Neutron spectroscopy with soft computing: development of a computational code based on Support Vector Machine (SVM) for reconstruction of neutron energy spectrum

Abstract

This paper presents a developed computational code based on Support Vector Machine (SVM) for reconstruction of energy spectrum of neutron source. To reconstruct unknown energy spectrum using known neutron pulse height distribution, the developed machine is trained by known neutron pulse height distribution of detector and corresponding energy spectrum of neutron source. Validation and testing are the next steps to verify the validity of the calculations done with the developed computational code. The calculated neutron pulse height distributions due to randomly generated energy spectrum using MCNPX-ESUT (MCNPX-Energy engineering of Sharif University of Technology) computational code are used as input data of the developed computational code for reconstruction of energy spectrum of neutron source. The considered energy spectrums of neutron source are output data. In the present study, the simulation of neutron pulse height distribution of 241Am-9Be neutron source in the NE-213 liquid organic scintillator is performed using MCNPX-ESUT computational code. In the constructed matrix form of Fredholm integral equation, the effective response matrix is usually close to singular or badly scaled and its condition number is so high. The solution of inverse problem in which the matrix has high condition number does not give accurate results. The motivation of the present study is the development of computational coded based on soft computing algorithm for reconstruction of energy spectrum of the neutron source with high accuracy. To this end, the computational code based on Support Vector Machine (SVM) is developed, where there is no need to solve the inverse problems. The unfolded energy spectra of 241Am-9Be neutron sources using the developed computational code have an excellent agreement with the reference spectrum provided by ISO. A comparison between the results of present study and results of developed computational code based on Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) is performed.