RESEARCH AND DEVELOPMENT OF POINT-KERNEL METHOD APPLICATION IN RADIATION SHIELDING (CROSBI ID 623648)
Prilog sa skupa u zborniku | izvorni znanstveni rad
Podaci o odgovornosti
Dučkić, Paulina ; Trontl, Krešimir ; Pevec, Dubravko
engleski
RESEARCH AND DEVELOPMENT OF POINT-KERNEL METHOD APPLICATION IN RADIATION SHIELDING
The point-kernel method is a widely used practical tool in shielding design. In the point-kernel approach the desired quantity (flux, fluence, or dose equivalent) is obtained by multiplying the part of the quantity resulting from uncollided flux with the parameter called buildup factor which accounts for the scattered radiation inside the shielding medium. It is therefore clear that the accuracy of the final result is highly dependent on the accuracy of the buildup factors used in the calculations. Although buildup factors have been widely used for gamma-ray shielding calculations, their application for neutron transport simulations is very limited. Recently a novel approach, based on support vector regression (SVR) technique, for multi-layer buildup factor calculation has been proposed and tested for gamma-ray shielding calculations, while for neutron shielding calculations some initial analyses have been conducted. Initial tests of the SVR model on mono-layer shields revealed that the relative average deviation (RAD) from the reference data (obtained by Monte Carlo calculation) in the estimation of gamma-ray buildup factors is 5.8% with maximum deviation of 34% observed for water shield 5.21 mfp thick at incident gamma-ray energy of 0.5 MeV. RAD for double-layer shields was 9.7% with maximum deviation of 68.77% observed for a shield comprised of 1.17 mfp of water and 5.75 mfp of iron, at 7.25 MeV gamma-ray energy. RADs for triple-layer and quadruple-layer shields were 24.3% and 29.7%, respectively. The initial tests illustrate clear potential of the novel SVM approach to the gamma-ray buildup factor determination. The physical complexity of the neutron radiation transport and the experience gained while developing the gamma-ray model indicate that the development of the SVR neutron buildup factor model is possible but would also require intensive research of the learning technique itself.
point kernel method; gamma shielding; neutron shielding; machine learning
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
Podaci o prilogu
402-407.
2015.
objavljeno
Podaci o matičnoj publikaciji
Proceedings of the 10th Symposium of the CROATIAN RADIATION PROTECTION ASSOCIATION
Petrinec, Branko ; Bituh, Tomislav ; Milić, Mirta ; Kopjar, Nevenka
Zagreb: CROATIAN RADIATION PROTECTION ASSOCIATION
Podaci o skupu
10th Symposium of the Croatian Radiation Protection Association with international participation
predavanje
15.04.2015-17.04.2015
Šibenik, Hrvatska