American Nuclear Society
Home

Home / Publications / Journals / Nuclear Science and Engineering / Volume 163 / Number 2

Step-Refined On-the-Fly Convergence Diagnostics of Monte Carlo Fission Source Distribution

Taro Ueki

Nuclear Science and Engineering / Volume 163 / Number 2 / October 2009 / Pages 99-117

Technical Paper / dx.doi.org/10.13182/NSE163-99

Step-refined on-the-fly convergence diagnostics have been studied for Monte Carlo fission source distribution. The diagnostics consist of progressive relative entropy, its differenced series, and the Wilcoxon signed rank sum. The new aspect in the present work is the examination of a smaller number of cycles preceding the current cycle at step 1 diagnosis and the increase of the number of cycles for examination at step 2 diagnosis. This refinement process proceeds along with the actual progression of cycles and continues until the number of cycles for examination becomes >66% of the declared convergence cycle or the declaration of convergence stays in the same cycle for five consecutive steps. In each diagnostic step the Wilcoxon signed rank sum of the difference series of progressive relative entropy is examined to check if the downward crossing of the median of the range of the Wilcoxon rank sum occurs. The feasibility test was conducted for the fresh fuel vault of a pressurized water reactor (PWR) with the checkerboard placement of fuel bundle units and water-filled units, the critical sphere known as Godiva, the three-dimensional whole-core model of a PWR, and the Whitesides' k-effective of the world problem. The performance was observed to be consistent with or more conservative than that of the posterior relative entropy diagnosis. The methodology can potentially be extended to problems in need of 5000 cycles or more until convergence. A safeguard different from posterior diagnostics is proposed.