Advanced Variance Reduction Strategies for Optimizing Mesh Tallies in MAVRIC

Conference · Sun Nov 11 04:00:00 EST 2007

OSTI ID:930967

Peplow, Douglas E. ^[1]; Blakeman, Edward D. ^[1]; Wagner, John C. ^[1]

---

1. Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States)

More often than in the past, Monte Carlo methods are being used to compute fluxes or doses over large areas using mesh tallies (a set of region tallies defined on a mesh that overlays the geometry). For problems that demand that the uncertainty in each mesh cell be less than some set maximum, computation time is controlled by the cell with the largest uncertainty. This issue becomes quite troublesome in deep-penetration problems, and advanced variance reduction techniques are required to obtain reasonable uncertainties over large areas. The CADIS (Consistent Adjoint Driven Importance Sampling) methodology has been shown to very efficiently optimize the calculation of a response (flux or dose) for a single point or a small region using weight windows and a biased source based on the adjoint of that response. This has been incorporated into codes such as ADVANTG (based on MCNP) and the new sequence MAVRIC, which will be available in the next release of SCALE. In an effort to compute lower uncertainties everywhere in the problem, Larsen's group has also developed several methods to help distribute particles more evenly, based on forward estimates of flux. This paper focuses on the use of a forward estimate to weight the placement of the source in the adjoint calculation used by CADIS, which we refer to as a forward-weighted CADIS (FW-CADIS).

• Research Organization: Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States)
• Sponsoring Organization: USDOE National Nuclear Security Administration (NNSA), Nuclear Criticality Safety Program (NCSP)
• DOE Contract Number: AC05-00OR22725
• OSTI ID: 930967
• Country of Publication: United States
• Language: English

Related Subjects

97 MATHEMATICS AND COMPUTING
ADVANTG
CADIS (Consistent Adjoint Driven Importance Sampling) Methodology
MAVRIC
Monte Carlo Methods
Nuclear Criticality Safety Program (NCSP)
SCALE