Statistical Tests for Diagnosing Fission Source Convergence and Undersampling in Monte Carlo Criticality Calculations

Adobe Acrobat PDF Document - on 2/17/23 at 11:58 AM
- Adobe Acrobat PDF Document on 2/17/23 at 10:40 AM

Monte Carlo criticality calculations involve a two-step procedure based on a stochastic implementation of the power iteration method: first, achieve convergence of the fission source distribution during the “inactive cycles”, and then sample from this source during the “active cycles”. Both phases involve distinct challenges in order to avoid potential issues related to lack of convergence (for inactive cycles) and undersampling and/or clustering (for active cycles). Over the past several years, under the auspices of the Nuclear Energy Agency (NEA), the Expert Group on Advanced Monte Carlo Techniques (EGAMCT) of the Working Party on Nuclear Criticality Safety (WPNCS) investigated the phenomena of clustering and undersampling in Monte Carlo critcality calculations. A previous Expert Group on Monte Carlo Source Convergence developed breakthrough methods for graphically assessing the initial convergence of the Monte Carlo fission source, using Shannon entropy or Brownian bridge metrics. Much was accomplished in understanding these phenomena, from both theoretical and practical approaches. Those efforts have led to a number of  ideas and challenges for new subgroup study topics. The WPNCS Subgroup 6 (SG-6), which ran during 2019-2020, is a direct follow-on to those previous efforts to improve the understanding of and capabilities for Monte Carlo criticality calculations.