Alternative implementations of the Monte Carlo power method.

Blomquist, R.N. ; Gelbard, E.M.

Argonne National Laboratory

关键词: Criticality;    99 General And Miscellaneous//Mathematics, Computing, And Information Science;    Fission;    Efficiency;    Monte Carlo Method;   

DOI  :  10.2172/793906
RP-ID  :  ANL-01/15
RP-ID  :  W-31-109-ENG-38
RP-ID  :  793906

美国|英语

来源: UNT Digital Library

【 摘 要 】

We compare nominal efficiencies, i.e. variances in power shapes for equal running time, of different versions of the Monte Carlo eigenvalue computation, as applied to criticality safety analysis calculations. The two main methods considered here are ''conventional'' Monte Carlo and the superhistory method, and both are used in criticality safety codes. Within each of these major methods, different variants are available for the main steps of the basic Monte Carlo algorithm. Thus, for example, different treatments of the fission process may vary in the extent to which they follow, in analog fashion, the details of real-world fission, or may vary in details of the methods by which they choose next-generation source sites. In general the same options are available in both the superhistory method and conventional Monte Carlo, but there seems not to have been much examination of the special properties of the two major methods and their minor variants. We find, first, that the superhistory method is just as efficient as conventional Monte Carlo and, secondly, that use of different variants of the basic algorithms may, in special cases, have a surprisingly large effect on Monte Carlo computational efficiency.