Application Research Of Genetic Algorithm In Water-pumping-injection Multi-sphere Neutron Spectrometer

At present,the multi-sphere neutron spectrometer is widely used in the measurement of neutron energy spectrum in various scenes.The traditional multi-sphere neutron spectrometer is composed of multiple polyethylene spheres with different diameters,which is expensive,low portability,and its measurement range is small,the number of probe spheres is large when measuring,and there is a problem of"sphere-sphere interference"when the interval is small.Therefore,the research group of this thesis on the National Natural Science Foundation of the project"study of multi-level concentric spherical pumps with wide range of new neutron energy spectrum detector research"?No.41774120?,National Defense Industry Science and Technology Bureau's technical basic research project"space radiation neutron Measurement and Dose Calibration Technology"sub-topic"Research on Spatial Radiation Neutron Spectral Analysis Technology"?No.JSJL2014404B001?,Sichuan Provincial Key R&D Project"Research on Water-Pumping-Injection Multi-sphere Neutron Spectra Measurement System"?No.2019YFG0430?Under the support of the company,we independently developed a water-pumping-injection multi-sphere neutron spectrometer.By pumping water into different water layer structures to change the different neutron moderating thickness,to achieve the response to different neutron energy,with a wide measurement range,low cost,portable,and does not have the advantages of"sphere-sphere interference".However,it takes time to pump or inject water into different water layers during the measurement process,which affects the measurement efficiency of the spectrometer during use.The time cost varies greatly with the change of the measurement sequence.How to shorten the pumping or injection time,to reduce the difference is one of the key technical problems to be solved in the water-pumping-injection multi-sphere neutron spectrometer.At the same time,when participating in the research team in the early stage,the BP neural network method was used to perform the unfolding work.It was found that the method needs to perform t-SNE dimensionality reduction classification on the response function before the unfolding work.This dimensionality reduction method has randomness.The results of sub-classification will cause errors in the input layer data of the BP neural network,and the result of the demodulation will fall into the local optimum.How to eliminate these random errors is the second key technical problem to be solved in the water-pumping-injection multi-sphere neutron spectrometer.Therefore,this paper aims at the above two key technical problems,using genetic algorithms to simulate the natural evolution process to search for the optimal solution,study the optimal time and measurement sequence under different water layer combinations and under the response function without dimensionality reduction classification processing to seek a global optimal solution.Through the research,the main results are as follows:1.First,make a subtraction between the pumping and injection time of any two water layer structure switching processes in the water layer structure used in the neutron count measurement process of the water-pumping-injection multi-sphere neutron spectrometer,thereby obtaining a time matrix,by The time cost function obtained by accumulating the time units in the time matrix is used as the fitness function in the time optimization algorithm model.Second,the single-point crossover operator and the partially matched crossover operator in the genetic algorithm are applied to the time optimization algorithm model.And in turn,the time optimization effect of three types of experiments,such as 2⁵?32?random combination of water layer structure,18 random selection of random water layer structure,and 6 random combination of random water layer structure,was selected and compared.The results show that under the partially matched crossover operator,the shortest measurement time can be found for the above three types of experimental objects,respectively 3212 s,2000 s and 1568 s,and gave the best measurement permutation sequence.Therefore,a time optimization algorithm model based on genetic algorithm was established by combining the time cost function and the partially matched crossover operator.The algorithm model was applied to 32 water layer structures and 18 randomly selected water layer structures in experiments.The total measurement time in any combination is optimized for time,which is 3.35 times and 3.12 times shorter than that without time optimization,which greatly shortens the measurement time and improves the measurement efficiency.2.The genetic algorithm is used to solve the energy spectrum of the neutron count measured by the water-pumping-injection multi-sphere neutron spectrometer to optimize the iterative process of the spectrum and finally find the optimal solution among many solution vectors.First,binary code the initial neutron spectrum solution space to ensure the continuity and simplicity of the solution process.Second,comparing the effect of using different fitness functions on the unfolding effect during the unfolding process of the water-pumping-injection multi-sphere neutron spectrometer.Select the square distance between the neutron count measured by the spectrometer and neutron count by convolution of response function and neutron energy spectrum as fitness function.Then,by comparing the influence of two-point crossover operator and uniform crossover operator on the result of unfolding in unfolding algorithm respectively,it is proposed to use uniform crossover operator for unfolding.Finally,combined with the determination of the unfolding process parameters,a water-pumping-injection multi-sphere neutron spectrum reconstruction algorithm model based on genetic algorithm was established,and the algorithm model was used for experimental verification.and five standard neutron energy spectra in the IAEA report,²⁵²Cf,²⁴¹Am-Be,the ²⁵²Cf neutron spectrum after D₂O moderation,²⁴¹Am-B neutron spectrum and ²⁵²Cf+²⁴¹Am-Be combined spectrum to verify.The overall trend of the energy spectrum is consistent with the standard neutron energy spectrum,and the evaluation factors Q_S and Q_R of the unfolding effect are close to 0,and the unfolding results are in high compliance with the above five standard energy spectra.3.Using Visual Studio 2019 development tools and SQL Server 2017 database to develop a set of water-pumping-injection multi-sphere neutron spectrometer software system.The system integrates the two key technical algorithms such as the genetic algorithm-based time optimization algorithm model and the energy spectrum reconstruction algorithm model,and realizes the functions of acquisition,control,processing,analysis,display,management,and output.The main innovations of this paper:?1?A time optimization algorithm model based on genetic algorithm was established and applied to the water-pumping-injection multi-sphere neutron spectrometer,and 32 water layers,18 water layers and 6 water layers were randomly combined into three model experiments were carried out to verify the accuracy and practicability of the algorithm model.?2?A water-pumping-injection multi-sphere neutron spectrum reconstruction algorithm model based on genetic algorithm was established.The neutron counts measured by water-pumping-injection multi-sphere neutron spectrometer were realized,and the energy spectrum was reconstructed and optimized,the neutron spectrum obtained by this algorithm has a high compliance with the standard neutron spectrum in the IAEA report.