Algorithm Groundwork
Implementation of Monte Carlo algorithms often requires some preparatory work or analysis.
We refer to this step as “groundwork.” In case of the algorithm for calculating an integral, at this step we choose a function \\(g\\left (x\\right )\\).
The function must be normalized and nonnegative (conditions A and B).
It also should be such that sampling fr.
,
Algorithm
1.
Generate a sample {ξ1, ξ2, …, ξN }, where the distribution of ξi is \\(g\\left (x\\right )\\); N≫ 1
,
Can statistical ap- proach be used to solve neutron diffusion and multiplication problems?
He had concluded that “the statistical ap- proach is very well suited to a digital treatment,” and he outlined in some de- tail how this method could be used to solve neutron diffusion and multiplica- tion problems in fission devices for the case “of ‘inert’ criticality” (that is, ap- proximated as momentarily static config- Fig. 1.
,
Estimator
According to Eq. (1.9), we need to estimate an expectation value.
We can use for that purpose the sample average provided that g(x) is chosen so that the ratio f(x)∕g(x) is finite for all x ∈ [a, b].
,
Example 1
Calculate the integral We choose and ensure that conditions A and B are satisfied: We also check that
,
How to determine thermal neutron diffusion length?
The thermal neutron diffusion length can be determined experimentally by mea- suring the axial neutron flux distribution in a long (with respect to mean free path) block of material with an isotropic thermal neutron flux incident on one end (e.g., fromthethermalcolumnofareactor).WithreferencetoFig.3.2,considerarectan- gularparallelepipedoflength c .
,
How to solve multi-group neutron diffusion equation?
Various kinds of numerical methods have been developed to solve the multi-group neutron diffusion equation such as:
- finite difference (FD)
- finite element (FE) and nodal methods (NM)
The nodal methods are more accurate, as well as, fast compared to FD for a given node or mesh structure.
,
Optimization
Here we present the importance sampling optimization technique.
We need to find a function g(x) that satisfies conditions A and B, and minimizes the variance of the ratio f(ξ)∕g(ξ).
The variance is In the second integral, g(x) is cancelled out, which means that we need to minimize the first integral To find the minimum of this integral, we write an.
,
Sampling The Distribution
We need to sample ξ from distribution g(x).
Given that g(x) can be any function that satisfies conditions A and B, it should be chosen so that the sampling algorithm is simple and fast.
The easiest to sample distribution is a uniform distribution in interval [a, b].
In that case g(x) is a constant.
However, we have not yet discussed optimization of.
,
Statistical Interpretation
Let g(x) be an arbitrary function that satisfies conditions A and B We also set g(x) = 0 for x outside the integration interval, [a, b].
Then g(x) can be interpreted as a probability density function of a random variable ξ, and the integral can be interpreted as the expectation value Again, ξ is a random number with distribution \\(g\\left (x\\right ).
,
What is the statistical structure of diffusion data?
In the method of diffusion the statistical structure of which leads to its redundancy is “dissipated” into long range statistics—i.e., 708 .