2026/07/30 by Xuran Yang, Liangzhi Cao, Qi Zheng +5
Engineering · Physics and Astronomy · #Nuclear Physics and Applications #Nuclear physics research studies #Nuclear reactor physics and engineering
paper · doi:10.1080/00295639.2026.2704151
crossref issued 2026/07/30 · crossref published 2026/07/30 · crossref published-online 2026/07/30 · openalex publication_date 2026/07/30 · crossref created 2026/07/30 · crossref deposited 2026/07/30 · crossref indexed 2026/07/30 · openalex created_date 2026/07/31 · openalex updated_date 2026/08/01
Focusing on the comparison and engineering application of direct and Iterative K-Source (IKS) Monte Carlo methods within flight-based (FB) and collision-Based (CB) frameworks for frequency-domain neutron noise, this study addresses the core challenges of numerical instability (i. e., particle explosion) and excessive computational cost in such simulations. To complement previous studies and ensure the integrity of this work, explicit expressions for the frequency-dependent numerical multiplication factor K(ω) were derived from linearized neutron transport and kinetic equations. The corresponding Monte Carlo power iteration module for K(ω) calculation is implemented and validated in the NECP-MCX code via an infinite homogeneous medium problem and then applied on the IAEA-3D pressurized water reactor benchmark. The IKS method was extended to the FB framework for the first time, forming four solution methods (FB-direct, CB-direct, FB-IKS, and CB-IKS). Quantitative tests on the IAEA-3D benchmark evaluated their computational time, figure of merit, and accuracy across 10−2 to 105 Hz with different CB hyperparameter eta values. Results show the FB method maintains K(ω)<1.0 across all frequencies, while the CB method becomes supercritical at high frequencies with fixed η, requiring careful η optimization. All four methods agree well in accuracy at 0.01 to 10 Hz; IKS methods boost efficiency significantly, with FB-IKS performing optimally at high frequencies. Engineering guidelines for frequency-dependent method selection are proposed, providing practical support for large-scale neutron noise database construction and data-driven reactor fault diagnosis.