Influence of Neutron Spectra Unfolding Method on Fast Neutron Dose Determination

Full text: Accuracy of knowing the fast neutron spectra has great influence on equivalent dose determination. In usual fast neutron spectrum measurements with scintillation detectors based on proton recoil, the main difficulty is confidence of unfolding method. In former ones variance of obtained result is usually great and negative values are possible too, which does means that we don't now exactly is obtained neutron spectrum real one. The new unfolding method based on Shanon's information theory, which gives non-negative spectrum and relative low variance, is obtained and appropriate numerical code for application in fast neutron spectrometry based on proton recoil is realized. In this method principle of maximum entropy and maximum likelihood are used together. Unknown group density distribution functions, which are considered as desired normalized mean neutron group flux, are constlucted using only constrain of knowing mean value. Obtained distributions are consistent to available information (counts in NCA from proton recoil), while being maximally noncommittal with respect to all other unknown circumstances. For maximum likelihood principle, distribution functions around mean value of counts in the channels of MCA are taken to be Gauss function shape. Optimal non-negative solution is searched by means of Lagrange parameter method. Nonlinear system of equations, is solved using gradient and Newton iterative algorithm. Error covariance matrix is obtained too. (author)