skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Importance Sampling Variance Reduction in GRESS ATMOSIM

Abstract

This document is intended to introduce the importance sampling method of variance reduction to a Geant4 user for application to neutral particle Monte Carlo transport through the atmosphere, as implemented in GRESS ATMOSIM.

Authors:
 [1]
  1. Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
Publication Date:
Research Org.:
Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
Sponsoring Org.:
USDOE National Nuclear Security Administration (NNSA)
OSTI Identifier:
1356092
Report Number(s):
LA-UR-17-23432
DOE Contract Number:
AC52-06NA25396
Resource Type:
Technical Report
Country of Publication:
United States
Language:
English
Subject:
47 OTHER INSTRUMENTATION

Citation Formats

Wakeford, Daniel Tyler. Importance Sampling Variance Reduction in GRESS ATMOSIM. United States: N. p., 2017. Web. doi:10.2172/1356092.
Wakeford, Daniel Tyler. Importance Sampling Variance Reduction in GRESS ATMOSIM. United States. doi:10.2172/1356092.
Wakeford, Daniel Tyler. 2017. "Importance Sampling Variance Reduction in GRESS ATMOSIM". United States. doi:10.2172/1356092. https://www.osti.gov/servlets/purl/1356092.
@article{osti_1356092,
title = {Importance Sampling Variance Reduction in GRESS ATMOSIM},
author = {Wakeford, Daniel Tyler},
abstractNote = {This document is intended to introduce the importance sampling method of variance reduction to a Geant4 user for application to neutral particle Monte Carlo transport through the atmosphere, as implemented in GRESS ATMOSIM.},
doi = {10.2172/1356092},
journal = {},
number = ,
volume = ,
place = {United States},
year = 2017,
month = 4
}

Technical Report:

Save / Share:
  • The author derives a transformed transport problem that can be solved theoretically by analog Monte Carlo with zero variance. However, the Monte Carlo simulation of this transformed problem cannot be implemented in practice, so he develops a method for approximating it. The approximation to the zero variance method consists of replacing the continuous adjoint transport solution in the transformed transport problem by a piecewise continuous approximation containing local biasing parameters obtained from a deterministic calculation. He uses the transport and collision processes of the transformed problem to bias distance-to-collision and selection of post-collision energy groups and trajectories in a traditionalmore » Monte Carlo simulation of ``real`` particles. He refers to the resulting variance reduction method as the Local Importance Function Transform (LIFI) method. He demonstrates the efficiency of the LIFT method for several 3-D, linearly anisotropic scattering, one-group, and multigroup problems. In these problems the LIFT method is shown to be more efficient than the AVATAR scheme, which is one of the best variance reduction techniques currently available in a state-of-the-art Monte Carlo code. For most of the problems considered, the LIFT method produces higher figures of merit than AVATAR, even when the LIFT method is used as a ``black box``. There are some problems that cause trouble for most variance reduction techniques, and the LIFT method is no exception. For example, the author demonstrates that problems with voids, or low density regions, can cause a reduction in the efficiency of the LIFT method. However, the LIFT method still performs better than survival biasing and AVATAR in these difficult cases.« less
  • The GRESS Version 0.0 code system was developed to automate the implementation of derivative-taking capabilities in existing FORTRAN 77 computer models. The GRESS CHAIN option is used to calculate and report first derivatives of model results with respect to user selected input data by application of the calculus chain rule. The GRESS ADjoint Matrix GENenerator (ADGEN) option is used to calculate first derivatives of selected model results with respect to all input data. The first part of this paper presents the mathematical foundations and algorithms as presently implemented in GRESS Version 0.0. Examples are used to describe the implementation ofmore » both the CHAIN and ADGEN options. Due to excessive execution time and memory requirements with the CHAIN option users are often limited to propagating derivatives for just a few parameters. The ADGEN option allows an almost unlimited number of parameters (i.e., input data); however, the data storage requirement for an ADGEN application was more than 322 megabytes for a code that executes in 1 minute on a VAX 8600 computer. The purpose for this paper is to present three new algorithms that could easily be implemented in GRESS Version 0.0 to dramatically reduce he data storage requirements and execution time for application of the ADGEN option. The new algorithms are described with examples. Test versions of these algorithms were implemented and tested. The application of these algorithms to the GRESS enhancement of the PRESTO-II computer model resulted in a significant reduction in execution time and a reduction in data storage requirements from 322 megabytes to 97 megabytes without any loss in the generality of the approach. 23 refs., 14 figs., 8 tabs.« less
  • The Fokker–Planck approximation to the Boltzmann equation, solved numerically by stochastic particle schemes, is used to provide estimates for rarefied gas flows. This paper presents a variance reduction technique for a stochastic particle method that is able to greatly reduce the uncertainty of the estimated flow fields when the characteristic speed of the flow is small in comparison to the thermal velocity of the gas. The method relies on importance sampling, requiring minimal changes to the basic stochastic particle scheme. We test the importance sampling scheme on a homogeneous relaxation, planar Couette flow and a lid-driven-cavity flow, and find thatmore » our method is able to greatly reduce the noise of estimated quantities. Significantly, we find that as the characteristic speed of the flow decreases, the variance of the noisy estimators becomes independent of the characteristic speed.« less
  • Reactor properties can be investigated by observing fluctuations in the numober of counts occurring in a fixed time-interval in a neutron counter. The statistical parameter variance/mean of the count population may be estimated from the observed statistic N(N-l!/sup -1/ x variance/mean calculated for a sample of counts obtained in N time-intervals. It is shown that this method of estimation is biased, and a bound is derived for the magnitude of the bias. The bias can be severe when N is small and the mean number of counts per time interval is also small. The expected distribution of the sample statisticmore » N(N-1!/sup -1/ x variance/mean is described. The results of the study are illustrated by analysis of actual reactor fluctuation data. (auth)« less
  • Data acquired form the sampling of Yakima River water July 24, 1947, is presented. The water was analyzed for hardness, turbidity, dissolved oxygen, ph value, alkalinity, chlorine, NH(sub 3), NO(sub 3), and total solids.