You need JavaScript to view this

Probability Density Estimation Using Neural Networks in Monte Carlo Calculations

Abstract

The Monte Carlo neutronics analysis requires the capability for a tally distribution estimation like an axial power distribution or a flux gradient in a fuel rod, etc. This problem can be regarded as a probability density function estimation from an observation set. We apply the neural network based density estimation method to an observation and sampling weight set produced by the Monte Carlo calculations. The neural network method is compared with the histogram and the functional expansion tally method for estimating a non-smooth density, a fission source distribution, and an absorption rate's gradient in a burnable absorber rod. The application results shows that the neural network method can approximate a tally distribution quite well. (authors)
Authors:
Shim, Hyung Jin; Cho, Jin Young; Song, Jae Seung; [1]  Kim, Chang Hyo [2] 
  1. Korea Atomic Energy Research Institute, 1045 Daedeokdaero, Yuseong-gu, Daejeon 305-353 (Korea, Republic of)
  2. Seoul National University, San 56-1, Shillim-dong, Gwanak-gu, Seoul 151-742 (Korea, Republic of)
Publication Date:
Jul 01, 2008
Product Type:
Conference
Resource Relation:
Conference: PHYSOR'08: International Conference on the Physics of Reactors 'Nuclear Power: A Sustainable Resource', Interlaken (Switzerland), 14-19 Sep 2008; Other Information: Country of input: France; 11 refs.; proceedings are available as a CD-ROM on request to info'at'physor08.ch
Subject:
97 MATHEMATICAL METHODS AND COMPUTING; 73 NUCLEAR PHYSICS AND RADIATION PHYSICS; 22 GENERAL STUDIES OF NUCLEAR REACTORS; ABSORPTION; APPROXIMATIONS; DENSITY; DISTRIBUTION; EXPANSION; FISSION; FUEL RODS; MONTE CARLO METHOD; NEURAL NETWORKS; POWER DISTRIBUTION; PROBABILITY; PROBABILITY DENSITY FUNCTIONS; CALCULATION METHODS; FUEL ELEMENTS; FUNCTIONS; NUCLEAR REACTIONS; PHYSICAL PROPERTIES; REACTOR COMPONENTS; SORPTION
OSTI ID:
21391777
Research Organizations:
Paul Scherrer Institut - PSI, 5232 Villigen PSI (Switzerland)
Country of Origin:
Switzerland
Language:
English
Other Identifying Numbers:
Other: ISBN 978-3-9521409-5-6; TRN: CH10V0241119094
Submitting Site:
CHN
Size:
6 p. pages
Announcement Date:
Feb 07, 2011

Citation Formats

Shim, Hyung Jin, Cho, Jin Young, Song, Jae Seung, and Kim, Chang Hyo. Probability Density Estimation Using Neural Networks in Monte Carlo Calculations. Switzerland: N. p., 2008. Web.
Shim, Hyung Jin, Cho, Jin Young, Song, Jae Seung, & Kim, Chang Hyo. Probability Density Estimation Using Neural Networks in Monte Carlo Calculations. Switzerland.
Shim, Hyung Jin, Cho, Jin Young, Song, Jae Seung, and Kim, Chang Hyo. 2008. "Probability Density Estimation Using Neural Networks in Monte Carlo Calculations." Switzerland.
@misc{etde_21391777,
title = {Probability Density Estimation Using Neural Networks in Monte Carlo Calculations}
author = {Shim, Hyung Jin, Cho, Jin Young, Song, Jae Seung, and Kim, Chang Hyo}
abstractNote = {The Monte Carlo neutronics analysis requires the capability for a tally distribution estimation like an axial power distribution or a flux gradient in a fuel rod, etc. This problem can be regarded as a probability density function estimation from an observation set. We apply the neural network based density estimation method to an observation and sampling weight set produced by the Monte Carlo calculations. The neural network method is compared with the histogram and the functional expansion tally method for estimating a non-smooth density, a fission source distribution, and an absorption rate's gradient in a burnable absorber rod. The application results shows that the neural network method can approximate a tally distribution quite well. (authors)}
place = {Switzerland}
year = {2008}
month = {Jul}
}