Initialization bias suppression of an iterative MONTE CARLO Calculation
RICHET Yann - JACQUET Olivier - BAY Xavier - The Monte Carlo Method: Versatility Unbounded In A Dynamic Computing World Chattanooga, Tennessee, April 17–21, 2005, on CD-ROM, American Nuclear Society, LaGrange Park, IL
The accuracy of an Iterative Monte Carlo (IMC) calculation requires the convergence of the simulation output process. The present paper deals with a post processing algorithm to suppress the transient due to initialization using the Brownian bridge theory. It should be noticed that the initial transient suppression aims only at obtaining a stationary output series and cannot guarantee the convergence of the calculation in any way. As a consequence, it is necessary to run a sufficient number of iterations to ensure that the Iterative Monte Carlo calculation has converged. The simulation output process now standing for a converged seriesx – is tested as a stationary process. Basically, statistical tests inspired from Brownian bridge literature are performed at first on the entire original process x. If the test is positive, x is considered as stationary, otherwise the first observations of x are deleted and the test is performed again. This truncation procedure is repeated as long as the test is negative. This paper compares the efficiency of this iterative truncation procedure for different statistical tests and different first observations suppressions on experiment plans. Finally the most efficient truncation method is applied and evaluated on some criticality MC calculations.