A ``local`` exponential transform method for global variance reduction in Monte Carlo transport problems
- Los Alamos National Lab., NM (United States)
- Michigan Univ., Ann Arbor, MI (United States). Dept. of Nuclear Engineering
Numerous variance reduction techniques, such as splitting/Russian roulette, weight windows, and the exponential transform exist for improving the efficiency of Monte Carlo transport calculations. Typically, however, these methods, while reducing the variance in the problem area of interest tend to increase the variance in other, presumably less important, regions. As such, these methods tend to be not as effective in Monte Carlo calculations which require the minimization of the variance everywhere. Recently, ``Local`` Exponential Transform (LET) methods have been developed as a means of approximating the zero-variance solution. A numerical solution to the adjoint diffusion equation is used, along with an exponential representation of the adjoint flux in each cell, to determine ``local`` biasing parameters. These parameters are then used to bias the forward Monte Carlo transport calculation in a manner similar to the conventional exponential transform, but such that the transform parameters are now local in space and energy, not global. Results have shown that the Local Exponential Transform often offers a significant improvement over conventional geometry splitting/Russian roulette with weight windows. Since the biasing parameters for the Local Exponential Transform were determined from a low-order solution to the adjoint transport problem, the LET has been applied in problems where it was desirable to minimize the variance in a detector region. The purpose of this paper is to show that by basing the LET method upon a low-order solution to the forward transport problem, one can instead obtain biasing parameters which will minimize the maximum variance in a Monte Carlo transport calculation.
- Research Organization:
- Los Alamos National Lab., NM (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- W-7405-ENG-36
- OSTI ID:
- 10167411
- Report Number(s):
- LA-UR-92-2349; CONF-930404-1; ON: DE92018992
- Resource Relation:
- Conference: International topical meeting on mathematical methods and supercomputing in nuclear applications (M&C+SNA `93),Karlsruhe (Germany),19-23 Apr 1993; Other Information: PBD: [1992]
- Country of Publication:
- United States
- Language:
- English
Similar Records
Advanced computational methods for nodal diffusion, Monte Carlo, and S[sub N] problems
Advanced computational methods for nodal diffusion, Monte Carlo, and S{sub N} problems. Progress report, January 1, 1992--March 31, 1993