Pregled bibliografske jedinice broj: 826251
Neutron Buildup Factors Calculation for Support Vector Regression Application in Shielding Analysis
Neutron Buildup Factors Calculation for Support Vector Regression Application in Shielding Analysis // 11th International Conference of the Croatian Nuclear Society Conference Proceedings / Šimić, Zdenko ; Tomšić, Željko ; Grgić, Davor (ur.).
Zagreb: Hrvatsko nuklearno društvo, 2016. str. 051-1 (predavanje, međunarodna recenzija, cjeloviti rad (in extenso), znanstveni)
CROSBI ID: 826251 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Neutron Buildup Factors Calculation for Support Vector Regression Application in Shielding Analysis
Autori
Dučkić, Paulina ; Matijević, Mario ; Grgić, Davor
Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni
Izvornik
11th International Conference of the Croatian Nuclear Society Conference Proceedings
/ Šimić, Zdenko ; Tomšić, Željko ; Grgić, Davor - Zagreb : Hrvatsko nuklearno društvo, 2016, 051-1
ISBN
978-953-55224-8-5
Skup
11th International Conference of the Croatian Nuclear Society
Mjesto i datum
Zadar, Hrvatska, 05.06.2016. - 08.06.2016
Vrsta sudjelovanja
Predavanje
Vrsta recenzije
Međunarodna recenzija
Ključne riječi
neutron buildup factor; support vector regression; machine learning; point kernel
Sažetak
In this paper initial set of data for neutron buildup factors determination using Support Vector Regression (SVR) method is prepared. The performance of SVR technique strongly depends on the quality of information used for model training. Thus it is very important to provide representative data to the SVR. SVR is a supervised type of learning so it demands data in the input/output form. In the case of neutron buildup factors estimation, the input parameters are the incident neutron energy, shielding thickness and shielding material and the output parameter is the neutron buildup factor value. So far initial sets of data for different shielding configurations have been obtained using SCALE4.4 sequence SAS3. However, this results were obtained using group constants, thus the incident neutron energy was determined as the average value for each energy group. Obtained in this way, the data provided to the SVR are fewer and therefore insufficient. More valuable information is obtained using SCALE6.2beta5 sequence MAVRIC which can perform calculations for the explicit incident neutron energy, which leads to greater maneuvering possibilities when active learning measures are employed, and consequently improves the quality of the developed SVR model.
Izvorni jezik
Engleski
Znanstvena područja
Elektrotehnika
POVEZANOST RADA
Projekti:
HRZZ-UIP-2013-11-3522 - Razvoj paketa računalskih programa za naprednu analizu štitova od gama i neutronskog zračenja (DOCPAGANSA) (Trontl, Krešimir, HRZZ ) ( CroRIS)
Ustanove:
Fakultet elektrotehnike i računarstva, Zagreb
