Development of spectrum unfolding code for multi-sphere neutron spectrometer using genetic algorithms

In the process of neutron spectrum measurement using the multi-sphere neutron spectrometer,energy response functions and detector readings should be applied to neutron spectrum unfolding.Mathematically,there can be multiple solutions to this problem,but only one actual neutron spectrum exists.Compared with common numerical spectrum unfolding methods,genetic algorithms have the characteristics of global optimization and probabilistic search.Therefore,they are chosen to be the spectrum unfolding algorithms for the multisphere neutron spectrometer(MNS IL 100) developed by Tsinghua University.Firstly,the detector and different size polyethylene spheres of MNS IL100 were modeled to calculate the energy response functions by applying Monte Carlo simulation.Then based on the physical and mathematical properties of the spectrum unfolding problem by using genetic algorithms,effective search space and proper fitness function were determined to improve the efficiency of search and iteration.The elitism replacement scheme was used to ensure convergence and the pseudo-parallel strategy was used to inhibit premature convergence.According to the algorithms mentioned above,a spectrum unfolding code was developed and tested with several typical neutron spectra.At last,MNS IL100 and the spectrum unfolding code were used in actual experiment of 252 Cf neutron source spectrum measurement.The experimental result is in good agreement with the 252 Cf standard neutron spectrum,verifying the effectiveness and practicality of using genetic algorithms to unfold the neutron spectrum with combination of several processing strategies.