Error reduction techniques for Monte Carlo neutron transport calculations
Monte Carlo methods have been widely applied to problems in nuclear physics, mathematical reliability, communication theory, and other areas. The work in this thesis is developed mainly with neutron transport applications in mind. For nuclear reactor and many other applications, random walk processes have been used to estimate multi-dimensional integrals and obtain information about the solution of integral equations. When the analysis is statistically based such calculations are often costly, and the development of efficient estimation techniques plays a critical role in these applications. All of the error reduction techniques developed in this work are applied to model problems. It is found that the nearly optimal parameters selected by the analytic method for use with GWAN estimator are nearly identical to parameters selected by the multistage method. Modified path length estimation (based on the path length importance measure) leads to excellent error reduction in all model problems examined. Finally, it should be pointed out that techniques used for neutron transport problems may be transferred easily to other application areas which are based on random walk processes. The transport problems studied in this dissertation provide exceptionally severe tests of the error reduction potential of any sampling procedure. It is therefore expected that the methods of this dissertation will prove useful in many other application areas.
- Research Organization:
- Claremont Graduate School, CA (USA). Dept. of Mathematics
- OSTI ID:
- 5226583
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
Periodic orbits of hybrid systems and parameter estimation via AD.
Preface: Nonclassical Transport