On-the-Fly Diagnostics of Particle Population in Iterated-Source Monte Carlo Methods

On-the-fly diagnostics of the number of particles per iteration cycle in iterated-source Monte Carlo computation, which output diagnostic measures for a given spatial resolution of binning cells as iteration cycles progress, have been studied using relative entropy and chi-square distance. A source ratio vector is defined whose components are the ratio of the sources of adjacent iteration cycles at the individual binning cells. This enables one to define a problem-independent reference vector based on the integral equation representation of the static eigenvalue problem of particle multiplication. These vectors are normalized so that they represent discrete probability distribution. The relative entropy of the source ratio vector and the weighted difference between the relative entropy and the chi-square distance of the source ratio vector, all with respect to the reference vector, have been shown to be effective measures of particle population. Numerical examples presented include the initial core of a pressurized water reactor (PWR), the vault of PWR fresh fuel bundles, and the Whitesides' k_eff-of-the-world problem.