The application of artificial neural networks technique to estimate mass attenuation coefficient of shielding barrier

This study aims to investigate comparison of radiation attenuation property of harzburgite mineral calculated by Monte Carlo (MCNP) and Artificial Neural Network (ANN). Slab sample modeled in MCNP with 1, 2 and 4 cm thickness was irradiated with parallel beam of monoenergetic particles. Incoming and outgoing particle fluxes were computed with F1 tally. Beer-Lambert equation was used to obtain mass attenuation coefficients for photon energies between 40 keV and 20 MeV. Optimum ANN model was obtained after trying different structures in terms of iterations and hidden layer numbers. For ANN calculation, parameters considered in the study are dose, thickness and mass attenuation coefficient. Dose and thickness are used as inputs to ANN for the estimation of mass attenuation coefficient. Model results are evaluated using root mean square errors (RMSE) and determination coefficient (R-2) statistics. The estimates of selected ANN model were compared with MCNP results. Based on the comparison results, ANN was found good in prediction of mass attenuation coefficient for shielding material. Relationship between observed MCNP values and ANN estimates is noticeable with a high determination coefficient (R-2) of 1 and has a root mean square error (RMSE) of 0.0033.