Overview
Feynman-α analysis employs the bunching technique to synthesize neutron count data by aggregating counts across hierarchical bin widths. For each bin size T, the variance-to-mean ratio Y(T) is computed to determine the α parameter. The bunching process introduces correlations among Y(T) points that are systematically neglected by conventional uncorrelated fitting methods, resulting in biased uncertainty estimates for α. The fundamental challenge lies in accurately estimating the covariance matrix of Y(T) from empirical measurements within practical constraints on data acquisition and computational resources. This investigation addresses the gap between theoretical requirements for proper fitting and practical limitations in covariance matrix estimation.
Methods and approach
The research employs synthetic data validation to compare fitting approaches with and without covariance matrix incorporation. Two key refinements are introduced: thinning and batching procedures that substantially reduce the data volume required for reliable covariance estimation. The proposed snippet-based algorithm partitions measurement sequences into manageable subsets, enabling systematic covariance computation without exhaustive full-dataset processing. The method is evaluated against theoretical expectations and benchmarked for feasibility within reactor noise measurement constraints, specifically examining the data requirements and computational overhead necessary to achieve stable covariance estimates.
Key Findings
Uncorrelated fitting methods systematically underestimate uncertainties in the α parameter because they disregard correlations inherent to the bunching process. When an accurately estimated covariance matrix is incorporated into the fitting procedure, α estimates and their associated uncertainties align with theoretical expectations. The snippet-based algorithm demonstrates that approximately 200 independent snippets suffice to generate stable and reliable covariance estimates under typical reactor noise measurement conditions. Thinning and batching strategies reduce computational burden and data requirements by orders of magnitude compared to direct covariance matrix estimation from complete datasets, rendering the approach feasible for practical applications.
Implications
The findings validate the theoretical requirements for proper treatment of correlations in Feynman-α analysis and establish that neglecting the covariance structure introduces systematic bias in uncertainty quantification. The snippet-based algorithm provides a computationally tractable framework for practitioners to obtain accurate covariance information within realistic measurement and computational constraints. These results reinforce the necessity of covariance-aware fitting methods in nuclear reactor noise analysis and extend the practical applicability of Feynman-α techniques to diverse measurement scenarios where data acquisition may be resource-limited.
Disclosure
- Research title: A snippet-based algorithm for practical covariance estimation in Feynman- α analysis
- Authors: Tom Drechsler, S. Weichel, Antonio Hurtado, Carsten Lange
- Publication date: 2026-03-06
- DOI: https://doi.org/10.1016/j.anucene.2026.112241
- OpenAlex record: View
- Disclosure: This post was generated by Claude (Anthropic). The original authors did not write or review this post.