares algorithm tochange observation matrix, design optimization orthogonal matching pursuit (OMP)algorithm to improve the accuracy of the reconstruction purposes. Experimental resultsshow that the proposed reconstruction algorithm reconstruction efficiency is over90%,the reconstruction error is very small, which can effectively realize the reconstruction ofthe original signal under sparse conditions. Furthermore, in the basis of thereconstructed neutron signal, we use the data of the three-channel to doauto/cross-correlation function, auto/cross power spectral density ratio, time domaincharacteristics are analyzed, gets a series of important labels which reacts nuclearmaterials, lays a solid foundation for further study.②The paper studies random nuclear signal denoise method under Compressive Sensing framework. For the problem of noise interference in the data collection process,wavelet transform and Compressive Sensing are combined, the noise in the waveletdomain is not sparse and the coefficients are small, it cannot be preciselyreconstructed when using Compressive Sensing. the noisy nuclear signal is decomposedby wavelet nuclear, and high-frequency coefficients are processed with observationmatrix and OMP algorithm, in order to achieve high-precision signal reconstruction andremoval of noise. Simulation results show that the random nuclear signal de-noisingalgorithm based on Compressive Sensing can effectively suppress noise, peak signal tonoise ratio and mean square error parameters are greatly improved compared to theoriginal signal.This paper mainly studies processing and analysis techniques of252Cfsource-driven nuclear materials neutron signal, with Compressed Sensing theory as astarting point, the sparse characteristics of the neutron signal, doing researches aboutreconstruction and de-noised of the neutron signal, also access to a range of fruitfulresults. These studies undoubtedly has important theoretical significance and potentialpractical value for exploring new ways for neutron signal acquisition and processingand further broaden applications areas of high-speed nuclear signal measurementsystem.